UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Developing pyrometric and chemiluminescence optical diagnostics for investigation of modern alternative… Khosravi, Mahdiar 2019

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2019_september_khosravi_mahdiar.pdf [ 28.81MB ]
JSON: 24-1.0379727.json
JSON-LD: 24-1.0379727-ld.json
RDF/XML (Pretty): 24-1.0379727-rdf.xml
RDF/JSON: 24-1.0379727-rdf.json
Turtle: 24-1.0379727-turtle.txt
N-Triples: 24-1.0379727-rdf-ntriples.txt
Original Record: 24-1.0379727-source.json
Full Text

Full Text

Developing Pyrometric and Chemiluminescence OpticalDiagnostics for Investigation of Modern Alternative CIEngine Combustion StrategiesbyMahdiar KhosraviB.A.Sc., K.N.Toosi University of Technology, 2008M.A.Sc., University of Tehran, 2011A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFDoctor of PhilosophyinTHE FACULTY OF GRADUATE AND POSTDOCTORALSTUDIES(Mechanical Engineering)The University of British Columbia(Vancouver)June 2019© Mahdiar Khosravi, 2019The following individuals certify that they have read, and recommend to the Fac-ulty of Graduate and Postdoctoral Studies for acceptance, the dissertation entitled:Developing Pyrometric and Chemiluminescence Optical Diagnostics forInvestigation of Modern Alternative CI Engine Combustion Strategiessubmitted by Mahdiar Khosravi in partial fulfillment of the requirementsfor the degree of Doctor of Philosophyin Mechanical EngineeringExamining Committee:Patrick Kirchen, Mechanical EngineeringSupervisorKendal Bushe, Mechanical EngineeringSupervisory Committee MemberMartti Larmi, Mechanical EngineeringExternal ExaminerXiaotao Bi, Chemical and Biological EngineeringUniversity ExaminerDana Grecov, Mechanical EngineeringUniversity ExamineriiAbstractIts inherent economic and environmental advantages as a compression ignition (CI)engine fuel make natural gas (NG) an attractive alternative to diesel fuel. Lim-ited optical studies of the NG combustion strategies have been reported in litera-ture. The current work focused on developing optical characterization techniquesto study in-cylinder processes in cleaner combustion strategies, such as those in-volving natural gas. An experimental facility supporting optical diagnostics via aBowditch piston arrangement in a 2-litre, single-cylinder research engine was usedin this study.In order to facilitate quantitative soot analysis for low soot combustion strate-gies, the performance of the pyrometric method was improved by nearly 40% in-crease in the resolved signal fraction through modifications in numerical algorithm,calibration and implementation of the method, and image processing. The en-hanced pyrometry method was implemented simultaneously with high-speed OH*chemiluminescence imaging to pilot-ignited direct-injected natural gas (PIDING)combustion for the first time. The results revealed that a standard PIDING op-eration can be characterized by low-sooting non-premixed combustion of the NGalong the jet axes and of a partially-premixed charge at the wall region, followedby onset of detectable soot at the points of NG jet impingement on the bowl wall.This results in formation of a soot cloud adjacent to the wall, which then growstowards the center with continued soot formation and reflected momentum of theNG jets impinging on the bowl wall. The relative timing between NG injectionpulse and peak HRR, and the Pin j, showed strong influence in the rate and extentof soot formation and peak concentration levels.Rugged probe designs afford optical measurements from all-metal engines.iiiComparisons between 2D and probe based 0D pyrometry measurements were madeunder optical engine configuration, for the first time, to better characterize the 0Dprobe signal. The 2D and 0D results showed reasonable agreements, especiallywhen field-of-view geometry differences were taken into account. 0D two-colorpyrometry measurements in an all-metal engine led to similar conclusions on sootin-cylinder processes, albeit with signs of enhanced late-cycle soot oxidation, at-tributed to the conventional omega shaped piston bowl geometry.ivLay SummaryThe increasingly strict emission regulations for heavy-duty diesel truck engineshave driven the development of new engine technologies towards reducing green-house gas and soot emissions. These technologies possess fundamentally differentenergy production and pollutant formation characteristics, and dedicated researchto understand these phenomena is critical for optimization. High-speed imagingsystems were developed to investigate the combustion processes inside an opti-cal engine and improvements were identified to an optical diagnostic technique tofacilitate its application for low-sooting combustion conditions. The method wasimplemented to a direct-injected diesel-natural gas combustion strategy to evalu-ate soot concentration and temperature distributions. Based on these results andother simultaneous diagnostics, a conceptual understanding of the soot processesand the influence of the operating parameters thereof was proposed. In addition,the developed imaging tools were used to characterize the signal from a ruggedoptical probe design, which can be implemented for optical measurements in less-idealized all-metal engines.vPrefaceThe original vision of this research work with regards to the implementation oftwo-color pyrometry to natural gas combustion strategies in compression ignitions,as well as the thermo-optical methodology was developed by my supervisor Dr.Patrick Kirchen. However, the direction I took with regards to the design of thesimultaneous two-color pyrometry and OH* chemiluminescence imaging system,and analysis algorithms, were largely my own undertaking along with the imple-mentation of the developed imaging tools to the PIDING combustion strategy inthe optical engine.Commissioning of the single-cylinder research engine facility was a collab-orative effort involving several people. Jeremy Rochussen designed and com-missioned the main mechanical, fuel, and safety systems for the Ricardo Proteussingle-cylinder research engine. Jeff Yeo commissioned both the hardware andsoftware associated with engine control and data acquisition systems. I was re-sponsible for commissioning of the high-speed optical imaging systems.Optical engine tests were a combined effort between Jeremy Rochussen, JeffYeo, and myself. I was solely responsible for the calibration of the pyromet-ric imaging system and post-processing of all high-speed imaging results, whileJeremy Rochussen and Jeff Yeo completed processing thermodynamic and probedata, respectively. Jeff Yeo wrote the scripts to process the probe data based on thedeveloped imaging pyrometric algorithm, and I modified these to integrate themwith the imaging results.All data analysis presented in this thesis is my original work, with the exceptionof the routines used to process the thermodynamic results, including heat releaserates and fuel mass flow rates, which was developed by Jeremy Rochussen.viSimultaneous natural luminosity and OH* chemiluminescence imaging analy-sis of DIDF combustion discussed in Chapter 4 of this work has been presented andpublished as part of the 2016 Combustion Institute Canadian Section Conferenceand 2016 ASME-Internal Combustion Engine Conference. The enhancement tothe two-color pyrometry algorithm, described in Chapter 5 has been published atthe Journal of Automobile Engineering (Proceedings of the Institution of Mechan-ical Engineers, Part D). The Pyrometric imaging investigation of soot processesPIDING combustion, presented in Chapter 6, was ready for submission at the timeof writing this document. I was the lead author for these works and performed allwritten and analytical work, while received input and assistance from co-authorsJeremy Rochussen, Jeff Yeo, Dr. Patrick Kirchen, and Drs. Gordon McTaggart-Cowan and Ning Wu:• M. Khosravi, J. Rochussen, J. Yeo, and P. Kirchen, “Characterization ofDiesel-Ignited Dual-Fuel Combustion in an Optical Engine – Part I: Effectsof Fueling Parameters on Flame Structure”, Proceedings of the CICS Con-ference, 2016.• M. Khosravi, J. Rochussen, J. Yeo, P. Kirchen, G. McTaggart-Cowan, andN. Wu, “Effect of Fuelling Control Parameters on Combustion Characteris-tics of Diesel-Ignited Natural Gas Dual-Fuel Combustion in an Optical En-gine”, ASME-Internal Combustion Engine Division Fall Technical Confer-ence, 2016.• M. Khosravi, P. Kirchen, “Refinement of the Two-Color Method for Appli-cation in a Direct Injection Diesel and Natural Gas Compression IgnitionEngine”, Proc. Inst. Mech. Eng. D J. Automob. Eng., 2019.• M. Khosravi, P. Kirchen, “Pyrometric Imaging of Soot Processes in a PilotIgnited Direct Injected Natural Gas Engine”, under review.In addition to the works listed above I co-authored the following papers, wheremy contribution was primarily in the form of rebuilding the optical engine andrunning optical tests with Jeremy Rochussen and Jeff Yeo, as well as capturing andpost-processing the imaging data:vii• J. Rochussen, J. Yeo, M. Khosravi, P. Kirchen, “Investigation of Pilot-IgnitedDual-Fuel Natural Gas Combustion in an Optically Accessible Engine”, Pro-ceedings of the CICS Conference, 2015.• J. Rochussen, M. Khosravi, J. Yeo, P. Kirchen, “Characterization of Diesel-Ignited Dual-Fuel Combustion in an Optical Engine – Part II: Optical andThermodynamic Comparison”, Proceedings of the CICS Conference, 2016.• J. Yeo, M. Khosravi, J. Rochussen, P. Kirchen, “Application of an In-CylinderLine of Sight Two-Colour Pyrometry Probe in an Optical Pilot-Ignited Di-rect Injection Natural Gas Engine”, Proceedings of the CICS Conference,2017.• J. Rochussen, J. Son, J. Yeo, M. Khosravi, P. Kirchen, G. McTaggart-Cowan,“Development of a Research-Oriented Cylinder Head with Modular InjectorMounting and Access for Multiple In-Cylinder Diagnostics”, SAE Interna-tional Conference on Engines & Vehicles; 2017-24-0044, 2017.viiiContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viContents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xvList of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiList of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiv1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Role of Optical Diagnostics in IC Engine Development . . . . . . 11.1.1 Optical Diagnostics . . . . . . . . . . . . . . . . . . . . . 11.1.2 Diesel Engines . . . . . . . . . . . . . . . . . . . . . . . 41.1.3 Alternative Fuels for CI engines . . . . . . . . . . . . . . 61.1.4 Air Pollution . . . . . . . . . . . . . . . . . . . . . . . . 81.1.5 Developing Technologies . . . . . . . . . . . . . . . . . . 111.2 Objectives, Contributions, and Approaches . . . . . . . . . . . . . 12ix2 Background Information and Literature Review . . . . . . . . . . . 152.1 Optical Diagnostics for Combustion In-Cylinder Processes . . . . 152.1.1 Optical Access to the Combustion Chamber . . . . . . . . 162.1.2 Natural Luminosity Imaging . . . . . . . . . . . . . . . . 202.1.3 Chemiluminescence Imaging . . . . . . . . . . . . . . . . 212.1.4 Two-Color Pyrometry . . . . . . . . . . . . . . . . . . . 262.2 Modern Alternative Combustion Strategies . . . . . . . . . . . . . 362.2.1 DIDF . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362.2.2 PIDING . . . . . . . . . . . . . . . . . . . . . . . . . . . 402.3 Summary and Literature Gap . . . . . . . . . . . . . . . . . . . . 433 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . 473.1 Engine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473.1.1 Engine Facility Instrumentation . . . . . . . . . . . . . . 493.2 Imaging Systems . . . . . . . . . . . . . . . . . . . . . . . . . . 523.2.1 Natural Luminosity Imaging . . . . . . . . . . . . . . . . 543.2.2 Two-Color Pyrometry Imaging . . . . . . . . . . . . . . . 543.2.3 OH* Chemiluminescence Imaging . . . . . . . . . . . . . 593.2.4 Spectral Optics . . . . . . . . . . . . . . . . . . . . . . . 613.2.5 Optical Probe . . . . . . . . . . . . . . . . . . . . . . . . 633.3 Integrated System Operating Protocols . . . . . . . . . . . . . . . 663.3.1 Synchronization of optical diagnostics with the engine . . 673.3.2 Alignment and Calibration . . . . . . . . . . . . . . . . . 693.3.3 Optical engine operation . . . . . . . . . . . . . . . . . . 714 Optical Investigation of DIDF combustion . . . . . . . . . . . . . . . 734.1 Analysis Metrics and Engine Operating Conditions . . . . . . . . 734.2 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . 784.2.1 Effect of the Pilot Fuel Injection Pressure . . . . . . . . . 784.2.2 Effect of the PR and φCH4 . . . . . . . . . . . . . . . . . . 834.2.3 Influence of Diesel and Natural Gas on Each Other in Dual-Fuel Combustion Charge . . . . . . . . . . . . . . . . . . 884.3 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . 89x5 Two-Color Pyrometry Method Improvement . . . . . . . . . . . . . 925.1 Method Enhancement . . . . . . . . . . . . . . . . . . . . . . . . 925.1.1 Algorithm Selection . . . . . . . . . . . . . . . . . . . . 935.1.2 Experimental Considerations . . . . . . . . . . . . . . . . 985.1.3 System Calibration and Image Processing . . . . . . . . . 1005.2 Sample Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 1115.3 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . 1156 Optical Investigation of PIDING Combustion . . . . . . . . . . . . . 1176.1 Engine Operating Conditions and Analysis Metrics . . . . . . . . 1176.1.1 Analysis Metrics . . . . . . . . . . . . . . . . . . . . . . 1206.2 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . 1226.2.1 Baseline Operating Condition . . . . . . . . . . . . . . . 1236.2.2 Fuel mass (injection Duration) Effect . . . . . . . . . . . 1296.2.3 Relative Injection Timing Effect . . . . . . . . . . . . . . 1336.2.4 Injection Pressure Effect . . . . . . . . . . . . . . . . . . 1396.2.5 Updated Understanding of Soot Processes during PIDINGCombustion . . . . . . . . . . . . . . . . . . . . . . . . . 1416.3 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . 1627 Optical Probe Signal Characterization . . . . . . . . . . . . . . . . . 1657.1 2D vs. 0D two-color pyrometry measurements . . . . . . . . . . . 1667.1.1 The FOV effects . . . . . . . . . . . . . . . . . . . . . . 1687.1.2 Weighted function averaging . . . . . . . . . . . . . . . . 1707.2 Thermodynamic vs. optical engine configurations . . . . . . . . . 1777.3 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . 1828 Conclusions and Future Work . . . . . . . . . . . . . . . . . . . . . 1848.1 Summary of the Significant Findings . . . . . . . . . . . . . . . . 1848.1.1 Optical Investigation of DIDF combustion . . . . . . . . . 1858.1.2 Two-Color Pyrometry Method Improvement . . . . . . . 1868.1.3 Optical Investigation of PIDING Combustion . . . . . . . 1888.1.4 Thermo-Optical Analysis Tool-Kit . . . . . . . . . . . . . 1888.1.5 Concluding Remarks . . . . . . . . . . . . . . . . . . . . 191xi8.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1928.2.1 2D optical DIDF combustion analysis . . . . . . . . . . . 1928.2.2 2D optical PIDING combustion analysis . . . . . . . . . . 1938.2.3 Exhaust Gas recirculation (EGR) . . . . . . . . . . . . . . 1948.2.4 Optical probe 0D signal characterization . . . . . . . . . . 1948.2.5 Linking in-cylinder to engine-out soot measurements . . . 196Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198A High-Speed Imaging System Considerations . . . . . . . . . . . . . 215B Ensemble Averaging of Results vs. Raw signal . . . . . . . . . . . . 217C PIDING Fuel Injection Profile . . . . . . . . . . . . . . . . . . . . . 219D Optical and Thermodynamic Ignition Delays . . . . . . . . . . . . . 221E NG Jet Momentum Reflection Effects . . . . . . . . . . . . . . . . . 223F Mean Soot Cloud Growth Rate . . . . . . . . . . . . . . . . . . . . . 225G Ensemble Averaging of Results vs. Raw signal . . . . . . . . . . . . 226H Radiation Heat Loss through Quartz Window . . . . . . . . . . . . 228xiiList of TablesTable 1.1 Euro standards for heavy-duty vehicles . . . . . . . . . . . . . 11Table 2.1 Formation pathways and characteristic wavelengths of majorchemiluminescence sources in combustion of hydrocarbon fuels 22Table 2.2 Summary of reported two-color pyrometry uncertainties . . . . 32Table 3.1 Optical engine specifications . . . . . . . . . . . . . . . . . . 48Table 3.2 Engine sensors used in combustion analysis. . . . . . . . . . . 49Table 3.3 Optical Imaging Equipment . . . . . . . . . . . . . . . . . . . 53Table 3.4 Imaging Systems Assemblies . . . . . . . . . . . . . . . . . . 54Table 4.1 Operating points for investigation of DIDF combustion . . . . 75Table 4.2 Simultaneous OH* and NL imaging system setting for opticalinvestigation of DIDF combustion . . . . . . . . . . . . . . . . 76Table 4.3 Ignition delay and combustion duration. . . . . . . . . . . . . 87Table 5.1 Performance of considered numerical algorithms . . . . . . . . 96Table 5.2 Optical engine and imaging system specifications for two-colorpyrometry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99Table 5.3 Operating parameters for diesel and PIDING combustion. . . . 100Table 6.1 Operating points for investigation of PIDING combustion . . . 118Table 6.2 Imaging system specifications . . . . . . . . . . . . . . . . . . 119Table 7.1 Operating points discussed in Figure 7.3; RIT = 8 CAD, Pin j=180 bar, speed=1000 rpm for all points. . . . . . . . . . . . . . 169xiiiTable 7.2 Operating conditions investigated under thermodynamic and op-tical engine configurations using 0D optical probe measurements;Engine speed was 1000 rpm for all points. . . . . . . . . . . . 178Table D.1 Comparison of optical and thermodynamic ignition delays foroperating points described in Table 6.1 . . . . . . . . . . . . . 222xivList of FiguresFigure 2.1 Schematic of a sample optical probe light collection diagram . 17Figure 2.2 Schematic of a sample endoscopic light collection diagram . . 18Figure 2.3 Schematic and photo of the optical configuration of the engine 19Figure 2.4 Shadow and NL images of a single spray flame in a constantvolume combustion chamber . . . . . . . . . . . . . . . . . . 33Figure 2.5 fvL, HRR, NL, soot volume, and NL images for high-temperaturediesel operating condition . . . . . . . . . . . . . . . . . . . 34Figure 2.6 Sample 2D two-color pyrometry results and the correspondingfiltered NL images . . . . . . . . . . . . . . . . . . . . . . . 36Figure 2.7 OH*-chemiluminescence image sequences for high-load andlow-load DIDF operation in an RCEM . . . . . . . . . . . . . 39Figure 2.8 OH*-chemiluminescence image sequences for low-load andhigh-load DIDF operation with constant pilot mass . . . . . . 40Figure 2.9 Combined 15-cycle ensemble averaged 2D images of OH* chemi-luminescence and NL from PIDING combustion . . . . . . . 42Figure 3.1 Experimental facility process and instrumentation diagram forPIDING combustion experiments under optical configuration. 51Figure 3.2 Ricardo Proteus engine schematic and imaging system setupon the optical engine . . . . . . . . . . . . . . . . . . . . . . 52Figure 3.3 Image doubler design and two-color pyrometry system schematic. 55Figure 3.4 Chemiluminescence Imaging system setup . . . . . . . . . . 56Figure 3.5 The two-color pyrometry calibration setup . . . . . . . . . . . 57Figure 3.6 Response calibration with camera exposure vs. ND filters . . . 59xvFigure 3.7 Chemiluminescence Imaging system setup . . . . . . . . . . 61Figure 3.8 Transmission curves from the various optical filters used in thiswork . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63Figure 3.9 The engine head design for multiple in-cylinder diagnostics . 64Figure 3.10 Optical probe design used in this work . . . . . . . . . . . . . 65Figure 3.11 Optical probe assembly . . . . . . . . . . . . . . . . . . . . . 66Figure 3.12 Optical probe calibration system. . . . . . . . . . . . . . . . . 66Figure 3.13 Simplified wiring diagram of the integrated optical systems. . 69Figure 3.14 Adjustable optical table design . . . . . . . . . . . . . . . . . 70Figure 3.15 Standard optical engine operation skip-firing scheme . . . . . 71Figure 3.16 Optical measurements protocol flowchart. . . . . . . . . . . . 72Figure 4.1 Schematic of the simultaneous OH* and NL imaging systemused for optical investigation of DIDF combustion. . . . . . . 77Figure 4.2 Sample image series of ensemble averaged NL and OH* chemi-luminescence . . . . . . . . . . . . . . . . . . . . . . . . . . 79Figure 4.3 Overlaid ensemble averaged OH* chemiluminescence and nat-ural luminosity images . . . . . . . . . . . . . . . . . . . . . 80Figure 4.4 Overlaid ensemble averaged OH* chemiluminescence and nat-ural luminosity images for dual fuel operating points . . . . . 81Figure 4.5 Conceptual effect of pilot fuel injection pressure on reactionzone growth mechanism on DF combustion mode . . . . . . . 82Figure 4.6 Comparison of spatially averaged NL, OH* chemiluminescence,and HRR results . . . . . . . . . . . . . . . . . . . . . . . . 85Figure 4.7 OH* Chemiluminescence at the maximum bowl coverage. . . 86Figure 4.8 Reaction zones in dual-fuel and diesel-only combustion . . . . 89Figure 5.1 Soot cloud spectral emissivity vs. KL factor . . . . . . . . . . 94Figure 5.2 Residual error behavior around the root of two-color pyrome-try equation . . . . . . . . . . . . . . . . . . . . . . . . . . . 94Figure 5.3 Algorithm performance comparison and residual profile . . . 97Figure 5.4 Apparent temperature ratios map . . . . . . . . . . . . . . . . 98xviFigure 5.5 Two-color pyrometry imaging system implementation on theoptical engine. . . . . . . . . . . . . . . . . . . . . . . . . . 99Figure 5.6 Work flow to obtain temperature and KL distributions fromrecorded raw signal. . . . . . . . . . . . . . . . . . . . . . . 101Figure 5.7 Influence of linearized extrapolation on two-color pyrometryresults . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103Figure 5.8 Effect of spatial non-uniformities imaging system response . . 104Figure 5.9 Increase in detection envelope by the imaging system dynamicrange with matching aperture response. . . . . . . . . . . . . 105Figure 5.10 Parallax calibration target images . . . . . . . . . . . . . . . 106Figure 5.11 Impact of static and dynamic parallax corrections to increaseresolved pyrometric signal at different crank positions . . . . 107Figure 5.12 Comparison of pixel binning and filtering on calculated tem-perature fields . . . . . . . . . . . . . . . . . . . . . . . . . . 109Figure 5.13 Effects of individual and combined corrections on the resolvedimage data for the diesel operating mode . . . . . . . . . . . 110Figure 5.14 Temperature and KL distributions for diesel combustion usingproposed refinements . . . . . . . . . . . . . . . . . . . . . . 112Figure 5.15 Temperature and KL distributions for PIDING combustion us-ing proposed refinements . . . . . . . . . . . . . . . . . . . . 113Figure 5.16 Effects of individual and combined corrections on the resolvedimage data for PIDING combustion . . . . . . . . . . . . . . 114Figure 6.1 Simultaneous two-color pyrometry and OH* chemilumines-cence imaging system . . . . . . . . . . . . . . . . . . . . . 120Figure 6.2 Net soot formation and oxidation for baseline PIDING operat-ing point . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121Figure 6.3 Baseline PIDING operating point; Ignition and early stage com-bustion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123Figure 6.4 Baseline PIDING operating point; NG combustion and sootformation and oxidation . . . . . . . . . . . . . . . . . . . . 126xviiFigure 6.5 Area fraction distribution of two-color pyrometry soot temper-atures and KL factors at each crank position for the baselineoperating condition . . . . . . . . . . . . . . . . . . . . . . . 128Figure 6.6 Aggregate resolved temperature and KL factor cumulative dis-tribution for the baseline operating condition after 10 CAD aTDC129Figure 6.7 NG fuel mass effects at constant injection pressure; 2D OH*chemiluminescence and two-color pyrometry results . . . . . 130Figure 6.8 NG fuel mass effects at constant injection pressure; Spatiallyaveraged imaging and thermodynamic results . . . . . . . . . 131Figure 6.9 NG fuel mass effects at constant injection pressure; Two-colorpyrometry soot temperature and KL factor area fraction coverage132Figure 6.10 Relative injection timing effect; 2D OH* chemiluminescenceand two-color pyrometry results . . . . . . . . . . . . . . . . 134Figure 6.11 Relative injection timing effect; Spatially averaged imagingand thermodynamic results . . . . . . . . . . . . . . . . . . . 135Figure 6.12 Relative injection timing effect on pilot combustion . . . . . . 136Figure 6.13 High temporal resolution 2D OH* chemiluminescence and two-color pyrometry results for pilot combustion at RIT=2 CAD . 136Figure 6.14 Relative injection timing effects; Two-color pyrometry soottemperature and KL factor area fraction coverage . . . . . . . 137Figure 6.15 Fuel injection pressure effect; 2D OH* chemiluminescenceand two-color pyrometry results . . . . . . . . . . . . . . . . 138Figure 6.16 Fuel injection pressure effect; Spatially averaged imaging andthermodynamic results . . . . . . . . . . . . . . . . . . . . . 140Figure 6.17 Fuel injection pressure effects; Two-color pyrometry soot tem-perature and KL factor area fraction coverage . . . . . . . . . 141Figure 6.18 HRR and OH* chemiluminescence images of combustion re-action zones through a typical PIDING combustion event . . . 143Figure 6.19 Correlation between the peak HRR and detected onset of sootsignal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144Figure 6.20 Correlation between peak KL factor and NG EOI . . . . . . . 145Figure 6.21 Radial KL factor distribution time series for GPW=1.85 ms andPin j= 140 bar operating conditions . . . . . . . . . . . . . . . 148xviiiFigure 6.22 Normalized net soot formation rates for GPW and Pin j sweeps 150Figure 6.23 The influence of Pin j sweep on mean soot cloud growth rate . 151Figure 6.24 The effect of relative timing between onset of soot signal andcGSOI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153Figure 6.25 The influence of considered parameter sweeps on mean netsoot formation rates . . . . . . . . . . . . . . . . . . . . . . . 155Figure 6.26 Soot processes during a typical PIDING combustion event . . 157Figure 6.27 Baseline PIDING vs. Diesel combustion; 2D imagin results . 159Figure 6.28 Aggregate resolved temperature and KL factor cumulative dis-tribution for baseline PIDING and conventional diesel operat-ing conditions . . . . . . . . . . . . . . . . . . . . . . . . . . 160Figure 7.1 2D and 0D Two-color pyrometry measurement system on op-tical engine configuration. . . . . . . . . . . . . . . . . . . . 166Figure 7.2 Line-of-sight measurement effects in a partially transparentsoot cloud . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167Figure 7.3 0D probe vs. mean 2D two-color pyrometry results . . . . . . 169Figure 7.4 Generation of the weighted spatially averaged ROI for 2D two-color pyrometry . . . . . . . . . . . . . . . . . . . . . . . . . 171Figure 7.5 WSA ROI mask at three crank positions and the correspondingoptical probe cone projection cross-sections . . . . . . . . . . 172Figure 7.6 0D probe vs. WSA 2D two-color pyrometry results . . . . . . 174Figure 7.7 Probe FOV geometry effect on WSA results . . . . . . . . . . 175Figure 7.8 Two-color pyrometry temperature and KL factor populationmaps for the baseline operating condition at 5 and 8 CAD aTDC.176Figure 7.9 0D probe vs. WSA 2D two-color pyrometry results, and theircorresponding apparent temperatures and 2D KL factor resultsvs. probe FOV coverage during soot oxidation stage in thebaseline operating point . . . . . . . . . . . . . . . . . . . . 177Figure 7.10 0D probe two-color pyrometry and HRR results for the base-line operating condition under optical and thermodynamic en-gine configurations . . . . . . . . . . . . . . . . . . . . . . . 179xixFigure 7.11 0D probe two-color pyrometry and HRR results for Pin j=220 bar,Pin j=140 bar, and long RIT operating conditions under opticaland thermodynamic engine configurations . . . . . . . . . . . 181Figure 8.1 2D and 0D natural luminosity/two-color pyrometry and OH*chemiluminescence measurement system on optical engine con-figuration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185Figure 8.2 Conceptual effect of pilot fuel injection pressure on reactionzone growth mechanism on DF combustion mode . . . . . . . 186Figure 8.3 Two-color pyrometry method enhanced performance depictedas improvements in the resolved signal fraction throughout theensemble averaged cycle and at a sample crank position forPIDING combustion . . . . . . . . . . . . . . . . . . . . . . 187Figure 8.4 The four principal stages of soot processes during a typicalPIDING combustion event . . . . . . . . . . . . . . . . . . . 189Figure 8.5 WSA mask at 25 CAD aTDC and comparison of the WSA 2Dand 0D optical measurements for select operating points . . . 190Figure B.1 Effect of ensemble averaging on the two-color pyrometry tem-perature distributions instead of using it as a SNR enhancementapproach for the raw data . . . . . . . . . . . . . . . . . . . . 218Figure C.1 Estimated PIDING fuel injection profile . . . . . . . . . . . . 220Figure E.1 Soot cloud advection effects with reflected NG jet momentumafter impingement onto the walls . . . . . . . . . . . . . . . . 224Figure F.1 The influence of Pin j on 2D two-color pyrometry temperaturedistribution . . . . . . . . . . . . . . . . . . . . . . . . . . . 225Figure G.1 Alternative WSA procedures flowchart . . . . . . . . . . . . 227Figure G.2 Alternative WSA procedures; sample results . . . . . . . . . . 227Figure H.1 A simplified radiation heat loss model for comparison of thetwo engine configurations . . . . . . . . . . . . . . . . . . . 228xxList of SymbolsAλ soot volume fraction optical constantα Hottel & Broughton empirical soot constantC2 Diatomic carbonCA50 Crank angle corresponding to 50% heat releaseC6H6 BenzeneCH MethylidyneCH4 MethaneCHO Formyl radicalCH2O Formaldehyde radicalCO Carbon monoxideCO2 Carbon dioxideCN Cyanido radicalEinput Total fuel energy inputfv Soot volume factionh Planck’s constantIηth Indicated thermal efficiencyIλ Spectral radiant intensityK Absorption strength of the soot cloudκλ Absorption coefficientL Characteristic optical path lengthλ Wavelengthmdiesel Diesel mass flow rateNOx Nitrogen oxidesOH Hydroxyl radicalOH* Electronically excited hydroxyl radicalOH*-CL OH* ChemiluminescencePin j fuel injection pressureTa Apparent soot temperatureφCH4 Methane equivalence ratioφglobal Global equivalence ratioΘSoot Onset of soot signalτign,NG NG optical ignition delayτign,pi Pilot optical ignition delayτSoot Onset of soot signal delayxxiList of AbbreviationsAFR Air-Fuel-RatioaTDC after Top Dead CentrebTDC before Top Dead CentreCAD Crank Angle DegreeCCD Charged Couple DeviceCHTS Cumulative Histogram Time SeriesCI Compression IgnitionCMOS Complementary Metal Oxide SensorcGSOI command of Gas Start of InjectioncPSOI command of Pilot Start of InjectioncSOI command of Start of InjectionCR Compression RatioCRP Color-ratio pyrometryCWL Central WavelengthDAQ Data AcquisitionDRP Diesel Rail PressureDIDF Direct Injection Dual-FuelDISI Direct Injection Spark-IgnitedECU Engine Control UnitEGR Exhaust gas recirculationEOI End of InjectionFOV Field-of-ViewFPGA Field-Programmable Gate Arrayfps frames-per-secondFWHM Full Width Half MaximumGIMEP Gross Mean Effective PressureGPW Gas Pulse WidthHPDI High Pressure Direct InjectionHRR Heat Release RateHTC high-temperature combustionIC Internal CombustionIR InfraredIRO Intensified relay opticsLHV Lower Heating ValueLII Light Induced IncandescenceLIF Light Induced FluorescenceLOS Line of SightLTC Low Temperature CombustionxxiiNBP Narrow Band-PassND Neutral DensityNIR Near InfraredNL Natural LuminosityNG Natural GasOD Optical DensityPIV Particle Image VelocimetryPIDING Pilot-Ignited Direct-Injected Natural GasPR Pilot RatioRIT Relative Injection TimingPM Particulate matterRCEM Rapid Compression/Expansion MachineSI Spark IgnitedSNR Signal to Noise RatioSOI Start of InjectionTDC Top Dead CentertHC Total HydrocarbonTTL Transistor-Transistor LogicuHC Unburned HydrocarbonsUV Ultra-VioletVOC Volatile Organic CompoundsWSA Weighted Spatially AveragingxxiiiAcknowledgementsThe presented work would not have been possible without the financial supportfrom Westport Fuel Systems and collaboration of its resourceful personnel, in par-ticular Dr. Sandeep Munshi for his technical insight and encouraging remarks, Dr.James Saunders for sharing his experience in setting up an optical imaging setup,Dr. Gordon McTaggart-Cowan for his consistent involvement in progression ofthe work through our monthly meetings, and others including Drs. Ashish Singh,Ning Wu, and Jim Huang. In addition, I would like to acknowledge further finan-cial support provided by the Natural Sciences and Engineering Research Councilof Canada (NSERC) Collaborative Research and Development (CRD), the Cana-dian Foundation for Innovation (CFI), John Evans Leaders Fund (JELF), and theNSERC Discovery Grant Program. I would also like to thank Dr. Steven Rogak,whose scientific and engineering experience I always benefited from during ourrecurrent group meetings. I am also grateful to Oliver Terry for the design andfabrication of the optical table and to Bob Parry for his technical support.I would like the express my most sincere gratitude to Dr. Patrick Kirchen,my supervisor, whose unfaltering oversight, mentorship, and evident passion forresearch was undoubtedly a principal driving force throughout my research. I couldoften perceive him also as a colleague when he would join us in late evening huntsfor source of issues in the test cell, or spend hours discussing possible explanationsfor the perplexing problems.Of the pleasant memories I will keep from my days at UBC for the rest of mylife, are moments spent with my friends, in sharing happiness and laugh as wellas in frustration and cursing programming scripts or equipment. Among these,I had the honor of being lab mates and working on the same test facility withxxivJeremy Rochussen, Jeff Yeo, and Pooyan Kheirkhah, whose help and support Iinfinitely appreciate and find hard to be able to equally return. I would also liketo thank David Sommer, Rene Zepeda, Aditya Prakash-Singh, Michael Karpinski-Leydier, Jeff Son, Jeff Meiklejohn, and Miayan Yeremi, my other colleagues atClean Energy Research Center.Finally, I would like to thank my family for everything they have made possiblefor me to achieve to this date, especially my mother, to whom I will always remainin debt for the sacrifices she made, and my beloved wife, Banafsheh, to whomI owe my prevailed sanity throughout my research, joy of today, and hope for abrighter future.xxvI dedicate this thesis to my mother and my wife, for everything I have...xxviChapter 1IntroductionThe motivation for the presented work is elucidated in this chapter. The pressingneed to further limit internal combustion (IC) engine emissions through modernalternative combustion strategies is discussed and a brief history around applica-tion of optical diagnostics for combustion in-cylinder soot characterization is pre-sented. The major objectives of the current work and an overview of the approachesadapted to achieve those objectives are subsequently discussed.1.1 Role of Optical Diagnostics in IC EngineDevelopmentThe major focus of the work presented herein was developing and implementingoptical diagnostic tools to better understand combustion processes in reciprocatingengines. As such, in what follows a brief overview of the application of opticaldiagnostics in engines is provided, along with some general background informa-tion regarding these applications. This is followed by an overview of compressionengine operating strategies, with a focus on the associated pollutant emissions andthe health and environmental effects thereof.1.1.1 Optical DiagnosticsThe scientific and technological advancements in imaging applications and manu-facturing imaging systems have both come a long way since the earliest surviving1recording of an image in mid-1820s, where many hours or even days of exposurein the camera were required to successfully record the image of a bright landscape.Improvements of the optical diagnostic tools ever since has led to breakthroughs inhealth and cognitive sciences, with the likes of magnetic resonance imaging (MRI)and various computed tomography (CT) scan techniques and shaped our under-standing of the universe through the likes of NASA’s Great Observatories programseries of telescopes (Hubble, Compton, Chandra, and Spitzer) and ESA’s largestever launched Herschel infrared telescope. Today, optical diagnostics has becomean indispensable component of modern non-intrusive and nondestructive test meth-ods and inspection techniques. Optical techniques play an extremely importantrole in wide variety of research and industrial applications, such as biomedical andbiomechanical research, military applications and forensics, ballistics, astronomy,industrial inspection, analytical chemistry, fluid dynamics, and combustion.The extensively developed optical techniques are today applied to institutionalinternal combustion engine development. On the research level, these methods aimat understanding the combustion event itself. These investigations can be outsidethe engine with various bench-top flame structures and geometries, inside constantvolume chambers simulating the engine in-cylinder conditions, or actual in-situoptical measurements of combustion in an engine that offers optical access to thein-cylinder combustion processes. These analyses are often accompanied with re-active computational fluid dynamics simulations to compare and verify the findingsand consolidate the method development tools. The developed methods and pro-vided insights then lead to combustion system developments and new operationstrategies, where certain component performance effects are investigated underthe constraints of approaching more realistic and optimized final engine operat-ing modes. These methods closely analyze a wide range of combustion in-cylinderphenomena, including but not limited to, flame kernel formation and flame propa-gation in spark ignited (SI) engines; diffusion flame formation, flame-wall interac-tions, and soot temperature and concentration in compression ignition (CI) engine;fuel spray propagation, droplet size, species concentration and temperature; andflow field velocity distribution.Optical methods can be divided into major categories of 1: The passive tech-niques relying on the natural or spontaneous light emission from the source species2(no external light source) and 2: techniques focusing on the interactions betweenthe species under investigation and an external light source illuminating these tar-get species. These interactions can result in stimulated light emission from thetargeted species or alterations in the illuminated light in the form of absorption,scattering, partial transmission, refraction, and interference.Natural emission of light from combustion of hydrocarbon fuels can be dividedinto two categories of thermal radiation and spontaneous light emission. The solidsoot particles, formed as a result of incomplete combustion of the hydrocarbon fueland suspended in the high temperature flame emits a spectrum electromagnetic ra-diations. At combustion temperatures of typically much higher than the Draper’spoint threshold (~798 K), soot particles emit significant thermal radiation in thevisible spectrum, referred to as incandescence. Spontaneous light emission stemsin the quantum mechanics view of the light emission mechanisms and refers to theprocess during which a quantum in the form of a photon is released as a result ofthe transition of a quantum mechanical system (certain gaseous species within theflame in this case) from an excited energy state to a lower energy state, referred toas the ground energy state. In the case of combustion of hydrocarbon fuels, ma-jority of spontaneous light emission comes from the molecular emissions from thediatomic carbon (C2) radicals at a number of exited vibrational states in the visiblespectrum, forming the Swan band [66], and weaker broadband emissions of CO2*radicals [50]. This phenomena is responsible for the blue flame associated withthe more complete combustion of a hydrocarbon fuel. A more detailed discussionof combustion spontaneous light emission sources and the optical investigationsthereof is presented in the next chapter.A wide variety of optical diagnostic techniques rely on application of an ex-ternal light source and investigation of the interactions the light rays make withcertain species among the reactants, intermediate products, or final products andbyproducts of the combustion process. Characteristics of such external sources areselected and designed depending on the physics of the process under investigation.This class of optical investigation can be as simple as illuminating the combustionchamber with a generic broadband white light source to visualize the injection pro-cess of the liquid diesel fuel through Mie scattering or require sophisticated andcostly design and intricate post-processing such as in Coherent anti-Stokes Raman3spectroscopy temperature measurements for in-cylinder thermal boundary layer in-vestigations [48].From a different perspective, optical diagnostic techniques can be divided intoqualitative and quantitative methods. Qualitative methods present valuable infor-mation such as the general shape of the flame kernel or diffusion flame forma-tion, interactions with the walls, propagation of the liquid fuel spray, and relativedistribution of signal intensities from various species. However, description ofsoot particles and gaseous species temperature and concentration distribution andthe effects of wall impingement on these distributions, and droplet size and flowfield velocity distributions are only possible through application of the quantita-tive methods. The latter measurement category, however, requires a calibrated orcharacterized light source specific to the technique under consideration, or a refer-ence beam and/or higher fidelity combustion chamber illuminating system. Thesein-cylinder diagnostic techniques can also be categorized based on the geometryof the optical access provided into the combustion chamber. Creating a large ac-cess, in the order of the combustion chamber boundaries geometry (through thepiston or walls), offers substantial depth of information and valuable insight intothe in-cylinder processes, albeit at the expense significant geometric modificationsand operational restrictions. Alternatively, a very small optical access can be intro-duced without major modifications to the engine design and alterations in standardoperating conditions. Although this eliminates the challenges and restrictions ofa large optical access, the information that could be extracted would be limitedand lack spatial attributes. These two design approaches to optically access thecombustion chamber were both utilized in this work; the large optical access isprovided in the “optical engine configuration”, allowing 2D optical measurements,and the small optical access, leading to 0D measurements, can be implemented inthe more realistic “thermodynamic engine configuration”. More details about thesetwo designs is provided in § Diesel EnginesIn a similar way the invention of steam engines powered the industrial revolution,the compression ignition (CI) engine is regarded as a piece of technology that has4transformed the modern world and has been the driving force behind the globaliza-tion of today’s economy. When fuelled with liquid diesel, CI engine approximatelyfollows the thermodynamic Diesel cycle and presents many attractive characteris-tics. These systems are very robust and reliable, very versatile, more efficient thangasoline fuel engines, and deliver high torques at low speeds [20]. Vast majority ofworld’s commercial, industrial, agricultural, and military machinery and vehiclesare powered by CI engines burning diesel fuel (diesel engines).When it was invented in 1890s, diesel engine was the most efficient heat en-gine ever made with nearly 20% higher efficiency than the gasoline engines and theprevalent labour-intensive steam engines. Soon, together with gasoline engines,they started to replace steam engines in mills and factories. During the first worldwar and European nations striving for improved submarine designs, the higher reli-ability of diesel engines was well recognized. After the world war, with innovationsby Harry Ricardo, specifically the comet swirl chamber design and indirect dieselinjection fuel delivery mechanism, diesel engines started to become the primaryenergy generation source for heavy-duty trucks, buses, trains, and the global seatrades transportation. Also, they proved to be a reliable source for stationary powergeneration in remote areas and emergency situations. During 1970s diesel engineswon the race with gasoline engines on all forms of land transportation except forthe important category of light-duty vehicles. During the same decade, the oil crisisin middle east and skyrocketed gasoline prices, became a motivation for improvingdiesel engines further so that they can compete with the higher performance gaso-line engines. This lead to the first direct injected diesel engine manufactured byDelphi. It was a turning point for diesel engine design and with further improve-ments of the direct injection process and introduction of high pressure commonrails, today nearly half of the new light-duty vehicles in Europe are using dieselengines. Furthermore, the marine applications is perhaps where diesel engineshave demonstrated their highest impact on modern global economy. Today’s worldintercontinental global trade depends on ocean transport to move billions of tonesof resources every year, where marine diesel engines produce efficiencies of over50%. Measured by the distance that goods have to travel from the manufacturing orcollection source to the retail, more than 90% of the global trade is diesel poweredmaking it the most indispensable engines of our era.51.1.3 Alternative Fuels for CI enginesDiesel engines originally operated on peanut oil, but through early experiments aspecific type of heavy hydro-carbon fuel, later called the Diesel fuel, became thestandard fuel for the CI engine. This fuel was cheaper to refine from crude oilthan gasoline, and it was less volatile and less likely to cause explosions. However,there are two major concerns associated with using conventional diesel fuel in theconventional diesel engine designs. First, heavily relying on this diminishing cat-egory of fossil fuels raises the concerns from the viewpoint of power generationstability policies. Despite the continuing percentage drop in oil products consump-tion since the beginning of the 21st century, it still constitutes the largest share inthe global primary energy consumption at nearly 35%, followed by coal at 27%,primarily used in electricity generation sector. Furthermore, the average global oilproduction and consumption in 2017 both reached positive rates of 0.6 and 1.8%,respectively [49]. Secondly, and perhaps more importantly, these systems producedifferent levels of various air pollutants. Production of less greenhouse gas emis-sions in diesel engines compared to other in-use fossil fuel based power sources,as a result of their higher efficiency, has been a significant advantage from the en-vironmental impacts point of view. However, these engines were soon realizedto produce other unpleasant emissions due to the higher levels of sulfates and ni-trates in the less refined long-carbon-chain diesel fuel and fundamental principlesof CI engine operation. Common pollutants emitted from diesel engines are par-ticulate matter (PM), nitrogen oxides (NOx), unburned hydrocarbons (uHC), andcarbon monoxide (CO), as well as toxics such as benzene (C6H6) and formalde-hyde (CH2O) in the form of volatile organic compounds (VOCs), which are signif-icantly higher for diesel fuel combustion compared to other in-use transportationmotive power sources [35, 97]. These emissions pose serious and complex risksto environment and have proven carcinogenic and dramatic respiratory system andcardiovascular health effects, leading to serious respiratory tract diseases such asbronchitis, pneumonia, asthma. Thus, they are all regulated in various emissionstandards, including US EPA, European Commission, and also regulations intro-duced by Environment Canada under the Canadian Environmental Protection Act,1999 (CEPA 1999), to minimize the risks and undesirable consequences.6One feasible alternative fossil fuel that can diminish the two major concernsregarding the use of conventional diesel fuels in CI engines is natural gas (NG)[70]. It has lower NOx emissions than diesel and gasoline as a result of burningat lower temperatures (having lower adiabatic flame temperatures). In addition,its low carbon-to-energy ratio combustion produces lower CO2 emissions and lesspolycyclic aromatic hydrocarbon (PAH) and solid carbon (soot) and less PM isemitted [152]. Nevertheless, in comparison with diesel fuel, natural gas less read-ily self-ignites at standard pressures and temperatures and forced ignition systemsare deemed necessary. The ignition system can be a spark plug, a glow plug as ahot surface, or a dual-fuel pilot ignition mechanism [1, 10, 13, 102, 116, 163]. Inaddition, plasma assisted and corona discharge ignition techniques are receivingmore attention over the last few years [15, 138, 150]. Also in gasoline IC engines,inherent advantages of natural gas over gasoline fuel prompted its considerationas an alternative fuel [70]. Compared to gasoline fuel, it has higher Lower Heat-ing Value (LHV) and stoichiometric air-fuel-ratio (AFR) its higher octane numberallows higher compression ratios and, hence, higher efficiencies and boost in tur-bocharged engines, and lower knock sensitivity [19]. Nevertheless, air displace-ment by NG, slower flame propagation speed, and poor lean-burn capability of NGin gasoline engines result in performance loss, lower heat release rates and enginepower output, and increase in the fuel consumption and, therefore, its implementa-tion in these engines continue to be subject to research and improvement [19, 21].Conventional diesel-ignited dual-fuel (DIDF) combustion strategy in CI en-gines ignites a compressed lean premixed homogeneous charge of air and naturalgas by late cycle direct injection of pilot fuel. It produces efficiencies and brakemean effective pressures comparable to that of conventional diesel engines whileemitting much less NOx and PM. However, these systems have some drawbacksincluding high CO and unburned CH4 emissions and some challenges associatedwith fuel efficiencies at light-load operation [153]. Furthermore, at high load op-erating conditions the premixed combustion of this homogeneous charge results inhigh cylinder pressure rise rates and engine knock [80], which can have undesirableconsequences such as engine damage and excessive noise, and therefore restrictsthe operation limits of DIDF combustion compared to conventional CI combustionstrategies.7Pilot-Ignited Direct-Injected Natural Gas (PIDING) combustion strategy is amore modern dual-fuel pilot ignition mechanism for natural gas, which was intro-duced by Westport Fuel Systems [112]. In this combustion strategy, typically asmall amount of pilot diesel fuel is injected and self-ignite by the time the subse-quent late cycle direct injection of natural gas happens and therefore acts as theforced ignition mechanism in the form of a number of combusted kernels. Thedirect injected natural gas in this case can be converted via a mixing controlledflame, depending on its injection timing, at locations close to the injector, similarto conventional diesel engines. The phasing of NG and pilot diesel injection canchange in this method. This late cycle, direct injection approach resolves the highCO and CH4 emission problem, harmful excessive pressure rise rates, and engineknock attributed to conventional dual fuel engines.1.1.4 Air PollutionAnthropogenic emissions of unregulated air pollutants has marked a number ofcatastrophes in human history. The most prominent example is certainly the greatsmog of London (the pea soup fog) in 1952 with 4000 casualties as the direct re-sult of the smog and another nearly 6000 over the following months as a result ofthe incident [11]. Similar, not as deadly smogs formed in Kansai in Japan (1960s),Pennsylvania (1948) and California in US, where it has been a continuing challengesince its first reported dangerously high smog level (1943), and most recently Delhiin India (2016). The major culprits of smog formation are airborne particulate mat-ter, sulphur dioxide, carbon monoxide, NOx, VOCs, and unburned hydrocarbons,often aggravated by geographical properties of the region and atmospheric eventssuch as temperature inversion.The concurrent presence of sufficient concentrations of NOx and VOCs in theatmosphere, with enough UV radiation energy from the sun, favors a photochemi-cal reaction which yields airborne particles and ground-level ozone. Later anotherpathway from NOx to ozone was found that predominantly occurs in coastal areasvia formation of nitryl chloride when NOx comes into contact with salt mist [120].Through another chemical reaction, sulphur dioxide and nitrogen oxide react withthe water molecules in the atmosphere to produce acids, lowering the PH in the8smog. Road transport is the largest contributor to urban NOx emissions worldwidewith approximately 40%, nearly double the share from the next big source, energyproduction and distribution sector. Diesel engines are responsible for about 85% ofthis share, primarily in the form of NO, which typically constitutes 85–95% of thetotal NOx. With time, the colorless and odorless NO gradually converts to NO2,which is a brown-reddish gas with a pungent odor at normal temperatures and hasa five times greater toxicity level [22, 56, 86, 151].Particulate matter emission from coal and diesel power sources played a majorrole in the smog formation incidents and later realized to have significant health ef-fects even at much lower concentrations. Particulate matter (PM) is the term usedto refer to any airborne solid and liquid particulates. These particles can be directlyemitted from a source or as a result of chemical reactions between certain gaseousspecies in the atmosphere. PM emission can be naturogenic or anthropogenic, itcan comprise a wide variety of species, and its particle sizes vary depending on thecircumstances under which they are formed. An aerodynamic or effective diame-ter is the typical metric to describe particle sizes. Coarse particles with a diametergreater than 2.5 µm have relatively high settling velocities and settle out of at-mosphere in a matter of Hours or days. As the particles diameter decreases thesettling velocity decreases. Ultrafine particles (sub 100 nm diameter) tend to ag-glomerate and form larger particles in the range of 0.1 to 1.0 µm, which have lowsettling velocities and very long residence time [134]. PM emissions originatingfrom combustion processes, such as in diesel engines, have size distributions withinthis range and lower, either directly as products or through condensation of VOCsin the exhaust stream [97, 132]. Diesel engine PM emission consists of three ma-jor components: soot, soluble organic fraction (SOF), and inorganic fraction (IF).Engine out soot, the amorphous black carbon shaped into a porous agglomeratestructure through a complicated formation path, constitutes nearly 50 % of the to-tal PM emissions. SOF consists of adsorbed or condensed heavy hydrocarbons onthe soot surface, originating from the lubricating oil, uHC, and other byproduct ofthe combustion process [132, 144].All of the compounds mentioned above have proven negative environmentaland health effects. NO2 exposures can lead to irritation in the lungs and affectrespiratory tact immunity to infections such as influenza [56, 134]. The odorless9and colorless CO, when inhaled, binds to hemoglobin in bloodstream and inhibitsits capacity to transfer oxygen. Depending on the concentration, this can causeimpaired concentration, slow reflexes, dizziness, and ultimately asphyxiation lead-ing to unconsciousness or death [149]. Ozone is a powerful oxidant that decreaseslung function and produces toxic signals when reacts with the proteins and lipidswithin lung lining fluid compartment and, leading to inflammation and ultimatelydamage in pulmonary cells [103]. PM emissions have been the subject of numer-ous epidemiological studies and have shown strong correlations with impaired lungdevelopment, respiratory system irritation, cancers due to adsorbed or condensedVOCs such as benzene and formaldehyde, and overall increased cardiovascular andpulmonary system failure [31, 97]. Studies show that coarse particulates are un-likely to penetrate into the lungs and peak deposition occurs for particulates in therage of 0.01 tp 0.1 µm. These smaller particles are also more likely to find theirway to the bloodstream and a greater negative effect on the cardiovascular system[97]. Examination of diesel engine particulate diameters shows 90% of the sizedistribution bellow 1 µm [132]. There are also a variety of negative environmentaleffects associated with PM emissions, such as pollution of water and soil, soiling ofbuildings, reductions in visibility and agriculture productivity, and global climatechange [35, 43].Following the early high mortality incidents of dangerously high air pollutantlevels, the Clean Air Act (CAA) legislation was enacted in UK in 1956 and USin 1963, followed by similar regulations in many other countries to avoid futureoccurrences of such incidents. Although the original regulations successfully pre-vented such severe cases, epidemiological studies later showed that exposures topollutant levels much lower than those originally considered in the clean-air actregulations closely correlated to increased cases of serious health conditions andpremature death[31, 35, 56]. As a result of these investigations and in response tothe more in-depth understanding of the adverse effects of various emissions fromheavy-duty vehicles governments have been consistently introducing more strin-gent regulations on permissible exhaust emission. This has been in the form ofEuro I-VI standards by European Commission and Tier 1 to 3 EPA federal regu-lations in US, which has also adopted more strict regulation for the light-emissionvehicle (LEV I-III), legislated by Air Resources Board in California (CARB) given10their continuous battle with higher air pollutant levels. Canada also closely alignsits emissions regulations with those of EPA’s. As an example of evolution of theseregulations over time, Table 1.1 shows European Union’s Euro standards on fourof the major emissions from the heavy-duty diesel engines [110].Table 1.1: Euro standards for heavy-duty vehiclesCO uHC NOx PM[g/kWh] [g/kWh] [g/kWh] [g/kWh]Euro I 4.5 1.1 8.0 0.61Euro II 4 1.1 7.0 0.15Euro III 2.1 o.66 5.0 0.13Euro IV 1.5 0.46 3.5 0.02Euro V 1.5 0.46 2.0 0.02Euro VI 1.5 0.13 0.4 Developing TechnologiesA comparison of the permissible pollutant levels between Euro I (implemented in1993) and Euro VI (implemented in 2014) in Table 1.1 shows much tighter emis-sion limits of 67% for CO, 88% for uHC, 95% for NOx, and 98% for PM. Theeffectiveness of these corrective measures are conspicuous from the collected sta-tistical emission data. In Europe, for instance, there has been a 60% drop in NOxand non-methane VOC emission levels since 1990 and a 30% drop in PM emis-sions since 2000 [2]. However, despite the improvements made by introductionof new technologies to address the air pollution challenge, a significant role isstill played by the exhaust stream aftertreatment procedures to meet the legislatedstandards. These effective techniques, in addition to design and operational com-plexities and capital costs, might require additional work output from the engineto expel the exhaust gas from the system. As such, further investigations and con-tinuous improvements on these developing technologies are essential to increaseengine efficiency and reduce the production of criteria emissions and the burdenon the exhaust aftertreatment systems to meet the ever-increasing stringent emis-sion regulations is of great value.From the perspective of optical measurement techniques, however, the de-11scribed improvements to reduce formation of various pollutant species during thecombustion in-cylinder processes can introduce challenges in terms of the signal-to-noise (SNR) ratio of the recorded data. This will limit the information providedby the qualitative measurements, but more importantly, it will directly affect therange of properties to be calculated in the quantitative measurements. Also, thelow-SNR issue potentially can be more of a concern in the case of natural emis-sion optical measurement techniques, where the collected data SNR is solely deter-mined from the characteristics of the source species and there is no additional con-trol factors such as the intensity of an external light source. As a consequence, spe-cific considerations might be necessary to improve the performance of the imagingsystem so as to make it more suitable for these “low light” combustion strategies.The enhanced measurement techniques, along with introduction of novel analy-sis procedures, would present a valuable toolkit to better understand combustionin-cylinder processes within the existing and developing CI engine technologies.With reflection of the acquired insights from the likes of these investigations intothe design of the next generation of these technologies and their operation proto-cols, the 19th century diesel engine invention can continue to evolve and adapt withthe needs of our era and remain the the most important engine of the 21th century.1.2 Objectives, Contributions, and ApproachesThe motivation for more in-depth characterization of and further improvements tomodern alternative CI engine strategies, and a brief overview of the current tech-nologies was presented in the previous section. With this introduction, the majorobjectives and contributions of the current work can be listed as follows:1. Development of standard high-speed imaging diagnostics, and the support-ing analysis tools, to further study combustion strategies using NG in CIengines.2. Enhancement of a quantitative in-cylinder soot concentration and tempera-ture measurement technique (two-color pyrometry) for low-SNR signal sit-uations associated with the these and any similar combustion strategies.3. Implementation of the developed and enhanced techniques to multiple com-12bustion strategies to investigate various parameter effects and compare ob-served combustion trends with existing understanding of various combustionmodes.4. Facilitate and assess the correlations between 2D and 0D quantitative in-cylinder soot measurement techniques, as well as 0D measurements in op-tical vs. thermodynamic engines, as an attempt to extend the in-depth 2Doptical results to 0D measurements under more realistic operating condition.The approach adopted to achieve these objectives is detailed throughout themain body of this thesis. A literature review of the relevant optical diagnostictechniques and investigations of the DIDF and PIDING combustion strategies wasperformed and presented in Chapter 2 to establish current knowledge and elucidatethe existing gap motivating the presented work. The standard high-speed imag-ing diagnostic systems developed in this work are described in Chapter 3, alongwith other experimental facilities used, as well as development of a simultaneous2D optical measurement system to facilitate analysis of the results acquired fromtwo optical methods. The spatially resolved (2D) measurements aimed at advanc-ing interpretation of combustion analysis results obtained from other experimentalmeasurement techniques as well as producing the baseline data required later inthis work. As such, the commissioned simultaneous high-speed 2D optical mea-surements of DIDF combustion, as a case study, is discussed in Chapter 4. Thetwo simultaneous optical diagnostics considered were both qualitative measure-ment techniques. As a prerequisite to performing the quantitative in-cylinder sootmeasurement technique in low-signal conditions of the combustion strategies con-sidered for this work, Chapter 5 identified and assessed improvements to variousstages of the standard technique. These included refinements to the measurementsystem calibration and configuration, image post-processing, and solution algo-rithm. The enhanced method was then implemented, simultaneously with measure-ment of spontaneous emissions characterizing high temperature reaction zones, foroptical investigation of soot processes in PIDING combustion under a range ofoperating conditions. These results, presented in Chapter 6, also highlighted themajor differences of PIDING, as an advanced natural gas ignition technology, incontrast to diesel and conventional DIDF strategies. The insight provided into the13influence of engine level control parameters on in-cylinder processes could helpimprove this combustion strategy; However, the anticipated changes in in-cylinderprocesses, imposed by the geometric modifications and restricted operational con-siderations in an optical engine, promotes application of the less processes-altering0D measurements in a less idealized engine environment. Therefore, a series ofconcurrent 2D and 0D measurements of soot concentration and temperature wascarried out in the optical engine configuration and presented in Chapter 7. An anal-ysis methodology was developed for more meaningful comparison of the 2D and0D results and better assessment of the correlations between these two approaches.This was followed by implementation of the 0D soot measurements in the thermo-dynamic engine configuration to investigate the different combustion environmenteffects as well as to better understand the 2D vs. 0D measurement effects. The pre-sented efforts to achieve a more characterized and reliable 0D optical measurementtool to study combustion in-cylinder processes, is in fact an approach to assess thefeasibility of extending the detailed optical diagnostics to engine operating condi-tions that are more representative of real-world on-road operation.14Chapter 2Background Information andLiterature ReviewIn this chapter, some background information and review of the relevant literatureon the optical diagnostic techniques utilized in this work, as well as the combus-tion strategies these techniques were applied to, are presented. This will lay thegroundwork for all the analysis results and interpretations made throughout thesubsequent chapters of the current document.2.1 Optical Diagnostics for Combustion In-CylinderProcessesOptical diagnostics in nature is essentially the study of the energy exchange in theform of electromagnetic radiation and its interactions with matter. In the field ofcombustion research, these methods aim at understanding the fuel conversion pro-cess and providing insight into characteristics of the injected fuel (liquid dropletsor gaseous cloud) and the produced particles and gaseous species. Parameters un-der investigation are majorly the instantaneous amounts and distributions of thereactants, intermediate, and product species, under various operating conditions;and the influence of different operational control parameters on these characteris-tics. Ultimately, in any combustion system, these findings are intended to optimizethe system such that an operational space be described where the constituent oper-15ating points lead to lower undesired system-out emissions, while maintaining highefficiencies.There is a wide range of optical methods applied in engine and combustion re-search, relying on thermal radiation from suspended particles; quantum emissionsfrom gaseous combustion species or introduced seeding agents; or alterations in thecharacteristics of an external light source passing through the combustion chamber.In what follows an overview of prominent optical methods relevant to IC enginestudies is presented, along with a brief overview of the involved background theoryin each case.2.1.1 Optical Access to the Combustion ChamberIn order to implement optical diagnostic methods to understand combustion in-cylinder processes, an optical access to the combustion chamber must be providedfirst. This, however, is a challenging task that induces various levels of complexityand changes in the combustion chamber geometry and conditions, depending onthe type of optical access considered. In the absence of such an optical access toan engine, or in the case of more fundamental analysis of combustion, constantvolume chambers (CVC) [62, 159] or rapid compression and expansion machines(RCEM) [133] have been used to mimic in-engine conditions and present valuablespatial information on combustion processes to complement traditional analyses re-sults performed on the basis of calculated heat release rates (HRR) and engine-outemissions. However, being able to take optical measurements from the in-cylinderprocesses of an engine would undoubtedly provide more relevant information re-quired to optimize operation or develop new combustion strategies under currentdesigns, or improve future designs. optically accessible engines approaches can bedivided into three major categories: optical probe design, endoscopic design, andBowditch piston design [16].Optical Probe DesignIn optical probe designs, light is collected from a small control volume, definedby the field-of-view (FOV) of the probe optical lens, and the signal output, inte-grated over the control volume, is transmitted through fibre optic assemblies to16Figure 2.1: Schematic of a sample optical probe light collection diagram.Reprinted from [155] with permission from publisherreach sensors. Today, these sensors are typically solid-state Silicon photodetectorsor photomultiplier tube arrays, depending on the strength of the recording signals.The result would be single-value light intensities recorded with the temporal reso-lution of the sensors, possibly filtered around certain wavelengths depending on thephenomena under investigation. Figure 2.1 shows an schematic example of opticalprobe assembly.Endoscopic DesignEndoscopic probe design is an extension to the optical probe design discussed inthe previous section, where a coherent fibre bundle, maintaining the relative po-sition of individual fibers in either end, transmits the high f-# collected light inthe field-of-view. On the other side, a set of lenses are in-line mounted and focuson the fibre bundle exit port, magnifying the transmitted image to project onto acamera sensor. This will provide valuable 2D light measurements from in-cylindercombustion process, albeit within a limited field-of-view and from a small regionwithin the combustion chamber. Figure 2.2 provides an schematic of such assem-bly and sample natural luminosity images taken from a diesel-ignited dual-fuelcombustion case study.17Figure 2.2: Schematic of a sample endoscopic light collection diagram,reprinted from [135] (left) and sample endoscopic image from DIDFcombustion, adapted from [18] (right) with permission from publishersBowditch DesignThe most common approach to gain a full optical access to the combustion chamberis a Bowditch optical piston extension design [16]. In this design, a relatively thickwindow (typically quartz or sapphire) is mounted inside a modified piston crown.A hollow engine block extension and an elongated hollow piston extension, withthe window in its piston crown, are mounted on top of the main block and a dummypiston. A 45◦ mirror is then mounted on the block extension through the pistonextension. With the long enough piston extension hallow area (stroke + mirrorheight + tolerance), the stationary mirror reflects the light from the combustionchamber, transmitting through the piston window, towards the imaging equipment,without interference from engine moving parts. Figure 2.3 shows the Bowditchoptical engine design.The modifications imposed by the optical engine configuration makes the pro-cesses no longer necessarily identical with those in the conventional “thermody-namic” configuration and impose a series of limitations on operating conditions,due to the optical parts mechanical properties. A thorough overview of the dif-ferences between thermodynamic and optical engine operations is presented byKashdan et al. [72] and Aronsson et al. [5]. Among these are changes in the com-pression ratios as a result of altering the piston crown geometry and rings design,18Figure 2.3: Schematic and photo of the optical configuration of the enginelower heat losses through the walls due to the lower heat conductivity of opticalparts and a less efficient cooling system [5], lower thermal conductivity of theglass compared to aluminum, which may promote the formation of hot spots, andthe necessity of a skip-firing procedure to reduce window failure probability andcleaning requirements. Kashdan et al. also considered an “effective” compressionratio for the optical engine as a result of the dynamic and thermal loading of theextended piston-liner assembly [72]. The effects of these changes manifest both inthe heat release and combustion phasing, and in the engine-out emissions [5, 72].There are some practical approaches attempting to compensate for these differ-ences. A common practice is to provide the engine with heated and pressurizedair in order to compensate for the effects of the lower compression ratio in opticalengines, as was done by Taschek et al. [142] and Dembinski [27], for instance, us-ing an external boosting device. Intake air temperature and pressure is usually setbased on estimations on the engine charge conditions at the top dead center (TDC)19for an equivalent thermodynamic engine. This estimation process becomes morechallenging when heat transfer effects are to be considered as well or additionalphysics is involved, such as in exhaust gas recirculation (EGR) mode. Moreover,even if the thermodynamic conditions are matched at TDC, the total mass of thecharge gas and its temperature and density away from TDC would be higher inoptical engines [109].2.1.2 Natural Luminosity ImagingThe natural luminosity (NL) imaging, as the name suggests, relies on the informa-tion that can be collected from the visible portion of the electromagnetic radiationspectrum from the self-radiating combustion products and scattered light (around380-700 nm). This basic imaging technique attempts to qualitatively characterizethe combustion in-cylinder processes by looking at the broadband emissions. Thenatural luminosity light from combustion of hydrocarbon fuels generally comprisesa “blue” flame and a “yellow” flame [29]. The yellow flame attributed to the in-candescence from glowing suspended particles, and the blue flame originates fromchemiluminescence light from intermediate species, discussed in the next section[46]. During combustion, soot particles reach temperatures much higher than theDraper point and thermally radiate a broadband light, called incandescence andconsidered to be diffuse.Natural luminosity imaging of the flame and the associated incandescent sootparticles is the oldest optical diagnostic technique applied to better understandcombustion in-cylinder processes. In an early CI engine optical investigation in1934, Rothrock showed that the diesel flame first appears on the spray envelopeand from there spreads to other parts of the combustion chamber; and that thecourse of combustion is highly dependent on the air pressure and temperature inthe chamber between the injection and start of combustion [129].When imaging the entire visible range, the chemiluminescence light is usuallydominated by the incandescence generated from soot particles at elevated temper-atures, when present concurrently. For typical diesel fuels and under typical CIengine operating conditions, natural luminosity from soot incandescence has beenreported 4-5 orders of magnitude stronger than chemiluminescence [104]. Despite20the technological advancements and improved theoretical understandings of the de-tails of in-cylinder processes, natural luminosity imaging remains a popular basic,easy-to-implement, technique to qualitatively characterize combustion, often usedin conjunction with other optical methods [29, 104, 108, 136, 140, 142]. In addi-tion to the interpretation of the broadband natural luminosity light, researchers haveused the signal level disparity between the chemiluminescence and incandescencelights to segregate the blue and yellow flames. Taschek et al. [142], along withother optical diagnostic tools, adjusted camera exposure to discuss the soot cloudbehavior and chemiluminescence from intermediate species separately in an opti-cal CI engine, in order to study the effect injector nozzle holes geometry on sprayformation, mixing, ignition, combustion, and soot formation. The high intensitysoot incandescence required relatively much shorter camera sensor exposure timeto record images with an acceptable dynamic range, while the low intensity chemi-luminescence is expected to be negligible at such low exposure times. In contrast,high exposures required for imaging low intensity chemiluminescence light prop-erly, would lead to sensor saturation in presence of intense soot incandescence.Alternatively, Singh et al. [136] and Musculus et al. [108] used Neutral Density(ND) filters in order to lower the transmitted spatially resolved natural luminos-ity signal level (without altering spectral information) and eliminate the weakerchemiluminescence light. The images of the stronger yellow flame were used tosupport the spatially-integrated quantitative measurements of in-cylinder soot. Intwo more recent studies by Khosravi et al. [75] and Hatzipanagiotou et al. [54]dichroic mirrors were used in the natural luminosity light path that blocked signalsfrom lower wavelengths (350nm and 340nm, respectively), and thus, eliminatedpart of the chemiluminescence signal (spectral behavior of the chemiluminescencelight from hydrocarbon fuels is discussed in the next section).2.1.3 Chemiluminescence ImagingChemiluminescence is a sub-class of the spontaneous emission of light, where theatoms or molecules transition to an excited energy state by the energy releasedthrough chemical reactions and photons are released as they relax back to theground energy state. In the case of combustion of hydrocarbon fuels, chemilumi-21nescence emitting molecules are mostly short-lived electronically excited interme-diate species formed during the fuel pyrolysis chemical reactions and naturally emitphotons in the UV and visible range [47]. The diatomic carbon (C2*), hydroxyl(OH*), methylidyne (CH*), and carbon dioxide (CO2*) radicals are the majorsources of chemiluminescence emission from hydrocarbon fuel combustion, whilechemiluminescence from formyl (CHO*), formaldehyde (CH2O*), cyanido (CN*),carbon monoxide (CO*) can also be significant under certain operating conditions[6, 19, 46, 50, 53, 101, 140, 145]. Chemiluminescence from these species areoften used as a quantitative diagnostic in combustion studies, e.g., for local equiv-alence ratio estimations in gaseous and liquid fueled systems [3, 51, 53, 106, 137],tracking locus of the flame front [61, 140], and analyzing heat release noise andinstabilities in combustion dynamics investigations [6, 85]. Table 2.1 illustrates theformation reactions and emission peaks/bands for the major chemiluminescencesources in combustion of hydrocarbon fuels. In the listed chemical reactions inthis table, M is a third-body chemical compound that participates in thermal deac-tivation of the excited molecules through inelastic collisions, a process referred toas “collisional quenching” [40, 46].Table 2.1: Formation pathways and characteristic wavelengths of the majorchemiluminescence sources in combustion of hydrocarbon fuels [9, 19,46]; prominent peaks are printed in bold font.Species Reactions Wavelengths [nm]OH* CH + O2 −−→ CO + OH*H + O + M−−→M + OH* 281, 307, 342OH + OH + H−−→ H2O + OH*CH* C2H + O2 −−→ CO2 + CH*C2H + O−−→ CO + CH* 365, 389, 431.5C2 + OH−−→ CO + CH*C2* CH2 + C−−→ H2 + C2* 474, 516, 563, 655CO2* CO + O + M−−→M + CO2* 340-650The spontaneous spectral emissions from these species can be in the form of22peaks of radiation at one or a few certain wavelengths with a relatively narrowbandwidth, as in the case of intermediate species such as OH*, CH* [46, 53].Alternatively, these emissions might occur within a wider wavelength bandwidthat lower signal levels, as in the case of CO2* [53] and CHO* and CH2O* undercertain operating conditions [145].C2* radicals chemiluminescence emission spectra comprises a number of ex-ited vibrational states in the visible spectrum (473.7, 516.5, 563.5, and 655 nm[118]), making C2* a major contributor to the blue flame [29, 66]. Among thesepeaks, the 516.5nm peak is the most prominent one and the target of C2* chemi-luminescence measurements [53, 137, 140, 145]. In one of the early studies in thesearch for a relationship between chemiluminescence and heat release rate, Priceet al. [123] showed a linear relationship between the C2* chemiluminescence andthe mixture flow rate, independent from the conditions of turbulence. Langhorne[82] measured C2* chemiluminescence emission to study reheat buzz, a combus-tion instability phenomenon that occurs in the afterburners of jet engines. C2*chemiluminescence has often been used in conjunction with other chemilumines-cence signals to quantitatively assess local heat release rates and equivalence ratios[53, 77, 131].OH* and CH* radicals are two of the most significant chemiluminescencesources in combustion diagnostics. They are often used to locate flame fronts orto analyze local flame structure [19, 46, 61], as well as to study the heat releaserate and equivalence ratio [6, 9, 46, 53]. De Leo et al. [89], for instance, observedhigh intensities of OH* and CH* chemiluminescence emitted in the reaction zoneof opposed flow methane oxy-flames and concluded that their peak intensity is agood indicator of the flame front. The OH* radical shows a strong peak of natu-rally emitted chemiluminescence signal at a narrow wavelength band centered at307 nm, which is the result of integration over multiple OH* (0, 0) bands at 306.4,306.8, 307.8, and 309.0 nm and the OH* (1, 1) bands at 312.2, 312.6, 313.5, and314.7 nm [77]. Besides this strong peak which is often the subject of chemilumi-nescence measurements, OH* spectra shows a weaker peak at 280 nm [101]. CH*spectra peak at 431.5 nm offers the stronger transition bands and is often used forchemiluminescence measurements [53, 140, 145].In pollutant formation studies, CH is recognized to play a significant role in23NOx formation through the prompt NO mechanism and oxidation of CH is one ofthe major pathways of producing chemically excited OH. It is very unstable andtends to chemically react with other species very quickly, and exists in a narrowspatial and temperature region within the flame, and can provide insight into theC2 reaction chain [19, 148]. The CH* radical formation is a strong function oftemperature [45] and high intensity CH* chemiluminescence indicates high localtemperatures. Since it is produced in a very narrow width region of the flame front,CH* can be tracked to mark the locus of flame front, especially in turbulent flames[46, 101]. CH* is also recognized as a marker for hydrocarbon fuel break-downinto simpler molecules and high CH* chemiluminescence emission was reportedto indicate the continuing diesel fuel vaporization process [19].OH* chemiluminescence measurements has been widely used for combustioncharacterization. Investigation of the non-sooting flame or high temperature reac-tion zones is perhaps the most significant application of OH* measurements [159].The intensity of OH* chemiluminescence depends on the local temperature andAFR [46]. Thus, high OH* chemiluminescence intensities in a region can indicatenear stoichiometric AFR and higher temperature [19, 142]. The ignition processin a diesel engine has been described to occur in three stages: starting with a coolflame reaction, followed by formation of chemiluminescent flame, and ending withthe ignition of CO [105, 142]. During the simulation of ignition reaction mecha-nisms for n-heptane in a homogeneous reactor, OH* appeared as one of the firstspecies with significant concentrations [105], followed by CH* shortly. This re-sult was later confirmed by optical in-cylinder investigations of Bertsch [12] andKoyanagi et al. [79]. Based on this description, OH* chemiluminescence was per-ceived as a good marker for the ignition process [142]. OH* is also believed to bea good indicator for soot oxidation [94], and its presence has been used to definethe flame area [25]. In addition, Ayoola et al. [6] compared spatially resolved OH*and CH* chemiluminescence and the flame surface density and heat release rateand demonstrated good correlations, which led them to the conclusion that boththe OH* chemiluminescence and flame surface density measurements reasonablyestimate heat release rate. In gasoline engine studies, the described relation be-tween OH* chemiluminescence and local temperature and heat release has beenused to distinguish between the unburned charge mixture and the burned gas re-24gions [101].The ratio of various chemiluminescence peaks have been used to describe localequivalence ratios in engines [3, 19], given the dependence of these chemilumines-cence intensities to the local AFR. OH* and CH* measurements of perfectly pre-mixed flames showed a linear correlation between chemiluminescence intensitiesand charge mass flow rate and an exponential one between these intensities and theequivalence ratio [55]. Therefore, the ratio of two chemiluminescence peaks willbe only a function of the equivalence ratio. Ratios of CH*/C2*, CH*/OH*, andC2*/OH* chemiluminescence demonstrated good estimates of local equivalenceratios [77].The overlapping quantum emission bandwidths and signal interference, espe-cially due to the existence of broadband emissions from certain species, should beconsidered during chemiluminescence measurements. Chemiluminescence fromCH* radical at 431.5 nm, for instance, is identified to accompany backgroundemission signal from CO2* from 340-350 nm to somewhere between 500-650 nm[6, 9, 19]; convolution of the CH2O* Vayda band from 250 nm to 410 nm, andCHO* Emeleus flame band from 330 to 523 nm [19, 46, 101]. Hardalupas andOrain [53] evaluated the magnitude of background intensity from CO2* relative tothe chemiluminescence signal from OH, CH, and C2* radicals, using a Cassegrainoptics system, in counter-flow natural gas flames with different equivalence ratiosand strain rates. They observed that although at near stoichiometric and rich mix-tures the ratio of the CO2* background intensity over CH* and OH* chemilumi-nescence intensities are similar (~10%), the background/CH* ratio can be as highas 30% on the leaner mixture side. Another important source of signal interferencefor chemiluminescence measurements, especially when non-premixed flames areconsidered, is thermal radiation from generated soot particles. This effect is moresignificant for liquid fuels as they tend to produce more soot [161]. Backgroundemission from incandescent soot particles is negligible in the UV range near theOH* chemiluminescence peak, but can be substantial around the peak of CH* rad-ical emissions [63, 71].252.1.4 Two-Color PyrometryAs discussed in the previous chapter, characterization of the in-cylinder soot for-mation and oxidation processes is a valuable tool for reducing the engine out soot.However quantitative in-cylinder measurements are not trivial in light of the com-plex thermochemical processes and challenging access to the combustion chamber[147]. Several methods are available for characterization of the in-cylinder soot.Multi-color pyrometry is used to measure the soot concentration and temperaturebased on the radiation from a soot cloud, along with an empirical wavelength-dependent representation of the soot cloud emissivity [162]. The higher fidelitylaser-based methods, such as Laser Induced Incandescence (LII) and light scatter-ing methods, are also commonly used for quantitative in-cylinder soot characteri-zation. In what follows, first a brief comparison between these two approaches arepresented; the fundamentals of the two-color pyrometry measurement and analysisconcepts are then reviewed and a few examples of its application for in-cylindersoot characterization is discussed.Comparison with Laser-Based MethodsLaser Induced Incandescence (LII), light extinction method, Laser Induced Scat-tering (LIS), and Laser Induced Fluorescence (LIF) are the prominent laser basedtechniques used for quantitative in-cylinder soot characterization [88, 146, 147,162]. LII provides planar distribution of the soot concentration on a laser sheetradiated to a soot cloud to increase soot particles temperature well above the sur-rounding gas. Light scattering can be divided into two groups of elastic and in-elastic methods, depending on whether the energy exchange mechanism interfereswith the irradiated/radiated frequencies. The light extinction method is an elastictechnique based on measurements on the attenuation of emitted light (extinction)passing through a cloud of soot particles to calculate soot concentration [162]. LISis another elastic scattering technique which uses simultaneous imaging of scat-tered light polarized in two directions to study spatial and temporal variations ofsoot based on interpretation of the polarized ratio [87]. LIF is perhaps the mostimportant inelastic light scattering technique in soot formation and oxidation stud-ies. In this technique, a species of interest is excited with a laser beam/sheet and26then emits light at wavelengths higher than that of the excitation laser used (de-excitation). Different species have been subjected to LIF investigation, based onthe hypothesis of being precursors of either formation or oxidation of soot parti-cles, e.g., formaldehyde [78], OH* [147], and PAH [88]. LIF and LII are oftenperformed in conjunction to each other for comparison and cross-correlation pur-poses [78, 88, 147]. These elaborate techniques, however, are reported to havemore limited maximum detection ranges, as the laser light must penetrate the sootcloud with sufficient intensity to produce an adequate signal, which then must pen-etrate through the soot cloud to reach the camera sensor [147]. Typical concen-tration limits for LII were reported to fall in the 0.1–0.3 ppm range by Pinson etal. [119], while higher soot volume fraction upper limits of ~20 ppm were identi-fied in several two-color pyrometry studies [147]. Line of sight (LOS) extinctionmeasurements by Tree and Dec [146] and Musculus et al. [107] also suggestedthat, under the operating conditions considered in their work, soot concentrationsin diesel flames typically exceed the levels that can be measured using LII andLIS methods. In addition, the laser-based measurement techniques, such as theLII, LIS, and LOS extinction methods described above, require high capital invest-ments, primarily for the laser, especially if high temporal resolution measurementsare considered, as well as a second optical access to the combustion chamber tointroduce the laser light. In comparison, a multi-color pyrometry system would besimpler and more cost-effective to setup, as it only requires one optical access tothe engine and an imaging system, but no light source.Fundamentals and ApproachThe two-color pyrometry method considers the spectral radiant intensity at twowavelengths to evaluate the soot temperature and concentration. For this, thespectral emissivity of a soot cloud is assumed to be a function of the detectedwavelength, soot concentration, optical path length, and optical properties of sootdescribed by the complex refractive indices and soot particle size and shape (ex-tent of agglomeration). The semi-empirical relationship proposed by Hottel and27Broughton [57], is commonly used to describe the soot cloud spectral emissivity:ελ = 1− e−KL/λα(2.1)where the KL factor, as shown in equation 2.2, is directly proportional to thesoot volume faction, fv [147], and determines the amount of soot in the optical pathof the captured light.fvL =KLAλλα−1(2.2)where L is the geometric optical path length and Aλ is an optical constant deter-mined from the refractive indices of soot and was approximated by 6+0.27λ [µm][41]. The parameter α is an empirical constant. It has been noted elsewhere thata constant value (α=1.39) can be used for wavelengths in the visible range withnegligible effect on the calculated temperature and KL [96, 162]. More recently,Kamimoto and Murayama [67], re-examined the value of α for diesel soot col-lected from the surface of a quartz window in an optical engine and obtained aclose estimate of α=1.38. They further examined this value theoretically throughcalculation of spectral extinction using the Rayleigh-Debye-Gans (RDG) theory ofsoot aggregate scattering with an assumption of constant refractive index and ob-tained a value of α=1.31. Comparison of the acquired flame temperature and KLfactor using their alternative description of the soot cloud emissivity with those ofa conventional two-color pyrometry method showed an agreement of within 1-2%.Equation 2.1 originates from the Kirchoff and Bouguer-Lambert laws for ahomogeneous absorptive medium with no scattering resulting in:Iλ (L) = Iλ (0)exp(−κλL) (2.3)where Iλ is the spectral radiant intensity and κλ is the absorption coefficient.κλ is a function of the detected wavelength, complex refractive indices, soot con-centration, particles size and shape, As an approximation, Hottel and Broughtonsuggested the substitution κλ = K/λα , where K is the absorption strength of thesoot cloud [57], and is proportional to the number density of soot particles [162].As a result, the effects of the wavelength, particle size and shape, and complex28refractive index on the emissivity are described by λα and the product KL charac-terizes the total soot volume [147]. The dimensionless product κλL (or equivalentlyKL/λα ) is referred to as the optical thickness of the soot cloud. The recorded lineof sight signal might originate from a thick cloud of low soot concentration, ora thin cloud of high soot concentration, while both scenarios represent the samesoot cloud optical thickness. In the current work, a high KL factor signal will beattributed to an optically thick flame, and a low KL factor signal to an optically thinflame, regardless of the physical characteristic length.Combining equation 2.1, the definition of emissivity, and Planck’s distribution,and assuming that the KL factors at the two wavelengths are equal, result in aresidual function f :f (T,xi, j, t) = λα1 ln[1− eC2/λ1T (xi, j,t)−1eC2/λ1Ta,1(xi, j,t)−1]−λα2 ln[1− eC2/λ2T (xi, j,t)−1eC2/λ2Ta,2(xi, j,t)−1] (2.4)where C2 = 1.43x10−2 mK from Planck’s Distribution, T is the soot cloud tem-perature, and Ta,1 and Ta,2 are the apparent temperatures at λ1 and λ2, respectively.In a physical sense, apparent temperature at a certain wavelength, also referred toas the brightness temperature, is the temperature of a black body emitting the samespectral radiant intensity as the soot particles at that wavelength, as described inEquation.Iλb(λ ,Ta) = Iλ (λ ,T ) (2.5)Temporally and spatially resolved apparent temperatures are obtained usinga calibrated high-speed imaging system (see §3.2.2). Equation 2.4 is solved it-eratively for T at each image pixel (xi, j) and time step t. The KL factor is thenevaluated using:KL(xi, j, t) = λα ln[1− eC2/λT (xi, j,t)−1eC2/λTa(xi, j,t)−1](2.6)for either of the wavelengths. While the solution of equation 2.6 will pro-29vide the particle cloud temperature, Baker and Ryder,[8] have shown that equilib-rium time constants for soot particles and surrounding gases are sufficiently small(10−6−10−5 sec) that the soot particle temperature are indicative of the local gastemperature for engine relevant combustion processes.The range of wavelengths considered for this method span from visible (0.38-0.7 µm) to Near-infrared (NIR, 0.75–1.4 µm). In the temperature ranges relevantto diesel combustion, which is approximately 1800-2800 K [117, 160], the elec-tromagnetic radiation is dominated by the wavelengths in the NIR, which yieldsbetter signal-to-noise-ratio (SNR). However, shorter wavelengths in visible rangeare used in majority of the works for a number of reasons. In the visible range,the rate of change of spectral radiant intensity with respect to both the temperatureand wavelength, dIλ/dT and dIλ/dλ , is larger [162]. Also, the effect of radiationfrom the wall is significant for the NIR and reflection of the soot particles radiationfrom the opposite walls significantly contributes to the total error [96]. Finally, thequantum efficiency of the sensor in majority of regular digital high speed camerasdeteriorates starting from the end of the visible range (around 800nm) and also αcan be considered to be constant in this range. For diesel combustion, wavelengthsin the range of 550-750 nm are widely used [52, 62, 65, 117, 158, 160]. However,given the lower combustion temperature ranges for natural gas and less amount ofsoot expected to be produced, wavelengths around 700 nm and 800 nm were chosenin this work. For this selection, attention was paid to the relevant gaseous speciesspectra to ensure no interference from other species (notably CO2 and H2O).The pyrometric method described above has been used for spatially integrated[108, 136, 162], as well as spatially resolved (imaging) measurements [62, 64, 92,117, 158, 160]. For imaging purposes, two correlated images with known spectralintensities are required as inputs to equation 2.4. This has been achieved throughusing two calibrated monochrome cameras with a beam splitter [62], or an imagedoubler to project the filtered images on a single monochrome CCD or CMOS sen-sor [4, 117, 160]. As an alternative approach, Larsson [83], proposed extracting therequired spectral information from the three color channels of a single color CCDcamera using the Bayer filter mounted on the sensor (instead of the narrow band-pass filters); however, this method is susceptible to overlapping response of thecolors bands, which results in an untraceable signal interference (cross-talk). This30technique is explained in detail by Svensson et al. [141], and has been applied byseveral researchers [29, 158]. As an extension to the described application of acolor CCD camera, Ma et al. proposed a color-ratio pyrometry (CRP) techniqueusing a color CCD camera, where ratios of the signals obtained at three visiblewavelength channels (red, green, and blue) were used rather than the absolute sig-nal values. This was done as an attempt to avoid the challenges associated withquantitative absolute intensity measurements, such as calibration errors, opticalpath variations, and window fouling [92]. It was found, however, that better tem-perature estimates could be achieved assuming a constant KL, rather than solvingfor KL. The calculated temperatures using this method were noted to be insensitiveto the soot concentration as long as the signal ratios (red/green vs. red/blue) wereadaptively selected based on temperature ranges being considered.The uncertainties associated with pyrometry have been discussed in a com-prehensive review by Zhao and Ladommatos [162], as well as other works con-sidering the complex refractive index, and the shape of the soot agglomerates[30, 81, 95, 96, 147]. Furthermore, uncertainties due to wall reflections in thenear infrared spectrum (NIR), soot deposition on the window, wavelength sensi-tivity [95, 96] and digitization errors [117, 160] have been reported. Payri et al.[117] discussed the effect of the intensity non-uniformity along the line of sight,as well as the effect of digitization uncertainty on a CCD sensor. Line-of-sight un-certainties are most significant for flames with optically thick regions (KL¿1) [68],and higher gradients [108], particularly along the line of site. Correspondingly, theuncertainties are lower for homogeneous combustion strategies, lower temperaturecombustion strategies (due to generally lower gradients) [108], or later in the com-bustion process when the soot is more uniformly mixed [68]. Table 2.2 summarizesthe uncertainties relevant to two-color pyrometry as characterized in the literature.In general, the soot volume fraction and KL factor are more sensitive to uncertain-ties than the temperature. Furthermore, the uncertainties are more significant forlower soot concentrations.In a recent study, Kamimoto et al. [68] also investigated the presence of sootoutside the naturally luminous flame boundary, to identify soot cloud regions attemperatures below the detection limit. This was carried out using simultaneousnatural luminosity and shadow images of a single diesel spray jet flame within31Table 2.2: Summary of reported two-color pyrometry uncertaintiesErrorReferenceSource of 10 ppm; 0.1 ppm;uncertainty 700-800 nm range 700-800 nm rangeTemperature fvL* Temperature fvLSoot particlesize distribu-tion function<1% 2% <1% 2%Complexrefractiveindex<1% 8% 2% 12%di Stasio(1994)Soot particleshape<1% 18% 2% 18%High soot conc. Low soot conc.Temperature KL Temperature KLWall reflections(visible)<<1% 2% <<1% 10%Matsui (1980)Wall reflections(NIR)5−8% 30% 5−8% 50%High-T combustion Low-T combustionTemperature fvL Temperature fvLLine-of-sightgradienteffects8% 50% 3% 20%Musculus(2008)Temperature KLSoot depositionon the window1% 5% Matsui(1980)Pyrometry filterband-width1% 2% di Stasio(1994)Digitizationerror5% 20% Payri(2007)* fvL: product of the mean soot volume fraction and line-of-sight path length.This is a measure of the mean optical thickness of the soot cloud [41, 147].32Figure 2.4: Shadow and NL images of a single spray flame in a constant vol-ume combustion chamber (TASI: time after start of injection); operatingconditions available in the reference. Adapted from [68] with permis-sion from publishera constant volume combustion chamber, as shown in Figure 2.4. Generally, onecan claim the natural luminosity and soot cloud shadow regions appear to coincideacross the field-of-view. However, quantitative evaluation of the shadow imageswas not possible due to the significant background noise caused by the densityfluctuations of the surrounding gaseous species at elevated temperatures [68].Application in in-Cylinder Soot CharacterizationMulti-color pyrometry has been used by numerous researchers to study soot pro-cesses in conventional and alternative combustion strategies. For instance, in-tegrated light pyrometry was used in conjunction with high-speed spatially re-solved natural luminosity imaging by Singh et al. to demonstrate lower amountof in-cylinder soot resulting from representative low-temperature diesel combus-tion (LTC) scenarios, compared to their conventional high-temperature combustion(HTC) strategies [136]. Also, it was inferred from the results that, during high-temperature operating conditions soot formation starts upstream in the jet, whilefor the low-temperature operating conditions, the start of soot formation appearedto occur further downstream, closer to the bowl edges. Two-color pyrometry re-sults showed the onset of in-cylinder soot after the premixed burn and during themixing controlled combustion phase. Figure 2.5 shows the calculated fvL factor33Figure 2.5: fvL, HRR, NL, soot volume, and NL images for high-temperaturediesel operating condition (the number to the upper right of each imageis the crank angle aTDC). Adapted from [133] with permission frompublisherresults for a high-temperature operating condition studied in [136]. Also shown inthis figure are HRR, spatially averaged natural luminosity results, and a calculated“soot volume” parameter at each crank position, based on the soot cloud area indi-cated by the 2D natural luminosity images, as well as select instances of spatiallyresolved natural luminosity results. Although the detected onset of soot from NLand two-color pyrometry results are coincident, around peak HRR of the partially-premixed combustion phase, the peaks and trends in the two sets of results do notgenerally agree. The peak NL occurs closely after the peak HRR of the mixing-controlled combustion phase and end-of-injection (3 CAD aTDC), while the peakin the fvL factor was observed later during that phase.In a direct injection spark ignited (DISI) engine, Ma et al. [92] used a vari-ant of multi-color pyrometry technique to characterize soot and observed maxi-mum in-cylinder temperature, based on spatially averaged soot temperature distri-34bution, and coincidentally maximum heat release rate just after TDC. In a differentstudy, Zha et al. [158] used two-color pyrometry to compare the soot processes inbiodiesel blend and diesel combustion in a CI engine. For biodiesel blends, theyobserved soot formation in low amounts over large areas between fuel jets, whichwas associated with more splashing in the fuel during wall impingement and re-bounding fuel distribution over a thicker annual ring. Higher amounts of KL wasformed on the jet axes and away from the bowl edge, which was attributed to theremnants of the fuel spray. Diesel combustion, in contrast, showed soot formationexclusively at the piston bowl regions through what was described as pool flamesresulting from spray wall impingement. In comparison to the diesel fuel, the testedbiodiesel showed earlier and significantly higher soot formation, but also highersoot oxidation rates, inferred from lower late-cycle KL values. In addition, simi-lar to diesel combustion, application of higher fuel injection pressure for biodieselshowed to reduce late-cycle soot KL levels.In a more recent study, Kamimoto et al. [68] used KL factor distributions fromtwo-color pyrometry to study late-cycle soot oxidation in a heavy-duty diesel en-gine and an RCEM. In correlation with simulated soot mass traces, they specifieda surface-specific soot oxidation rate in each case to comment on underlying ox-idation mechanisms based on the Arrhenius diagram. They calculated a highersoot oxidation rate for combustion in the heavy-duty diesel engine than that of theRCEM, and, based on a soot oxidation model presented by Boba et al. [14], con-cluded that this higher soot oxidation rate could be attributed also to the well-mixedsoot oxidizer structure in the flames, in addition to the higher soot temperatures.This was in contrast with the hypothesis proposed by Huestis et al. [60], whichsuggested in-cylinder soot temperature was more important than jet-induced tur-bulent mixing during the late-cycle soot oxidation processes. Figure 2.6 showssample two-color pyrometry results from [158] and [68], as well as their filterednatural luminosity counterparts (raw images). The resolved distribution of tem-perature and KL factors presents valuable information regarding the relevant sootprocesses; however, there are regions in the raw data that have not been resolvedinto KL and temperature values, despite the available high soot signal.35Figure 2.6: Sample 2D two-color pyrometry results and the correspondingfiltered NL images from (a) diesel combustion in an RCEM [68] and (b)biodiesel combustion in a diesel engine [158]. Image sequences wereadapted from references with permission from publishers2.2 Modern Alternative Combustion StrategiesAn overview of the alternative natural gas combustion strategies was presentedin Chapter 1. Here, a review of the current understanding of these processes isprovided and the areas with a potential for improvements are emphasized.2.2.1 DIDFDIDF engine operation is an attractive strategy for the application of natural gas inconventional CI engines because of the minimal modification to existing engine de-signs necessary and the relatively high efficiencies possible [130, 153, 154]. DIDFengine operation is characterized by premixing a high octane number fuel and thenigniting it with a late-cycle direct injection of a high cetane number pilot fuel.Typically, the bulk of the energy is supplied by combustion of the premixed fuel.36Despite the significant reduction of the NOx, CO2, and PM emissions, whichoriginally made DIDF a promising combustion strategy, this technique poses a ma-jor deficiency by increased emission of unburned combustion species such as COand CH4 (when NG is used for the premixed charge), particularly at low engineloads. Higher CH4 emission is particularly concerning, given it was shown to pos-sess 21—72 times the greenhouse warming effect of CO2, and has made the low-load DIDF engine operation a more intense field of ongoing research [44, 124].Emissions of unburned species can be related to the distinct combustion modespresent throughout the dual-fuel combustion process [69, 127]. These combustionmodes and the transitions between them are complex and not well understood. Atthe same time, the combustion behavior within dual-fuel engines have been shownto be highly dependent on the mass, timing, pressure, and fuel composition of thepilot injection, all of which can readily and predictably controlled [7, 127]. Karimintroduced a model, where dual-fuel combustion was broken down into three dis-tinctive stages: premixed combustion of the pilot fuel, immediately followed bycombustion of the entrained premixed charge at the vicinity of the ignited pockets,and then conversion of the rest of the premixed charge through a turbulent flamepropagation process [69]. In fact, Karim looked into the influence of premixedequivalence ratio on the shape of HRR curves to define different stages of com-bustion based on dominating fuel component in the charge in the proportion ofreleased energy. With a similar approach although including ignition delay stagesfor both the pilot and primary fuels, Sahoo et al. [130] divided the DIDF combus-tion process into five stages: Pilot ignition delay followed by the premixed pilotfuel combustion, the NG ignition delay stage as a result of the higher auto-ignitiontemperature of methane, rapid combustion of the primary fuel, and finally, a latestage low-intensity diffusion combustion stage as a result of the decreasing com-bustion reactions rates with expanding cylinder volume and combustion diluents.In another work, Rochussen et al. [127] divided the dual-fuel combustion processinto two main stages based on shape of HRR curves. The first stage was foundto be dominated by auto-ignition and rapid combustion of pilot fuel and entrainedCH4. This was then followed by the second stage, where conversion of premixedNG was regarded as the dominating process.While such thermodynamic and emission based investigations can provide a37description of the combustion process, optical investigations of in-cylinder pro-cesses provide significantly more of these descriptions. There are a limited numberof studies that have optically assessed DF combustion. These include visualizationof the distribution of high temperature reaction zones and their correlation to un-burned hydrocarbon emissions; spectroscopic analysis of the reactive charge, in-termediate species and end products; visualization of soot particle production andoxidation [18, 33, 133].Carlucci et al. [18] investigated the effect of inlet bulk flow, port-fuel injectorpositions, pressure and quantity of the diesel pilot injection on dual-fuel combus-tion at light load using single-shot endoscope natural luminosity imaging. Theport-fuel injector positioning was modified to control charge mixture stratifica-tion and inlet port setup was modified to control swirl and tumble in the inducedcharge. Their analysis of the averaged NL curves showed a great influence fromthe injected diesel pilot properties under various conditions. In particular, lowerpilot fuel injection pressure and higher injected mass, both led to higher recordedNL signal. Schlatter et al. [133] studied diesel pilot ignited dual-fuel in an RCEMby looking at spatially resolved OH*, and integrated OH*, CH*, C2*. The focus ofthis work was to study the effect of sweep of methane equivalence ratios (0-0.65)and changes in the number of pilot jets (3 or 6) on the auto-ignition and com-bustion behavior, while the pilot injection pressure and timing was held constant(400 bar and 4.1 ms bTDC). Under both high and low load operating conditions(φCH4 = 0.37 and φCH4 = 0.67), the high temperature reactions were localized onthe pilot jet axes, denoted by the high intensity OH* chemiluminescence in theseregions, for both pilot injectors. Figure 2.7, reprinted from [133], illustrates thisprocess. In this work, authors also observed an increase in ignition delays, asindicated by first measurable OH* signal, with increasing amounts of premixedmethane.Dronniou et al. [33] focused on the effect of the premixed charge equivalenceratio using time-resolved natural luminosity and single-shot OH* chemilumines-cence imaging in an optically-accessible 0.5L single-cylinder engine. Single-shotPLIF-tracer1 imaging under equivalent motored conditions was also considered to1Planar Laser Induced Fluorescence of a tracer species added to pure nitrogen instead of air38Figure 2.7: OH*-chemiluminescence image sequences for high-load (left)and low-load (right) DIDF operation in an RCEM, with 6 and 3 pilotjets, respectively. Adapted from [133] with permission from publisher.characterize spatial distribution of the pilot fuel and provide complimentary infor-mation during the ignition process. The major goal of the work was to identifyprincipal combustion regimes characterizing the dual-fuel combustion processes.More specifically, whether the DIDF combustion can be described by auto-ignitionor sequential auto-ignition, flame propagation, diffusion, or a combination of thesecombustion modes. NL and OH* chemiluminescence results showed that the ig-nition consistently occurred in the bowl periphery throughout a sweep of globalequivalence ratios (φglobal) in the range 0.33-0.94, while the injected pilot mass washeld constant. From there, the high temperature reactions show increasingly moreintense OH* chemiluminescence and develop further towards the central bowl re-gion. Although it was acknowledged that differentiating sequential auto-ignitionfrom flame propagation would be challenging, the reaction zone growth was inter-preted as flame propagation based on experimental and numerical laminar flamespeed analyses for methane-air mixtures [32] and repeating experiments with alower octane fuel (iso-octane) [33]. The extent of reaction zone growth towards thecenter was observed to correlate with the equivalence ratio. Full growth and ho-mogeneous high intensity OH* chemiluminescence was recorded at φglobal = 0.94,while a notable part of the bowl area was left with no OH* signal detected atφglobal = 0.56. The described reaction zone growth comparison is shown in Fig-ure 2.8, reprinted from [33]. It was concluded that the observed higher unburned39Figure 2.8: OH*-chemiluminescence image sequences for low-load (left)and high-load (right) DIDF operation with constant pilot mass. Adaptedfrom [33] with permission from publisherhydrocarbon emissions during fuel-lean DIDF operating conditions, is to a greatextent, an artifact of the reaction zone not being able to propagate further throughthe globally leaner mixture and reach the central bowl regions.The DIDF combustion mode illustrated in Figure 2.8 significantly differ fromthose described by Schlatter et al. in Figure 2.7, where the flame propagation ap-pears to initiate around the injected pilot jets, despite the similar injection pressuresand injected masses. This discrepancy is further discussed later in Chapter 4. In ad-dition to the more conventional DIDF operations, optical investigations on alterna-tive dual-fuel applications are available. Florea investigated ethanol and n-heptanedual-fuel partially-premixed combustion using time-resolved NL and single-shotOH* measurements [42] and Mancaruso et al. studied bio-ethanol and diesel dual-fuel using single-shot spectroscopic measurements [93].2.2.2 PIDINGThe Pilot Ignited Direct Injected Natural Gas (PIDING) combustion strategy, is acommercially viable technology that addresses the common problems associatedwith conventional dual-fuel combustion[33, 75, 80, 153], while matching perfor-40mance of conventional diesel combustion. The PIDING injector uses concentric,hydraulic-servo actuated needles to control both natural gas and diesel injectionand enables diesel pilot ignited, direct injected natural gas combustion. Near topdead center, high pressure natural gas (NG), which constitutes approximately 95%of the total energy, is injected and ignited by the combustion of a small amount ofdirect injected diesel. In a typical operating condition, the diesel pilot injection andignition precede the NG injection. This combustion approach has been commer-cialized by Westport Fuel Systems Inc. as High-pressure direct-injection (HPDI)of NG [112].Previous understanding of the impacts of the pilot ignition and subsequent fuel-air mixing and combustion process of the gas jets have depended on experiments[34, 38, 97, 99, 100], or CFD simulations [54, 90, 114]. PIDING experimentalinvestigations have been based on conventional thermodynamic analyses using in-cylinder pressure traces and engine boundary conditions, or engine-out emissionsmeasurements of soot and various gaseous species. Based on these extensive sur-veys, the PIDING strategy is perceived as being predominantly governed by non-premixed NG combustion, preceded by a fraction of premixed NG combustionbefore transition to the non-premixed stage, similar to conventional diesel combus-tion [38, 99]. Fraction and mode of the premixed NG combustion phase [38], deter-mined by engine control parameters such as combustion timing and relative delaybetween the pilot and NG injection [99, 100], and fuel rail pressure and combustionduration [34, 98], have been characterized as the major factors that could be “tai-lored” to control engine-out PM emissions, while also affecting CO, tHC and NOxemissions during PIDING combustion. The relative pilot and NG injection timing,and the interlace angle between the pilot and NG nozzle holes, was reported todetermine the state and locus of the ignited pilot kernels and the corresponding in-teraction with the NG jets, significantly impacting the NG ignition processes [97].Ultimately, the variations to the ignition processes impact soot formation, oxida-tion, and engine-out emissions [34, 38, 90]. Advancing the gas injection relativeto the diesel ignition can lead to a more premixed combustion strategy, especiallyat lower engine loads, reducing PM and CO emissions substantially at the expenseof increased tHC and NOx emissions [34, 99, 100]. As in conventional diesel en-gines, high thermal efficiencies can be achieved through increasing the fuel rail41Figure 2.9: Combined 15-cycle ensemble averaged 2D images of OH*chemiluminescence and NL from PIDING combustion. Adapted from[54] with permission from publisherpressure, which results in higher NG injection rates and more rapid combustionprocesses [34]. Higher fuel rail pressure is reported to substantially improve fuelconversion efficiency and particulate oxidation through enhanced mixing after theend of the injection event, leading to improved efficiency and reduced PM emis-sions [98]. However, the more rapid non-premixed burning rates will also producehigher flame temperatures and was observed to increase NOx emissions [34].To date, only one optical investigation of PIDING combustion has been pub-lished [54]. In this work, simultaneous 2D OH* chemiluminescence and filterednatural luminosity measurements were used to analyze the mixture formation andignition of the gas jets, and to validate numerical simulation of these processes. A2.13 L single-cylinder engine with a 17.3 compression ratio, available in both ther-modynamic “all-metal” and optical configurations with an omega-shaped quartzglass piston window, was used in this work. PIDING combustion was brokendown into four combustion phases of diesel pilot combustion (1), ignition of gasjets at diesel pilot zones (2), main combustion of natural gas with main heat release(3), and soot post-oxidation (4). Figure 2.9 shows an image sequence of PIDING42combustion produced based on combined 2D ensemble averaged images (from 15cycles) of the OH* chemiluminescence and natural luminosity. More detailed dis-cussion of the these results, in conjunction with the findings of the work presentedherein, is presented in Chapter 6. After validating their simulation methodologyagainst thermodynamic engine measurements of the pressure traces, heat releaserates, and “end of engine” soot, Hatzipanagiotou et al. compared local combus-tion features and observed a good agreement between optical measurements andsimulations.2.3 Summary and Literature GapThe complexities and intricacies of the phenomena involved in diesel-ignited dual-fuel combustion strategies go well beyond the already analytically challengingphysical and chemical processes occurring in conventional diesel engines. Theport fuel injection of NG during DIDF, for instance, significantly changes the com-pression processes and the properties of the premixed charge and, consequently,the combustion chamber conditions at the time of direct injection of the diesel fuel.This has been shown to influence the pilot fuel ignition processes [91, 115], which,as a result, significantly affects the combustion of the premixed charge itself andthe resulting gaseous and PM emissions [69, 115, 154]. Also, while the PIDINGcombustion strategy is generally perceived to have similarities with diesel engineoperation characteristics and can achieve diesel-like efficiency with lower PM andgreenhouse gas emissions[34, 112], significant differences in the soot processesare expected to result from replacing the liquid diesel sprays with non-premixedgaseous jets.The highly complex and interrelated chemical and physical processes occur-ring during these combustion processes demands for application of a wide rangeof experimental and numerical investigations in order to provide more insight intoand better characterize these seemingly perplexing mechanisms. This better under-standing would be imperative for making improvements to the current designs withthe goal of surpassing the performance of existing technologies and replacing themwith more efficient technologies with less pollutant emissions. The conventionalthermodynamic analysis of in-cylinder pressure traces and engine boundary condi-43tions, analysis of engine-out emissions measurements, and CFD simulations, havealready provided invaluable information regarding DIDF and PIDING combustionstrategies, as reviewed in §2.2. In the case of PIDING combustion, these experi-mental investigations and numerical simulations highlighted some of its fundamen-tal characteristics differing from those of conventional diesel and DIDF combustionstrategies.As also elucidated in §2.2, optical measurements of the combustion in-cylinderphenomena offer a unique perspective into the problem, providing a more in-depthanalysis of the information that is otherwise inaccessible. Nevertheless, limitednumber of optical investigations of these combustion processes and their controlparameter effects are reported in literature (discussed later in this chapter), andmany aspects of their conceptual behavior are yet to be studied and characterized.As an example, discrepancies were observed in the ignition processes and reactionzone growth behavior of the DIDF combustion in optical measurements by Dron-niou et al. and Schlatter et al., which is not explained [33, 133]. As such, a sectionof the current work has focused on addressing some of these unanswered ques-tions. More importantly, however, none of the available optical studies have imple-mented detailed spatially resolved quantitative measurements of soot concentrationand temperature distributions, using techniques such as the two-color pyrometry,to study in-cylinder conditions of NG combustion in diesel engines. Diagnostictechniques are of great significance for better characterization of the hypothesizedcombustion mechanisms developed based on prior experimental investigations andwould allow further optimization of the existing CFD tools.The examples reviewed in §2.1.4 illustrated the utility of multi-color pyrom-etry for in-cylinder soot analysis in conventional diesel engines. Closer examina-tion of the calculated soot temperature and KL factor distributions presented acrossthese studies partly reveals its challenges, appearing occasionally in the form ofregions with unresolved data, as shown in §2.1.4. Majorly contributing to the lackof such quantitative optical investigations for diesel-ignited dual-fuel operation todate is the additional challenges associated with the low-SNR optical signal avail-able from these and other low-sooting combustion strategies. Essentially, since thetwo-color pyrometry method is an unstimulated emission technique, its implemen-tation becomes more challenging at lower soot concentrations and temperatures44where the SNR decreases. This can be particularly problematic for strategies suchas Low Temperature Combustion (LTC), or PIDING combustion [112], where ex-haust stream soot is relevant but in-cylinder concentrations and temperatures arelower than in conventional diesel combustion. Some studies have applied the in-tegrated light two-color pyrometry technique to low-sooting, low-temperature ap-plications. Singh et al. [136] used this technique, along with high-speed spatiallyresolved natural luminosity imaging to characterize the lower soot concentrationsrepresentative of low-temperature diesel combustion (LTC). In a more recent study,Hatzipanagiotou et al. [54] used integrated light two-color pyrometry for analysisof PIDING combustion. Both of these works attempted to use 2D natural lumi-nosity as a means to roughly describe soot distributions. However, comparison ofthe integrated combustion light intensities at three wavelengths, where two wereused for two-color pyrometry and the third was used as a reference soot luminositysignal, in general, did not show a close correlation between natural luminosity andpyrometry soot signal trends [108, 136], as can be seen in Figure 2.5 ( fvL vs. com-bustion luminosity curves). Spatially resolved two-color pyrometry measurementsof these lower-soot combustion methods, however, is more challenging and has notbeen reported in the literature. As a major contribution, this work will demonstrateimprovements to the two-color pyrometry method to facilitate its application forlow-sooting combustion conditions. The improved method is then implemented toPIDING combustion and a conceptual understanding of the in-cylinder soot for-mation and oxidation processes is presented.The undeniable advantages and valuable information the spatially resolved op-tical measurements of combustion in-cylinder processes offer, however, is onlypossible at the expense of alterations to the engine operating conditions. As dis-cussed in §2.1.1, these changes are imposed by the necessary major modificationsto the combustion chamber geometry and measures are adopted to compensatefor them primarily through manipulation of intake air conditions and fuel injec-tion parameters to match charge conditions at TDC, combustion phasing, and theshape of the HRR profile. Nevertheless, a remaining valid concern is the discrep-ancies these intrinsically different engine operations might cause for in-cylinderoptical measurements and the extent that the optical measurements findings wouldbe relevant to more realistic and less idealized conditions of a thermodynamic en-45gine operation. Several studies have been performed on highlighting the need tocharacterise the differences between optical and all-metal engine measurements[5, 23, 72]. These include investigating the effects of piston surface temperaturedifferences, using cycle-resolved thermocouples in DISI engine [139] and materialproperties of optical components (windows), and the differing heat transfer char-acteristics [5], as well as simplified flat bowl vs. more representative toroidal bowldesign [79] and implementing (simulated/synthetic) EGR, on in-cylinder phenom-ena in the two engines. Study of implications of the described differences, andadditional factors such as dynamic loading of the Bowditch piston, on charge mix-ing, combustion, and emissions under LTC diesel operating conditions indicatedsignificant discrepancies and several necessary measures for valid comparison ofthe results [72]. The described complexities associated with spatially resolved op-tical measurements from in-cylinder combustion phenomena, motivates 0D opticaldiagnostics of these processes under all-metal engine configurations. The majordeficiency of such measurements is the lack of spatial attributes in the spatiallyintegrated light from a control volume whose geometry changes with piston move-ment. Thus, as another novel contribution, the current work focuses on charac-terizing the integrated light measurement signal from an optical probe design (see§2.1.1) through comparisons with simultaneously recorded spatially resolved mea-surements. The better characterized optical probe is then used for comparison ofthe pyrometric measurements of in-cylinder soot processes in optical and thermo-dynamic engines, which has bot been reported in literature.The characterized rugged optical probe, together with thermodynamic analysismetrics (e.g., HRR and fuel flow rates), and exhaust stream measurement tools,would provide a valuable “Thermo-Optical” analysis toolkit to study combustionin-cylinder processes. Such analysis toolkit could help optimize the operating pro-tocols and improve future designs of various combustion strategies.46Chapter 3Experimental SetupIn this chapter, a description of the experimental facilities and measurements sys-tems developed and utilized in this work are presented. These details elucidate themeasurement capabilities of the integrated diagnostics system as well as the opera-tional restrictions, the range of possible measurements, and limitations imposed tothe interpretation of the results.3.1 EngineThe experimental facility used in this work is a single-cylinder optical researchengine (Ricardo Proteus) with a Bowditch design. The geometry is representativeof a 2.0 Litre heavy-duty compression ignition engine. The core feature of thefacility is the ability to convert between two distinct configurations:1. Thermodynamic Configuration; which has a geometry similar to a conven-tional IC engine, also referred to as the “all metal” engine configuration.2. Optical Configuration; which provides large optical access into the combus-tion chamber, affording 2-D visualization of the combustion processes in thepiston bowl, at the expense of engine and bowl geometry modifications.Relevant specifications of the engine are provided in Table 3.1 and are com-pared to a typical heavy-duty engine.47Table 3.1: Optical engine specificationsRicardo Proteus Typical HD Diesel/NGType 4-stroke, compression ignitionNumber of cylinders 1 6Displacement 1998 cm3 2500 cm3Bore 130 mm 135 mmStroke 152 mm 170 mmMax. cylinder pressure170 bar (thermodynamic)180 bar117 bar (optical)Compression Ratio13.9:1 (thermodynamic)17:112.9:1 (optical)Max. speed2050 rpm (thermodynamic)2050 rpm1200 rpm (optical)Max. charge pressure 3 bar turbocharger limitedVisible bore 83 mm; ~40% of bore area N/AThe Ricardo Proteus engine design is based on the Volvo TD120-family of en-gines. These were 12-litre, in-inline 6-cylinder (I6), overhead valve, heavy-duty CIengines, originally designed for on-highway applications. Designed as a research-oriented single-cylinder engine (single head, cylinder, piston, and valve train as-sembly from the stock Volvo engine), with the capability to transition between thetwo described configurations, the Ricardo Proteus engine provides a platform forcombustion diagnostics focusing on the in cylinder processes of a heavy-duty CIengine. Under the “all metal” thermodynamic configuration, an eccentric toroidalbowl design is used and there is no major modifications to that of a conventionalCI engine, except for small diagnostic ports to allow in-cylinder measurements ofcertain combustion characteristics. As such, more realistic operating conditions,resembling those on-road applications, can be investigated. Under the optical con-48figuration the engine, a large optical access to the combustion chamber is affordedthrough use of a Bowditch piston design (see §2.1.1). This results in a square bowldesign as a result of the installed flat quartz window. Although this optical accessopens up extensive opportunities for acquiring more insight into combustion in-cylinder processes through spatially and temporally resolved optical diagnostics,the necessary major modifications to the engine geometry inevitably introducesrestrictions to the range of operating conditions, operating protocols, and interpre-tations to be made through the analysis of the results. This is discussed in furtherdetail in §3.3 and § Engine Facility InstrumentationThe described research engine and the test-cell it is operated in are equipped withmany auxiliary systems and diagnostic tools. Many of these sensors are responsi-ble for monitoring engine health and test-cell safety and not utilized combustionanalysis (e.g, thermocouples, pressure sensors, dyno speed sensors, and chemicalalarms). Sensors that were crucial in the collection of combustion related enginedata are listed in Table 3.2 and details of how they are connected to and controlledby data acquisition and engine control unit systems are available in [156].Table 3.2: Engine sensors used in combustion analysis.Measurement Sensor DescriptionMethane mass flow rate Endress & Hauser Promass 80A Coriolis Me-terDiesel mass flow rate Mettler Toledo Viper Ex MB SM12 Gravi-metric ScaleIntake mass airflow rate Bosch OEM Hot Film SensorIntake manifold absolute pres-sure transducerKistler 4005B Piezo-resistive Pressure Sen-sor; Kistler 4618A AmplifierIn-cylinder pressure transducer Kistler 6125C Piezo-electric Pressure Sensor;Kistler 5010B Charge AmplifierEngine position and crank-synchronous timingBEI H25 Incremental Quadrature Optical En-coder; Hamlin 55505 Hall Effect Sensormounted on flywheel; Hamlin 55505 Hall Ef-fect Sensor mounted on cam-gear49As discussed in the previous chapter, the developed optical diagnostics in thework presented herein were designed for and implemented to DIDF and PIDINGcombustion strategies (see §2.2). For DIDF combustion experiments, port injec-tion of methane was executed via a Bosch NGI2 injector and pilot diesel fuel wasdirectly injected using a Bosch CRIN2 injector. For PIDING combustion exper-iments, a research-level High-Pressure Direct Injection (HPDI) injector was usedfor direct injection of Natural Gas (NG) and diesel (separate internal fuel lines).This injector was supplied by Westport Fuel Systems and is derived from theirproduction first-generation HPDI fuel system. More details regarding this injectordesign and operation was presented in §2.2.National Instrument’s Powertrain Control hardware was used to control the fuelinjection system. The control hardware is based on a field-programmable gate ar-ray (FPGA) and uses the NI cRIO-9068 (CompactRIO) chassis with selected Pow-ertrain Control modules as hardware drivers for the various sensors and actuatorson the engine. A simplified process and instrumentation diagram for the PIDINGcombustion experiments under optical configuration of the engine is presented inFigure 3.1.The commissioning of the Ricardo Proteus engine under thermodynamic con-figuration and the first major research investigations conducted using this enginefacility was performed prior to this work and recently reported by other groupmembers [124, 156]. A complete description of the facility can be found in [124].This work focused on the development of the spatially and temporally resolvedoptical measurement (i.e., high-speed imaging) systems.50Figure 3.1: Experimental facility process and instrumentation diagram forPIDING combustion experiments under optical configuration.513.2 Imaging SystemsA major focus of this work was to develop spatially and temporally resolved opticaldiagnostics, integrate them with the existing engine facility, and discuss the mea-surement results in conjunction with the thermodynamic and spatially-integratedoptical measurement results. As such, preparations for three different classes ofspatially resolved optical measurement techniques, namely natural luminosity, chemi-luminescence, and two-color pyrometry, were considered. This was followed bythe calibration of the optical probe in order to better characterize its spatially-integrated signal through comparisons with the spatially resolved measurements.Ultimately, these comparisons would help extend the interpretations that can bemade from the results obtained through restricted optical access available underthermodynamic configuration of the engine.Figure 3.2: Schematic of thermodynamic and optical configuration of the Ri-cardo Proteus engine and imaging system setup on the optical engine.52Table 3.3: Optical Imaging EquipmentHigh-Speed CamerasCamera Phantom V7.1 Photron SA-1Sensor design CMOS CMOSSensor ISO 4,800 10,000Bit-depth 12 12Max. resolution 800×600 pix2 1024×1024 pix2Pixel size 22 µm 20 µmMax. frame rate (full res.) 4,500 fps 5,400 fpsMin. exposure duration 2 µsec 1 µsecSpectral response 200 nm - 1,100 nm 200 nm - 1,100 nmPaired objective lens 60 mm-f/2.8Micro-Nikkor98 mm-f/2.8 CercoUVTrigger signal input TTL TTLDigital interface Gigabit Ethernet Gigabit EthernetPrimary lens mount F-mountImage IntensifierDesign LaVision two-stage high-speed IROSpectral response 200-900 nmMin. gate 100 nsMax. operation frequency 2 MHzMax. MCP amplification 1000×Image size 25 mmCamera Coupling 1:1 optical lens couplingPhotocathode S20Phosphor screen P46The thermodynamic and optical configuration of the Ricardo Proteus engine,together with simultaneous imaging equipment setup, are schematically shown inFigure 3.2 and imaging system general specifications are presented in Tables 3.3and 3.4. Details of individual components of the imaging equipment and the inte-grated optical measurement systems are discussed in the next section.53Table 3.4: Imaging Systems AssembliesImagingtechniqueSensing OpticsLight CollectionOpticsOptical Filters2D naturalluminosityPhantomcameraNikon lens optional HP filter2D two-colorpyrometryPhantomcameraNikon lensand doubler700 & 800 nm CWL;10 nm FWHM2D OH* CL Photron & IRO Cerco UV lens307nm CWL & 20nmFWHM (compound)0D two-colorpyrometryPDA36aSapphire rod lens,fibre optics andcollimators700 & 800 nm CWL;10 nm FWHM3.2.1 Natural Luminosity ImagingDetails of the NL imaging technique was discussed in §2.1.2. A Phantom V7.1high-speed camera, equipped with a 60 mm f/2.8 Nikon objective lens (Micro-Nikkor AF-D), was used for NL measurements in this work. The camera has a12-bit monochrome CMOS sensor with 800×600 pixel resolution and 22×22 µm2pixel area. Details of high-speed camera setup considerations are provided in Ap-pendix A.3.2.2 Two-Color Pyrometry ImagingThe two-color pyrometry technique fundamentals were discussed in §2.1.4. Thesame camera and objective lens system used for NL imaging was also used fortwo-color pyrometry measurements. An image doubler design and an integratingsphere are the other two essential elements of the two-color pyrometry imagingsystem. In what follows details of these two pieces of equipment are discussed.Image DoublerThe two-color pyrometry measurement principles discussed in §2.1.4 elucidatedthat in order to perform two-color pyrometry measurements using a single monochromecamera, two simultaneous images of an instantaneous combustion event should beprojected onto the camera sensor. This was carried out in the current study using54Figure 3.3: Image doubler design and two-color pyrometry systemschematic.a LaVision image doubler, mounted with a stand-off distance in front of the Phan-tom V7.1 camera, equipped with the 60 mm f/2.8 Nikon objective lens. The imagedoubler is an optical assembly that duplicates the original image and projects themonto the sensor side by side using adjustable flat and prism mirrors. Figure 3.3shows a schematic diagram and a photo of the optical assembly for the two-colorpyrometry imaging system. The pair of flat mirrors can be horizontally and verti-cally adjusted for proper projection of the individual duplicated images. The prismmirror sits on a rail parallel to the optical axis, which allows for controlling thelight interference from one channel to the other (referred to as the cross-talk be-tween channels). A pair of narrow band-pass filters are mounted onto the channellight input ports, to view each image within a different bandwidth.55Figure 3.4: Chemiluminescence Imaging system setupIntegrating SphereThe purpose of the two-color pyrometry imaging system is providing instantaneousmeasurements of the apparent temperatures of the combustion at the two wave-lengths being considered. This is executed through quantifying the spectral radiantintensity radiated from the combustion flames. Thus, it is necessary to calibrate theresponse of the sensor to the incident spectral radiant intensity. The instantaneousspectral radiant intensity Iλ (xi, j, t), which is the required input to the two-colorpyrometry equation 2.4 (see §2.1.4), is evaluated from the recorded images usinga reference characterized light source. A high-intensity integrating sphere setup,custom design by LabSphere, comprising a 150W NIST traceable calibrated tung-sten halogen lamp controlled by a precision regulated DC current power supplyand a 200 mm diameter integrating sphere with a 50 mm light port is used as thecalibration light source in this work. The calibration system output is monitored bya radiometer during calibration. A schematic diagram and photo of the integratingsphere are shown in Figure 3.4.The cone baffle and the internal Lambertian surface design of the integratingsphere provides a uniform light output, with 98% uniformity across its exit port.The large light port provides illumination of a large part of the sensor area at onceand enables the sensor response uniformity to be characterized.During the calibration process the integrating sphere is mounted on the op-56Figure 3.5: The two-color pyrometry calibration setup (left) and the integrat-ing sphere mounting structure (right).tical engine block in place of the cylinder head and the response calibration isdetermined with all relevant optical components in the light path. The integrat-ing sphere and the mounting structure are shown in Figure 3.5, along with thetwo-color pyrometry calibration setup. To assess the sensor non-uniformity, theintegrating sphere is translated such that the entire field-of-view of the two-colorpyrometry measurement system is illuminated. At each position, an average of 30images is considered to eliminate the effect of shot noise and is spatially averagedover a smaller area at the center of the sphere target image to avoid potential pixelbias.The integrating sphere was selected over commonly used tungsten-ribbon lamps[64, 117], as the latter provides a small calibration target (1.6x20 mm2 [117]), withonly slightly higher spectral radiant intensity levels (~5e3 vs. 8e3 W/m2.sr.µm).Other lower intensity light sources have also been used, including tungsten lightsources [52, 92, 160] and low intensity blackbody sources [158, 160]. Thesesources consequently require more significant extrapolation to combustion intensi-57ties under consideration.Despite using a high intensity light source for calibration, the maximum spec-tral radiant intensity provided by the integrating sphere was still lower than that ofthe combustion process (e.g., 9.5e4 W/m2.sr.µm maximum intensity was detectedin this work), as is typically the case with two-color pyrometry system calibra-tion [52, 64, 92, 117, 158, 160]. During preliminary engine combustion imaging,optimum objective lens aperture to be used for both the calibration and combus-tion imaging was identified and sensor exposure duration was adjusted to properlyrecord higher intensity combustion images with minimal saturation. Although thelens aperture setting (f-number) can be incorporated into the signal calibration pro-cess, bench top measurements showed that the factor of two estimation in lightintensity change due to light-stop increments is an approximation that introducessmall, yet preventable errors in the calculated results. The ratio of the sensor ex-posure durations for calibrating integrating sphere signal and the experiment wasused as the scaling factor for the spectral radiant intensity extrapolation in the cal-ibration curves (Xτ ).In order to calibrate the camera sensor response, measurements from a sweepof characterized spectral radiant intensities are required to obtain calibration curvesat the two pyrometry channels. This can be achieved through selective attenuationof the integrating sphere signal using neutral density filters with a range of op-tical densities, without altering the spectral information of the incident light (see§3.2.4). However, bench top measurements suggest that exposure duration timecan be used to mimic physically attenuated integrating sphere signal through useof neutral density filters. This would significantly expedite the sensor response cal-ibration process, typically performed prior to a two-color pyrometry measurementcampaign. Figure 3.6 shows an example of a response calibration curve acquiredusing the integrating sphere and also compares the two different approaches to per-form the required reference intensity sweep as described above. As can be seenhere, the use of camera exposure for sensor response calibration also allows fora high resolution calibration curve, which can be used to examine the linearity ofthe correlation between the illuminating light and recorded charge on the CMOSsensor. The final stage in the sensor response calibration process is attributing therecorded intensity values to characterized spectral radiant intensity values of the580 50 ; 22 100 ; 44 150 ; 66 200 ; 88 250 ; 111Exposure [µs] ; Transmission [%]050010001500200025003000350040004500Grey Tones (12 bit)camera exposure data pointsND filter data pointsOD-0.2OD-0.3OD-0.4NoFilterOD-0.1OD-1.0OD-0.6Figure 3.6: Sample response calibration data points obtained for 800 nm two-color pyrometry channel from sweep of camera exposure as well asusing a set of ND filters. The “no filter” data point was recorded at225 µsec camera exposure.integrating sphere (see Chapter 5).3.2.3 OH* Chemiluminescence ImagingDetails of the chemiluminescence imaging technique was discussed in §2.1.3. Inthis work, OH* chemiluminescence measurements were carried using a PhotronSA-1 high-speed camera, equipped with a high-speed image intensifier and a 100mm f/2.8 Cerco® UV objective lens. This camera holds a 12-bit monochromeCMOS sensor with 1024×1024 pixel resolution and 20×20 µm2 pixel area.As discussed in §2.1.3, chemiluminescence imaging of the combustion inter-mediate species targets relatively low intensity signals within small spectral band-widths around certain wavelength peaks. A fraction of the already weak signal isalso lost to attenuation in the optical path in the processes of guiding it towardsthe imaging sensor (engine window and mirror) and restricting the incoming light59to the desired spectral range (see §3.2.4). This necessitates a significant increasein the SNR and a shift the target wavelength before the signal can be recorded atthe spectral and ISO ranges of the high-speed CMOS sensors. Image intensifierdesigns, also referred to as the intensified relay optics (IRO), can fulfill this task.These are electro-optical devices, which in a conventional design comprise threemain stages:1. A photocathode plane, which when illuminated, transfers the incident photonquanta to the electrons and, as such, converts the light rays into electronbeams.2. A micro channel plate (MCP); which amplifies the signal as the electronspass through its numerous microscopic high-voltage semi-conductor chan-nels, acting as photomultipliers.3. A phosphor screen, which re-converts the amplified electron beams back tolight, after accelerating them using a high potential applied to it.For this setup, the objective lens focuses the image onto the photocathodeplane. The photocathode, in addition to photon-electron conversion, serves as avery fast global electronic shutter as it is connected to a high-voltage pulser, af-fording exposure duration values (referred to as “the gating”) in the order of a fewnanoseconds. The electron beams are then amplified traversing the high-voltageMCP, where the magnitude of electron amplification is controlled by the adjustablepotential across the MCP (referred to as “the intensifier gain”). The electrons leav-ing MCP are accelerated towards the phosphor screen by a fixed bias voltage ofseveral kV, to increase electron-photon conversion efficiency.In this work, a LaVision high-speed two-stage IRO, capable of relaying in-tensified images at high rates of 2 MHz, was used. The output image from theIRO was guided to the Photron SA-1 sensor for OH* chemiluminescence measure-ments. Given the complexities involved in the photocathode-intensive two-stagefocusing process of the chemiluminescence imaging system, the camera and IROwere fixed on a designed adjustable base after calibrated for the best intensifiedimage focus. A schematic diagram and a photo of assembled chemiluminescenceimaging system are shown in Figure 3.760Figure 3.7: Chemiluminescence Imaging system setup3.2.4 Spectral OpticsAs discussed in §2.1.3 and §2.1.4, depending on the considered application, thebroadband light from the combustion event often is required to be restitched to acertain spectral bandwidth. This is achieved through spectral filtering of the incom-ing combustion light before it reaches the camera sensor, through use of short-pass,long-pass, and narrow band-pass (NBP) filters. These filters have special coatingsthat would selectively transmit, absorb, and reflect wavelengths of incident light.For natural luminosity measurements, depending on the case under consideration,either the broad-band unfiltered light was recorded or a long-pass filter with 575 nmedge wavelength (Omega Optical-575HLP) was considered. The latter would focusthe measurements on the incandescent soot particles and eliminate the majority ofthe chemiluminescence signal (e.g., from OH*, CH*, CN, Swan bands; see §2.1.3).In the case of two-color pyrometry, a pair of hard coated narrow band-pass filterswith 700 nm and 800 nm central wavelength (CWL) and 10 nm FWHM bandwidth61(Edmund Optics) were used. These filters have optical density of more than 4 (OD-4) for out-of-band rejection (200-1200 nm range) to avoid signal interference fromoutside the pass-band. Equation 3.1 describes the relation between filter opticaldensity (OD) and transmission (Tfilter[%]).OD =− logTfilter/10010 (3.1)For OH* chemiluminescence measurements, a compound filter set comprisinga narrow band-pass filter with 310 nm CWL and 20 nm FWHM (Chroma; CT31020UV NBP) and a custom band-pass filter with 330 nm CWL and 80 nm FWHM wasused (Omega Optical; based on 330WB80 with higher transmissivity at 307 nm).The narrow band-pass OH* filter (Chroma) provides excellent transmissivity inthe pass-band and OD-6 out-of-band rejection in the 200-390 nm range, but suffersfrom interference after 400 nm (low out-of-band rejection); in order to block thisspectral interference region, the band-pass filter (Omega Optical) was consideredsince its pass-band overlaps that of the primary NBP filter, with a relatively hightransmissivity, and it provides OD-6 rejection in 380-700 nm and OD-5 in 700-1200 nm range. As such, the filter combination offers a high transmission of ~70%in the overlapping pass-band, while effectively rejecting out-of-band signal inter-ference. In addition, a UV enhanced long-pass dichroic mirror, with 350 nm edgewavelength (Chroma-ZT325DCLPXT), was used for simultaneous OH* chemilu-minescence and natural luminosity/two-color pyrometry measurements. The mir-ror reflects the UV light towards the chemiluminescence imaging system, with a98% reflectively at 307 nm and adds another OD-3 rejection to the higher wave-lengths in the visible range, expected from high intensity incandescence signalsfrom particles at elevated temperatures.In all cases discussed above, the significant drop in the CMOS sensor and thephotocathode quantum efficiency to nearly zero beyond 1,100 and 900 nm, respec-tively, defines the upper range for the out-of-band rejection limit considerations.Figure 3.8 shows the transmission curves for all optical filters used in this work.In addition to the interference filters discussed above, a series of metallic neu-tral density (ND) filters, available in a range of optical densities, were also used inthis work. These filters have metallic coatings on crown-glass substrates and, as62200 300 400 500 600 700 800 900Wavelength [nm]020406080100Transmissivity [%]OH*1 OH*2 High Pass Dichroic 700nm 800 nmFigure 3.8: Transmission curves from the various optical filters used in thisworkthe name suggests, are neutral to spectral information and uniformly attenuate theincoming light, mostly through reflection, given their OD value (see equation 3.1).3.2.5 Optical ProbeIn addition to the development of the spatially resolved optical diagnostics de-scribed above, a custom cylinder head, designed concurrent to this work [128] toenable mounting a wide range of fuel injectors and in-cylinder diagnostics, wasused to investigate the effect of spatially integrated measurements of similar phe-nomena. The spatially-integrated in-cylinder diagnostic tool of interest to this workwas a custom optical probe design, also designed and commissioned concurrent tothis work [156]. Simultaneous measurements and comparisons with the imagingresults would allow for better characterization of the spatially integrated measure-ments and extension of the optical diagnostics capabilities to more realistic higherload operating conditions. Figure 3.9 shows the engine head design with the in-cylinder diagnostic tools installed.The optical probe design utilized in this work comprises a trifurcated fiber63Figure 3.9: The engine head design with multiple in-cylinder diagnostics in-stalled [128]. The ICOS sensor uses infrared absorption for local fuelconcentration measurements (see [156] for more details)optical bundle and sapphire rod lens to collect the combustion light from insidethe combustion chamber. An interesting feature of the probe design was its self-cleaning capability [155, 156]. Considering the constraining effects of optics con-tamination and signal degradation, the self-cleaning feature was a significant de-sign characteristic as it would allow for longer operation durations, with the col-lection lens remaining clean under steady-state full load conditions. Figure 3.10shows the optical probe designed in the group [156] alongside the original designit was based on [155]. Further details of the optical probe design used in this workand the lens self-cleaning processes are available in [156].In this work, the optical probe was used for two-color pyrometry measurementsto provide estimations of in-cylinder temperature and soot concentration using thespatially integrated light collected from its conical field-of-view. As such, similarnarrow band-pass filters to those used in the spatially resolved two-color pyrometrysetup were mounted in the optical path of two of the optical fibers. The filters werepositioned after the collimating lens and before recording the signal using siliconswitchable photodetectors (Thorlabs). A schematic diagram of the optical probetwo-color pyrometry measurements and a photo of the optical probe assembly arepresented in Figure 3.1164Figure 3.10: Optical probe designs. Left: by Yeo [156] (used in this work),Right: by Yan and Borman [155].Similar to the spatially resolved two-color pyrometry imaging system, the op-tical probe requires calibration for sensor response in order to convert the recordedvoltage into spectral radiant intensity and, ultimately, apparent temperature values.Following this step, the temperature and KL values are estimated using the two-color pyrometry equation 2.4. Therefore, the integrating sphere also serves as thecharacterized calibration light source for the optical probe, although setup with adifferent layout to address probe measurement principles. Figure 3.12 shows theoptical probe calibration setup for two-color pyrometry measurements. In the ab-sence of a linear exposure duration setting used in the imaging system, neutraldensity filters (see §3.2.2) of a range of optical densities (0.1-0.6) were mountedin the filter to attenuate the sphere intensity. A calibration curve is obtained fromthese intensity measurements, which can be used to reach high intensities of com-bustion through extrapolation.The self-cleaning performance of the optical probe was evaluated through im-plementation of the calibration curves obtained for clean and dirty sapphire rodlenses, after nearly three hours of thermodynamic engine operation, on a sampleoperating condition [156]. This investigation showed an overestimation of 0.3%for temperature and 8% for the KL factor.65Figure 3.11: Optical probe assembly: Sensor assembly (Left) and probe as-sembly installed in the engine head, adapted from [156] (right)Figure 3.12: Optical probe calibration system.3.3 Integrated System Operating ProtocolsUnder both engine configurations, the engine and camera systems are controlledusing the in-house developed data acquisition (DAQ) and engine control unit (ECU)systems, primarily using National Instrument’s LabView programming environ-66ment and FPGA. The intricate details of the designed engine control and DAQsystem, and sensors and actuators involved in the successful and safe operation ofengine, can be found in [156]. The work presented here only focused on integra-tion of the optical diagnostics to the developed engine control and DAQ systems.In what follows the major details of imaging systems integration to the engine,calibration, and setup are presented.3.3.1 Synchronization of optical diagnostics with the engineAn optical encoder was used to synchronize the engine and the imaging systems,through determination of camera trigger TTL signal timings and/or individual im-age frame capture synchronous to engine rotation, as well as synchronized sensordata logging onto the DAQ. A 1440 pulses/revolution optical encoder (BEI) wasthe central electro-mechanical piece for the imaging system integration task.The engine synchronous sensor data logging with this encoder provides quartercrank-angle degree (CAD) temporally resolved optical probe and HRR (pressure)results. In the case of spatially resolved optical measurements, there are two syn-chronization factors involved: 1- The start of an image recording sequence, con-trolled by a TTL signal, setup a priori at a crank-position offset from the enginetop-dead-center (TDC), and 2- individual image frame capture timing. The lattercan be executed at a constant frequency, determined as the recording frame ratein the camera software. In this case, the camera sets its internal crystal clock atthis constant frequency and starts the imaging sequence on the first clock tick afterthe detection of the trigger signal rising edge (under normal settings). However,given the concentrated power delivery to the shaft of the single cylinder enginesetup, the system does not operate under a constant frequency and instead periodi-cally undergoes acceleration and deceleration in each cycle. As such, alternativelythe individual encoder pulses might be used to control the timing of the electronicshutter on the camera sensor chip and capture the individual image frames so as toaccurately match the timing with the engine synchronous data. However, the cam-era system capability to operate under a variable frame capture frequency needs tobe supported by the designed camera firmware, which was only the case for thePhantom V7.1 camera used in this work.67In addition to the high-speed cameras, the high-speed image intensifier is alsorequired to be synchronized with its paired camera (i.e., the Photron SA-1). Theimage intensifier is operated by a control unit, which in addition to the MCP gainand photocathode gate settings, manages system triggering and frequency. Trig-gering the intensifier has a slightly different concept from the high-speed cameras.An arming/disarming port at the intensifier controller back panel accepts a Booleanto energize the photocathode and start the photon-electron conversion. This signalis provided by the Photron camera signal output bundle, during an image recordingsequence (RecPos signal). Another output signal from the camera to the inten-sifier is the camera internal clock pulses, which provides the synchronization ofindividual image frames between the two. The frame exposure duration of the as-sembled chemiluminescence measurement system is determined by the smaller oftwo values: Photron camera exposure or intensifier gate settings. To avoid unnec-essary burden on the photocathode, the exposure of system is controlled by theintensifier gate settings. As such, the Photron camera exposure must be set to alonger duration, yet not too long to unnecessarily record the background noise (afew microseconds longer than the intensifier gate).A simplified wiring diagram of the integrated optical diagnostics are presentedin Figure 3.13. In this configuration, the images recorded by the Phantom cam-era (NL and pyrometry images) are synchronous with the HRR and optical proberesults, while the images recorded by the Photron camera (OH* chemilumines-cence images), captured with a constant frequency, are not necessarily synchronouswith other measurements. As such, the time stamps from the engine encoder andPhotron camera internal clock, recorded on the DAQ, were compared and inter-polated to attribute OH* chemiluminescence image frames to the closest encodertime stamps. It is worth noting that this process occasionally leads to repeated OH*chemiluminescence image frames for two consecutive registered encoder timings,given the higher than nominal frequency of engine rotation after TDC of combus-tion cycle, and appears as “frozen frames” in the phasing-adjusted sequence.68Figure 3.13: Simplified wiring diagram of the integrated optical systems.3.3.2 Alignment and CalibrationSeveral preparation steps are required before each new set of spatially resolvedoptical measurements of combustion in an IC engine. This procedure makes suchmeasurements relatively time-consuming and labour-intensive to setup. For a freshset of optical measurements, engine window, optical probe lens, and if necessary,the engine head are removed and cleaned, and certain mechanical parts are re-placed, if needed. After engine reassembly, the imaging systems should be po-sitioned at the suitable stand-off distance (see Appendix A) and be aligned andfocused, and in the case of two-color pyrometry, calibrated (see §3.2.2). It isworth noting that in the case of chemiluminescence measurement setup, choosinga higher intensifier gain setting and the lowest frame-rate setting on the controllingcamera (see §3.3.1) would reduce the photocathode use overhead during systemalignment and focusing process.Continuous optical in-cylinder soot measurements is challenging to maintaindue to mechanical and instrument-related limitations and the preparation proto-cols described above should be repeated often. Window contamination is typicallythe primary source of disruption in continuous measurements. Optical rings andwindow gaskets also need to be replaced due to wear occasionally during opti-cal engine maintenance, albeit less frequently under normal operation. In order69Figure 3.14: Adjustable optical table system to facilitate higher throughputand repeatable optical measurements.to improve optical measurements system throughput and repeatability, a portableoptical table was used with three DOF1 optical breadboard control. The bread-board can be fine-tuned for elevation, stand-off distance and lateral positioning,and the table structure, sitting on casters, can be rolled close to or away from theengine. Once the imaging systems and all the relevant optical elements were prop-erly mounted and locked in position on the breadboard, the entire imaging systemcan be conveniently removed and then put back in position for a new set of opticalmeasurements. A pair of mechanical table guides and a laser pointer aiming at atarget affixed to the optical engine block allow for accurate re-positioning of theoptical table in front of the optical engine mirror. The mechanical table guides areremovable to prevent table movements during engine operation. Figure 3.14 showsthe optical table structure and mounted imaging system setup.1Degree of freedom70Figure 3.15: Standard optical engine operation skip-firing scheme. 15 repeti-tions of the skip-firing sequence sets are performed during each opticalengine run.3.3.3 Optical engine operationAs discussed in §2.1.1, continuous firing should be avoided in order to allow forthe cooling of the quartz engine window between cycles. As such, to facilitatethe in-cylinder imaging of a more representative combustion event, a skip-firingprocedure was used in which three fired cycles followed by a certain number ofmotored cycles (15-17). Only the final of the three fired cycles was used for imag-ing purposes, as the first and second cycles were noted to be transitory while thethird cycle was observed to reach a steady state [124]. The described skip-firing setwas repeated for 15 times, restricted by the partitions limit of the Phantom camera.The ensemble average of these 15 cycles were considered to limit the cycle-to-cycle variability effects. A schematic illustration of the skip-firing procedure ispresented in Figure 3.15.Also as discussed in §2.1.1, the geometric modifications to afford the spatiallyresolved optical access into the combustion chamber usually results in a lower com-pression ratio and necessities measures to compensate for the discrepancy. In thecurrent optical engine setup, this was addressed through heating of the intake air (to50-80◦C, depending on the operating conditions) and boosting intake air pressure(~1 bar-gauge) to compensate for air density change. These adjustments would cre-ate similar conditions at TDC and result in close combustion phasing, characterizedby the 50% integrated heat release (CA50), to that of an equivalent thermodynamicoperating condition. Further fine-tuning of the combustion phasing is then executedthrough adjustments in fuel injection timings [124]. A summary flowchart of the71Figure 3.16: Optical measurements protocol flowchart.optical measurement protocol is provided in Figure 3.16. As illustrated here, theoptical engine preparation stage constitutes the majority of the operation time fora set of optical engine measurements. Depending on engine and optics conditions,engine preparation can take anything between one hour, with only engine windowcleaning required, to 4 hours, due to engine disassembly for required cleaning andregular maintenance of optical and engine parts and re-calibrating and re-aligningthe imaging systems.72Chapter 4Optical Investigation of DIDFcombustionFollowing the commissioning of the simultaneous spatially and temporally re-solved OH* chemiluminescence and natural luminosity measurement system, astudy on the effects of fueling control parameters on combustion characteristics ofDIDF combustion of methane was carried out, as a case study [75]. The currentchapter presents the findings of this investigation.4.1 Analysis Metrics and Engine Operating ConditionsThe metrics used to identify operating points in the current work are pilot fuelinjection pressure, the global equivalence ratio (φglobal) and pilot ratio (PR). Thelatter two are described through equations 4.1 and 4.2 [127]:φglobal =m˙diesel · AFdiesel,stoich + m˙CH4 ·AFCH4,stoichm˙air(4.1)PR =m˙diesel ·LHVdieselm˙diesel ·LHVdiesel + m˙CH4 ·LHVCH4(4.2)where m˙ is the mass flow rate and LHV is the lower heating value of each fuel.The thermodynamic results in [127] indicated that increasing pilot injection pres-73sure increased the fraction of fuel converted in the first stage of DIDF combustion.Up to a certain limit, increasing PR increases the proportion of energy released inthe first stage. Conversely, increasing methane equivalence ratio (φCH4) increasedthe energy fraction released in the second stage. The observed effects of the in-jection pressure and PR on dual fuel combustion can be further supported usingoptical measurements. In addition, lowering pilot injection pressure was observedto result in higher methane emissions. This observation, combined with the toroidcombustion behavior reported by Dronniou et al. [33], which suggests the possibleorigin of the increased unburned methane emissions might be the central regionpremixed charge, can be another motivation to see how manipulation of pilot in-jection pressure will affect the combustion behavior. This investigation presentedin this chapter examines the effects on dual fuel combustion incurred by changingthe pilot fuel injection pressure and the pilot ratio. These effects are studied opti-cally through use of simultaneous spatially and temporally resolved OH* and NLimaging.A combination of testing in thermodynamic and optical configurations was em-ployed for this investigation. The first round of testing was performed with the en-gine in thermodynamic configuration and detailed thermodynamic analysis resultswere presented by Rochussen et al. [127]. The primary goals of the thermody-namic testing, relevant to the discussion presented in this chapter, was to define theoperating points of interest (e.g., pilot injection parameters), and to characterizemetrics which are not readily measured during short optical measurements (e.g.,fuel and air flow rates, emissions, etc.).Thermodynamic measurements presented in [127] identified distinct regimesof dual-fuel combustion. A representative operating point from each combustionregime was selected for further investigation through optical measurements in thisstudy. After defining the operating points to be optically studied, the engine is con-verted into the optical configuration to more closely investigate the effects observedin the thermodynamic measurements. The selected operating points are presentedin Table 4.1 and will be discussed in detail below.The operating points outlined in Table 4.1 were selected because they demon-strated significantly different combustion characteristics such as the energy releasefraction in different stages of combustion, shape of HRR curves, occurrence of74Table 4.1: Operating points specifications. Mass of diesel was set based onthe injection duration determined in thermodynamic configuration. Einputis the total fuel energy input to the system and Iηth is the indicated ther-mal efficiency.Point φglobal[-]φCH4[-]PR[-]mdiesel[mg/cyc]Einput[kJ/cyc]GIMEP[bar]Iηth[%]CSOI[bTDC]A-1300 0.8 0.75 0.1 7.2 4.5 7.4 32 4A-300 0.8 0.75 0.1 7.2 4.5 6.2 27 8B-1300 0.7 0.60 0.14 15.6 4.1 6.4 31 6B-300 0.7 0.60 0.14 15.6 4.1 5.1 25 10C-1300 0.7 0.45 0.35 31.7 4.0 6.0 30 4flame propagation, and dependence of emissions and HRR curve on pilot fuelingparameters [127]. In the naming convention used here for each operating point,the letter corresponds to the fueling parameters (φCH4 ,φglobal, and PR), while thenumber indicates the injection pressure in bars.As discussed in §3.3.3, due to the mechanical limitations of optical engineoperation, a skip-fire operating protocol was used. To maintain a consistent heatrelease rate, this protocol consisted of three consecutive fired cycles, with imagesbeing recorded on the final cycle. This was then followed by 12 motored cycles.For each test point, a total of 15 fired cycles were recorded, as limited by thecamera buffer capacity. Also, as discussed in §2.1.1 and listed in Table 3.1, theoptical and thermodynamic configurations have different compression ratios and,therefore, the intake air temperature was adjusted using electric heaters prior torunning each optical test in order to compensate for this discrepancy and approxi-mately match the combustion phasing to previous thermodynamic measurements.The combustion phasing was characterized by the crank position at which 50% ofenergy is released (CA50). For this study, the engine was naturally aspirated andintake air was heated up to 80± 5◦C. The simultaneous OH* and NL imaging sys-tem and general engine specifications were presented previously in Table 3.3 and3.1; Table 4.2 here specifies the imaging and optical engine settings used for the75current investigation. A schematic of the simultaneous measurement system usedis shown in Figure 4.1.Table 4.2: Simultaneous OH* and NL imaging system setting for optical in-vestigation of DIDF combustionOptical EngineDirect Injector Bosch CRIN2Port Injector Bosch NG12Pilot fuel pump dieselPremixed fuel pure methaneIntake air condition naturally aspirated; heated to 80±5◦CSpeed 600 rpmNL Imaging SystemFrame Rate Variable (Externally triggered)Resolution 320×240Exposure 8 µsecAperture f/2.8OH* Chemiluminescence Imaging SystemFrame Rate 12,000 fps (Internal Clock)Resolution 640×640Exposure 84 µsecAperture f/5.6IRO Gating 10 µsecIRO Gain 70%Interference Filter 20 nm FWHM band-pass; 307 nm CWLHere, the Phantom v7.1 camera was synchronized with the thermodynamicdata using the crank-shaft encoder signal (0.25 CAD resolution), while the OH*chemiluminescence images were captured at a constant frequency 12,000 frames/sec-ond. This frame rate corresponded to ~0.3 CAD resolution at the 600 rpm enginespeed used for this study.The recorded images were processed for image enhancement, as well as tospatially and temporally align the images from the two imaging systems. The 12-76Figure 4.1: Schematic of the simultaneous OH* and NL imaging system usedfor optical investigation of DIDF combustion.bit raw images captured by the two cameras had the following post-processingsteps applied:1. Temporal synchronization of the image series from the two cameras.2. Image registration: adjustment of image size, scale, and positioning to beconsistent between the two series.3. Background and noise removal, intensity adjustments and image enhance-ments (gain, brightness, and gamma corrections).4. Ensemble averaging, morphological operations and edge detection, imagethresholding, preparation of compatible color maps, alpha blending, and im-age overlay.774.2 Results and DiscussionFor the A and B operating points two Pin j are compared to illustrate the effect ofpilot fuel injection pressure on dual-fuel combustion with similar fuelling param-eters. Points A, B, and C, on the other hand, can be studied at constant Pin j (1300bar) to consider the effect of PR and φCH4 .In addition to the injection pressure and relative fuel quantities, the influence ofthe premixed methane on combustion of diesel was studied by comparing the dual-fuel operation with representative “diesel-only” points. For these points, the dieselinjection parameters are the same, but no CH4 is inducted into the combustionchamber. Such comparisons were carried out for B-1300, B-300 and C-1300.4.2.1 Effect of the Pilot Fuel Injection PressureThis section discusses the effects of different pilot fuel injection pressures on dual-fuel combustion behavior. In order to elucidate the approach taken to discuss opti-cal data in this study, two sample operating points are presented in Figure 4.2. Thisfigure presents two series of images for each of the two sample operating points,where OH* and NL processed signals can be compared at each crank position.This illustrates evolution of combustion at each operating point and affords com-parison of this process among different operating points. However, it is more usefulto present and discuss the information provided through both OH* and NL at eachcrank position on a single image. This will provide a more compact presentation ofdata and also enables comparison of spatial distribution of high temperature reac-tion zones, characterized by OH*, and high intensity NL regions, usually attributedto incandescent soot particles. The combined presentation of OH* and NL imagesis illustrated in Figure 4.3 for B-1300 operating point at 14.5 CAD after command-of-start-of-injection (aCSOI). In this context, Figure 4.2 is a reference to facilitateinterpretation of overlaid blended images in Figure 4.3. Here, the NL data is falsecolored with shades of yellow to dark red, as in figure 2. Thresholding is appliedto the OH* chemiluminescence data to reduce clutter in the overlaid images. Thethree regions, delimited by contours in dark blue (solid), cyan (dashed), and green(dotted) color, show the boundary of upper 75, 50, and 25% chemiluminescence,respectively. False coloring (shades of green) is used to indicate the highest OH*78Figure 4.2: Sample image series of ensemble averaged NL and OH* chemi-luminescence (left); and jet orientations from a split injection diesel testwith external illumination (right).intensities (>75%). A different OH* color map from Figure 4.2 was used in Figure4.3 in order to enhance the visualization of OH* signal in the overlapping regions.Figure 4.2 also shows the five diesel jets produced by the injector nuzzle andtheir orientation. This sample back-scattering image was obtained from a diesel-only combustion test (no NG), where the combustion chamber was illuminated viaan external light source for spray visualization. Due to the CCW swirl motion ofthe charge (swirl number≈ 3.2) and ignition delays, the jet structures in A-300 andB-300 points are rotated relative to the original sprays.Following the procedure described in Figure 4.3, the combined OH* and NLimages for all operating points in Table 4.1 are shown in Figure 4.4. 15 fired cycleswere imaged for each operating point and Figures 4.2-4.4 present the ensembleaveraged results. Each image series in Figure 4.4 starts at approximately 5% grossintegrated heat release (IHR) and for the sake of brevity, only images at 2 CADintervals are presented. The crank positions reported on each frame in these figuresare relative to the CSOI.For the pilot injection pressure comparison (points A and B), the pilot injec-tion duration was adjusted for equal pilot injection masses. The required durationsat each pilot pressure were determined during thermodynamic engine tests [127].The penetration of the pilot fuel and distribution of ignition sites were significantly79Figure 4.3: Overlaid presentation of ensemble averaged OH* chemilumines-cence and natural luminosity images for B-1300 at 14.5 CAD aCSOI.The gross IHR value up to this instance is presented at the top-right ofthe image frame.affected by the injection pressure. Comparisons of ignition site locations and reac-tion zone growth mechanisms between high injection pressure points (A-1300 andB-1300) with their corresponding low pressure cases (A-300 and B-300), demon-strate how the dual-fuel combustion mode can be manipulated by controlling theinjection pressure of the diesel pilot fuel.For the higher injection pressure, the liquid diesel atomizes better, entrains andmixes faster with charge gas, and penetrates further into the combustion chamber.This results in auto-ignition of a partially-premixed diesel mixture at the bowl pe-riphery (the first three columns in Figure 4.4). The reaction zone then propagatestowards the center, via what is commonly accepted to be turbulent flame propaga-tion, as was also observed by Dronniou et al. [33]. The dual-fuel combustion inthese cases produces fewer high intensity localized NL regions when compared tothe lower injection pressure, which, for instance, depicts multiple black pockets at13 CAD aCSOI in A-300 case. Since these high intensity regions are associatedwith radiation from soot in locally rich regions, their absence can be indicative ofcombustion initiation in lean, well-mixed regions.80Figure 4.4: Overlaid ensemble averaged OH* chemiluminescence and natu-ral luminosity images for dual fuel operating points described in Table4.1. Crank positions are relative to CSOI and gross IHR values are pre-sented at the top of each frame.It should be noted, however, that high-intensity NL regions that show highspatial gradients (i.e., are more localized) and emerge at earlier crank positions andcloser to CSOI are more likely to indicate production of incandescent soot. Thisshould not be confused with high-intensity NL regions appearing later in the cyclewhich may be from soot or natural luminosity from gas phase products such asCO2, HCO, CH2O, and C2* [93].For the lower injection pressure (points A-300 and B-300), the diesel jet does81not penetrate as far or mix as well. As a result, the ignition sites are localizedaround the vicinity of the pilot fuel jets and a more heterogeneous auto-ignitionof the charge was observed. The auto-ignition in these cases occurs after the endof injection (EOI) and the “jet” refers to the region in which the injected fuel hasincreased the local equivalence ratio, rather than the actual liquid pilot fuel jet.In these auto-ignition regions, intense luminosity was observed, attributed hereto the presence of soot particles and indicative of a locally rich mixture (the lasttwo columns in Figure 4.4). The jet structures can also be recognized in OH*images from pilot fuel ignition towards later in the combustion cycle. Ignition ofthe premixed natural gas occurs at the proximity of the ignited pockets around thejet structures, observed across the entire chamber.Thus, the pilot fuel injection pressure and consequent effects on fuel mixingrates greatly affect the distribution of equivalence ratios and formation of fuel-richregions, potentially producing more soot and stronger NL intensities in some re-gions. This confirms the results of Carlucci et al. [18], where higher total timeintegrated and peak luminosity was observed for lower pilot fuel injection pres-sures.The conceptual representation of the described behaviors as the result of chang-ing Pin j is presented in Figure 4.5, which is based on OH* images from represen-tative high and low pilot injection pressure operating points.Figure 4.5: Conceptual effect of pilot fuel injection pressure on reaction zonegrowth mechanism (left: high pressure; right: low pressure) on DF com-bustion modeThe conceptual description of the effect of pilot injection pressure may alsoprovide an explanation for lower unburned methane emission with high Pin j, as82observed in thermodynamic investigations [127]. Concentration of auto-ignitionand high temperature reaction zones around the edges enhances the fuel conver-sion in these critical, wall-influenced regions. The propagation of reaction zonetowards the center consumes the rest of the charge in the higher temperature, cen-tral regions. For low injection pressures, the scattered auto-ignition and localizedreaction zones, not necessarily close to the walls, leaves more premixed pocketsbehind.Another observation made from Figure 4.4 was that the high temperature re-action zones (characterized by OH* chemiluminescence) are coincident with midto high level broad-band light emitting regions (high intensity NL), except for ear-lier stages in low Pin j points when intense luminosity is observed and suggestiveof early soot production in poorly mixed fuel-rich regions. This was consistentamong all operating points studied in this work. According to Karim’s thermody-namic analysis results [69], the pilot diesel fuel is assumed to be almost completelyconsumed in the first stage. This is in contrast with optical results acquired in thiswork, where, for the lower pilot injection pressure cases, the pilot jet structures arepresent in the later combustion stages. These jet structures are suggestive of higherlocal equivalence ratios and higher reactivity in those regions probably due to theconversion of the remaining diesel fuel.4.2.2 Effect of the PR and φCH4The other important fueling parameters investigated in this work were diesel pilotratio and methane equivalence ratio. Points A-1300, B-1300, and C-1300 have thesame pilot injection pressure, as do the points A-300 and B-300. They also havea similar φglobal and, therefore, can be used to investigate the combined effect ofdifferent PR and φCH4 values. Point A has the smallest PR value, while C has thelargest PR value. Because they have similar global equivalence ratios, this alsorequires that C have a lower φCH4 . From the OH* images (Figure 4.4), C-1300demonstrates high temperature reaction zone propagation from the periphery tothe center, similar to the other high injection pressure points (A-1300 and B-1300).However, the fuel conversion rate is more rapid for this higher PR point (C-1300).From natural luminosity images, C-1300 shows a more heterogeneous combus-83tion event than the other high injection pressure points, and high intensity incan-descent soot regions as a result of increasing PR. Observation of soot regions at thisincreased pilot ratio suggests that the otherwise effective charge mixing (atomiza-tion at 1300 bar injection, available turbulence and dwell time) are not sufficientlyeffective anymore to handle the increased diesel in the charge. This can also beinferred from Figure 4.6-a, where spatially averaged NL histories are comparedfor all operating points (ΣIx,y/N, where I and N are pixel intensity and number ofpixels, respectively). Point C-1300 looks different from the bell shape curves ofA-1300 and B-1300, and shows a sharp peak of natural luminosity shortly afterignition. The optical data in Figure 4.6 were not subjected to any non-linear in-tensity adjustments or histogram shifts; however, they are all normalized with themaximum signal intensity in point A-1300 to facilitate direct comparison.Although points A-1300, B-1300, and C-1300 show similar high-temperaturereaction zone propagation mechanism in the OH* chemiluminescence images, theextent and rate at which this growth takes place correlates with φCH4 (and there-fore PR). The results presented in Figure 4.4 (first 3 rows for these 3 points) andFigure 4.6-b, show that a higher initial rate of reaction zone growth can be ob-served for higher PR. Comparison of the slope of rising spatially averaged OH*signals amongst A-1300, B-1300, and C-1300 points in Figure 4.6-b expresses thisincreased growth rate for a higher PR.Figure 4.7 shows the 20% thresholded OH* chemiluminescence for each op-erating point at its maximum bowl coverage (i.e., the pixels registering intensitiesless than 20% of the dynamic range are set to zero). This thresholding techniquefacilitates comparison of the extent of spatial progression of the high temperaturereaction zones for different operating points. From Figure 4.7, it is evident that re-ducing φCH4 restricts the extent of reaction zone propagation in high Pin j operatingpoints (A-1300, B-1300, and C-1300). The reduced extent of reaction zones can belinked to increased unburned CH4 emissions observed for decreasing φCH4 in ther-modynamic engine configuration [127], suggesting a less complete fuel conversionprocess.A similar conclusion can be inferred from points A-300 and B-300 (Figure4.7). Less heterogeneity is observed in the OH* signal from point A-300 (havinglarger φCH4), which also appears to cover the whole chamber later in the combus-84Figure 4.6: Comparison of spatially averaged NL (a) and OH* chemilumi-nescence (b) determined by ΣIx,y/N, where I and N are pixel intensityand number of pixels, respectively, and HRR (c).85Figure 4.7: OH* Chemiluminescence at the maximum bowl coverage.tion cycle. Similar to what was discussed above, this slower rate of reaction zonepropagation can be attributed to smaller PR in A-300. Point B-300 appears to bemore dominated by the diesel fuel in the charge and follows the pilot fuel jet struc-tures till later in the cycle (Figure 4.4 and Figure 4.7).Variation of pilot injection pressure and pilot ratio has a marked impact on theignition delay. From a thermodynamic analysis perspective this is conventionallycharacterized by CA5. From the optical point of view of this work, the ignitiondelay can be characterized by the detection of the earliest OH* chemiluminescenceafter CSOI, which is indicative of the onset of high temperature reactions withrespect to CSOI. It should be noted that these two ignition delays are not the same.Figure 4.6 also describes these effects, where lower φCH4 (higher PR) results in anearlier rise of the OH* and NL signal among the high Pin j points. The ignition delaydefined by the crank position when the spatially averaged OH* and NL signalsexceed 5% of their maximum value are listed in Table 4.3. It is worth noting thata 2 CAD advanced CSOI for B-1300, relative to A-1300 and C-1300, does notaffect the ignition delay behavior among these three points. These observations areconsistent with results presented by Schlatter et al. [133], where higher values ofφCH4 promoted propagation of the premixed flame, as was seen in this work fromthe comparison of points A-1300 and B-1300, but delayed the auto-ignition of thepilot spray considerably.86Table 4.3: Ignition delay and combustion duration.A-1300 B-1300 C-1300 A-300 B-300Ignition delay (OH*)[CAD aSOI]7.75 5.5 4.25 7.75 8.25Ignition delay (NL)[CAD aSOI]8.75 5.0 3.5 6.0 6.75Ignition delay (CA5)[CAD aSOI]11 8.5 7.2 11.1 11.6Combustion phasing(CA50) [CAD aSOI]17 12.5 9 17 17.5Combustion duration(CA90-CA5) [CAD]21.6 17.2 15 37.5 41.2CSOI [CAD bTDC] 4 6 4 8 10As the injection pressure is decreased, a longer delay is observed, except forpoint A-1300. This is most likely a result of a more advanced CSOI for A-300and B-300, coupled with a very low diesel mass quantity in the dual-fuel charge ofA-1300. The enhanced mixing at this high Pin j resulted in a diluted pilot mixture,whereas at the low Pin j, the slower mixing and stratification in the charge made formore reactive regions for auto-ignition across the chamber.Figure 4.4 also presents gross IHR fractions corresponding to each image frame(top right at each frame). The heat release rates information also indicates in-creased fuel conversion rates and more aggressive combustion events for increasedpilot ratio. The IHR values in figure 3 show that in the first 8 CAD after start-of-combustion (aSOC), 84%, 69%, and 66% of the total apparent energy releaseoccurs for C-1300, B-1300, and A-1300, respectively. 47% of the total apparentenergy is released over only 1 CAD for the C-1300 operating point, which makesit the most aggressive point presented in this work and was observed as substantialoscillation of the flame structures in the single cycle image sequences (not dis-cernible in ensemble averaged results). This can also be inferred from the heatrelease rates presented in Figure 4.6-c, where larger slopes are observed for higherPR points.For lower pilot injection pressures, the energy is released at a slower rate of87around 57-59% of the total apparent energy release over the first 8 CAD aSOC.Thermodynamic analysis results do not reveal other effects discussed based onthe optical data. This investigation suggests that the shape of the heat release ratecurves are most strongly influenced by the pilot ratio. A transition from bell-shapedheat release rate, to one weighted more heavily towards earlier crank positions,regardless of the pilot fuel injection pressure, is observed with increasing PR. TheB and C HRR curves show the same two-stage combustion process discussed in[69].The deviation from a bell shape in the spatially averaged NL curve pointed outfor C-1300, was also observed for A-300 and B-300 (10-15 CAD) in Figure 4.6-a. The shape of these curves appears to have some correlation with HRR curvesand will be the focus of a future work. This can be important since, accordingto Karim’s conceptual model [69], deformations in HRR curves and presence ofmultiple peaks are indicative of different stages of combustion. The first defor-mation/peak can be associated with the partially-premixed pilot fuel and entrainedpremixed NG heat release, followed by the conversion of natural gas plus the restof diesel from the pilot. From this perspective, the sharp peak in spatially averagedNL and HRR curves of C-1300 is a result of a more profound first stage due to itshigher PR.4.2.3 Influence of Diesel and Natural Gas on Each Other inDual-Fuel Combustion ChargeThe optical results presented in the previous sections demonstrated that, as ex-pected, there is a strong interaction between the premixed methane and the dieselpilot. To further consider this, “diesel-only” cases were also considered for oper-ating points B-1300, B-300, and C-1300. These points have the same total dieselmass and mass flow rate as their corresponding dual-fuel point. Thus, each ofthe corresponding diesel-only and dual-fuel point pairs can be studied to comparethe effect of presence air vs. premixed NG on the combustion process. Figure4.8 shows an image series comparing OH* chemiluminescence from dual-fuel anddiesel-only combustion of these three points at 2◦ CAD increments. The crank po-sitions are relative to CSOI. The imaging system settings were identical for all ofpoints; however, the image gain was doubled for the diesel points to better visualize88Figure 4.8: Reaction zones (OH* images) in dual-fuel and diesel-only com-bustion. Crank positions are relative to CSOI and the image gain wasdoubled for diesel points for better visualization.the phenomena.All the diesel-only cases agree with the conclusions made for their equivalentdual-fuel case. That is, point B-300 shows the pilot fuel jet structures, whereaspoints B-1300 and C-1300, once again, show the start of the process at the periph-ery of the bowl and propagation directed inwards. Also, the larger total mass ofdiesel fuel injected in operating point C-1300 manifests as larger high temperaturereaction zone areas relative to B-1300 and more rapid growth of these regions.The comparison of these points also demonstrates the increased ignition delaydue to the premixed natural gas. In all instances presented here, the diesel-only caseshows the start of reaction zone growth at an earlier time than the dual fuel case.These observations are consistent with results presented by Schlatter et al. [133],where injection of diesel pilot in pure air advanced the auto-ignition considerably.4.3 Summary and ConclusionsThe work presented in this chapter, for the first time, has utilized simultaneous spa-tially and temporally resolved OH* chemiluminescence and NL imaging to inves-tigate effects of fueling control parameters on direct-injected diesel and premixed89natural gas dual-fuel combustion. The effects of pilot fuel injection pressure andratio on dual-fuel combustion were investigated. The results demonstrate the po-tential to manipulate combustion of premixed natural gas (ignition sites locationsand reaction zone growth mechanisms) through the fuelling strategy (Pin j, PR andφCH4).At high pilot injection pressures, the high temperature reaction zones and lu-minous ignition sites emerge around the edges of the piston bowl and propagatetowards the center. This observation is in agreement with previous results [33]despite the fact that a light-duty optical engine with a much smaller displacement(499 cm3) and bore size (85 mm) was used in that work. It was also noted that theextent and rate at which the growth occurs correlates with the equivalence ratio ofthe premixed CH4 charge (which is inversely proportional to the diesel pilot ratiofor a given φglobal, i.e., load). Increasing the pilot ratio, while keeping the globalequivalence ratio constant, resulted in a more rapid fuel conversion but limited thereaction zone propagation towards the center. This has the potential to be relatedto the emission of unburned species, which will be discussed in a future investiga-tion. The possibility of increasing flame propagation to reach the central region byincreasing φCH4 was also observed in [33].Lower pilot injection pressures resulted in a more heterogeneous combustionevent and locally fuel rich zones producing highly luminous soot particles. In gen-eral, the reaction zone initiated at the diesel pilot sprays and propagated outwardsfrom there. This is a significantly different mechanism than noted for the higherinjection pressures. Similar flame propagation mechanism was noted by Schlatteret al. in an RCEM with an 84 mm bore diameter with 400 bar injection pressure[133].The comparison of investigated dual-fuel operating points to their correspond-ing diesel-only cases indicated similar trends in the reaction zone growth mecha-nism. This indicates that the premixed methane conversion mode is strongly im-pacted by the diesel pilot injection strategy.The combustion modes observed at the different injection pressures are con-sistent with those proposed in previous thermodynamic studies [69, 127] and cansignificantly impact emissions and indicated efficiency. In addition, the diesel-onlyinvestigation indicated the hindering effect of presence of natural gas in the cham-90ber on the ignition of the charge, which has been noted elsewhere [133]. Furtherinvestigations are still required to assess the sensitivity of these observations toother engine operating speeds, loads and operation with EGR.91Chapter 5Two-Color Pyrometry MethodImprovementIn Chapter 2, the concept, fundamentals, and a review of the literature on imple-mentation of two-color pyrometry technique, to obtain spatially resolved distribu-tion of soot cloud temperature and concentration, was presented. Later in that chap-ter (see §2.3), it was discussed that the implementation of this method becomesmore challenging at lower soot concentrations and temperatures, as the two-colorpyrometry is an unstimulated emission technique relying on the naturally emit-ted incandescence from particles at elevated temperatures and the signal-to-noise(SNR) ratio significantly drops at those low concentration and temperature scenar-ios. Thus, this chapter has focused on identifying and assessing improvements totwo-color pyrometry in order to facilitate its application to conditions with lowerlight intensities, such as LTC, PIDING, or late cycle soot characterization. Theseimprovements were then applied to conventional diesel combustion as well as thelow-soot conditions for PIDING combustion.5.1 Method EnhancementSeveral stages were considered to improve the overall performance of the two-colorpyrometry method. In particular, improvements are identified through refinementsto the solution algorithm, pyrometry system configuration, calibration, and image92processing. A detailed discussion of the individual steps are presented in whatfollows.5.1.1 Algorithm SelectionThe solution of equation 2.4 provides the KL factor and temperature for each pixelof each image recorded during a measurement; however, the efficacy and effi-ciency with which this is done varies significantly between various algorithms.Notall pairs of apparent temperatures will result in a solution for equation 2.4, withfeasible temperature and KL values and thus the method is susceptible to signalnoise. As such, the applied solution methodology can have a significant impacton the total calculation time and robustness. The performance of several solutionmethodologies was compared on the basis of their accuracy and computation re-quirements. To elucidate these influences, the behavior of the pyrometry equationsare examined over the parameter ranges relevant to combustion studies, and severalsolution methodologies are compared.The range of KL values which may be evaluated using the pyrometry methoddepends on the considered wavelengths (700 and 800 nm in this work, see §2.1.4).As shown in Figure 5.1, for KL > 3 the emissivity is approximately unity and thesoot cloud acts like a blackbody and changes in KL are too difficult to assess as itbecomes very sensitive to any noise in the data. Thus, KL factors in the range of[0.01 3] were considered to assess algorithm performance.The solution of equation 2.4 can be achieved using a range of numerical rootfinding algorithms. To demonstrate the challenges associated with the solution ofthis equation, Figure 5.2 shows the residual ( f , see equation 2.4) for a representa-tive apparent temperature pair at 700 nm and 800 nm (T = 2708 K and KL = 0.796).Near the root (i.e. f = 0, T = 2708 K) the residual changes rapidly, which typicallyfacilitates a rapid and robust solution. However, if a poor initial estimate for T isused (e.g., >3000 K), the root may not be found. Furthermore, as the apparenttemperatures change, due to changes in T and KL, the shape of the residual willalso change.The characteristics of three classes of root-detecting algorithms were evalu-ated: Newton-Raphson, Brent-Decker, and non-linear least square optimization al-930 1 2 3 4 500. FactorEmissivityFigure 5.1: Soot cloud spectral emissivity vs. KL factor based on Hottel-Broughton’s empirical relationship (Equation 2.1) for typical engine rel-evant KL factor range.1000 2000 3000 4000 5000 6000Presumed Tempera ture  Solution [K]-0.02-0.0100. Error [-]Ta-700= 2600 KTa-800= 2550 KT = 2708 KKL = 0.8Figure 5.2: Residual error behavior around the root of equation 2.4. Tem-perature and KL were calculated using the non-linear least squaretrust-region dogleg algorithm (explained below). Solution: KL=0.8,T =2708 K.94gorithms. The Newton-Raphson method is relatively efficient but often encountersdifficulties, such as overshoot due to inflection points or failure at stationary points.It is also susceptible to poor initial estimates of the root. It is therefore common toplace limits on the number of iterations, bound the solution to an interval knownto contain the root, and combine the method with a more robust root-detectionalgorithm.The Brent-Dekker algorithm [17, 26] is a hybrid method which combines threedifferent methods: 1) the Secant method, which is a finite difference approximationof the Newton-Raphson method with a slower but still super-linear convergenceif the function meets certain criteria [122]; 2) the well-known bisection method,which is a relatively slow, yet robust method; and 3) inverse quadratic interpola-tion, which uses a recurrence relation based on Lagrange interpolation formula tofind a quadratic approximation of the inverse of the function [36]. A secant methodor inverse quadratic interpolation is used for fast convergence or, alternatively, bi-section method will be used if a more robust approach is necessary. The detailedprocedure on how the technique determines the suitable method for each iterationis described in [17].As an alternative to conventional root-finding algorithms, equation 2.4 can betreated as an optimization problem where the objective is to minimize f with anon-linear constraint applied to T (xi, j, t). For this, the trust-region dogleg methodwas considered, in which f is approximated using a simpler function ( f ) that hassimilar behavior near the point T (xi, j) (trust-region). The next iteration is evaluatedby minimizing f in the trust-region using Powell dogleg procedure [111, 121].The performance of each of the selected algorithms was evaluated over a rangeof temperatures and KL factors representative of engine combustion (T =1500-3000K; KL=0.01-3). The apparent temperatures for this T -KL space were used as in-puts to equation 2.4 to assess the ability of each algorithm to evaluate the cor-responding T and KL, as well as the computational time. The Newton-Raphsonmethod is susceptible to poor initial estimates and an initial guess of the form ofTinitial = Ta,700nm+g(Ta,700nm/Ta,800nm) was used and improved its performance no-tably (“assisted Newton-Raphson”).Table 5.1 summarizes the convergence time of the different methods for a rep-resentative condition (T = 2300 K and KL = 0.4). The reported calculation time for95Table 5.1: Performance of considered algorithms (Core i5 3.2 GHz processor,16 GB RAM). NLS: non-linear least squares.AlgorithmConvergence time [ms](KL = 0.4, T = 2300K)Fraction of T -KLspace with < 1%Newton-Raphson 118.6 (11 iterations) 80%; Increased error forAssisted Newton-Raphson63.8 (6 iterations) 0.5<KL<1 and KL>2,T <2200Brent-Dekker 2.960%; No convergence forTa,700nm/Ta,800nm > 0.994NLS optimization 5.5 99%Hybrid Algorithm 3.9 99%each method is for a single pixel, but it should be noted that a two-color pyrometryimaging analysis of a complete combustion process will require significantly moreevaluations (on the order of 107). The Newton-Raphson method is slower from theremaining algorithms by an order of magnitude.Also indicated in Table 5.1 are regions of the T -KL space where the algorithmsfailed (Brent-Dekker) or had reduced performance, as well as the fraction of theT -KL space that was resolved with < 1% error by each method. This is illus-trated more clearly in Figure 5.3, where the performance over the complete rangeof temperatures and KL factors is considered. Here, it is noted that the assistedNewton-Raphson and Brent-Dekker methods result in significant errors and unre-solved regions, respectively. The empty spaces in the error maps in each case showthe areas where a high accuracy solution was achieved. In comparison, the trust-region dogleg algorithm, with an error tolerance of 1e-8, successfully resolved allregions except those with KL < 0.05 (hence the mostly blank map in Figure 5.3).Lowering the error tolerance improves this limit at the expense of CPU cost. Alsoshown in Figure 5.3 are the residual errors for representative T -KL combinationswithin the unresolved and high error regions. In these cases, the residual asymp-totically decreases with increasing temperature, resulting in a false solution beingreturned.Further investigation of the Brent-Dekker algorithm failure mode revealed that961500 2000 2500 3000Temperature [K]0.511.522.53KL051015Error (%)1500 2000 2500 3000Temperature [K]0.511.522.53KL12345Error (%)1500 2000 2500 3000Temperature [K]0.511.522.53KL012345Error (%)1000 2000 3000 4000 5000 6000Presumed Temperature Solution [K]-0.0200.020.04Residual Error [-]T = 1800 KKL = 0.951000 2000 3000 4000 5000 6000Presumed Temperature Solution [K]-0.0200. Error [-]T = 2000 KKL = 1.71000 2000 3000 4000 5000 6000Presumed Temperature Solution [K]-0.02-0.0100.010.02Residual Error [-]T = 2600 KKL = 0.02Brent-DekkerNewton-Raphson Trust-region dogleglow errorregion (~1%)low errorregion (~1%)Unresolved Regionlow errorregion (~1%)low error over (almost) entire region (<1%)Figure 5.3: Algorithm performance comparison (temperatures) and repre-sentative solution residual profile: Assisted Newton-Raphson (left),Brent-Dekker (center), and trust-region dogleg with 1e-8 error tolerance(right).the regions which could not be resolved coincide with apparent temperature ratios(Ta,700/Ta,800 > 0.994) greater than 0.994 (Figure 5.4), which cause a very sharpovershoot of the residual close to the root, followed by the asymptote. Thus, a hy-brid algorithms comprising the Brent-Dekker and NLS optimization (trust regiondogleg) was selected. This hybrid algorithm utilizes the faster convergence timeof the Brent-Dekker and the robustness of NLS optimization (trust region dogleg).The Brent-Dekker algorithm is used for Ta,700/Ta,800 < 0.994, and the NLS op-timization is used elsewhere. The improved efficiency and increased robustnessof this hybrid algorithm provides faster processing of images or can be used forthe efficient generation of look-up-tables for image post-processing. Processing ofcombustion images from a single test in this work led to around 12 hours of totalcomputation time.971500 2000 2500 3000Temperature [K]0.511.522.53KL0.990.980.970.960.950.94Figure 5.4: Apparent temperature ratios map shows correlation withunresolved regions in Brent-Dekker and Newton-Raphson temperaturemap and sets a criterion to switch between Brent-Dekker andtrust-region dogleg with 1e-8 error tolerance5.1.2 Experimental ConsiderationsAccurate, correlated apparent temperature measurements are required as inputs toequation 2.4, though these measurements are affected by numerous experimentalfactors. To assess these, high speed two-color pyrometry imaging was applied tothe optical engine configuration described in §3.1 and the effects of parallax, cal-ibration, sensor response uniformity, and SNR improvement strategies were eval-uated. The engine and imaging system setup specifications used for the presentedanalysis are listed in Table 5.2 and the optical engine configuration for two-colorpyrometry is shown in Figure 5.5.The HPDI injector design described in §3.1 was used for direct injection ofNatural Gas (NG) and diesel. More details regarding the injector design and op-eration was presented in §2.2. The PIDING strategy presented here is not an op-timized operating condition calibrated for on-road use. Rather, it was selected forthe purpose of pyrometry method evaluation and development. Similarly, the dieseloperating point, carried out using the same HPDI injector, is utilized to demon-strate the sensitivity of two-color pyrometry system to imaging parameters anddata post-processing procedure. As such, the presented diesel operating point is98Table 5.2: Optical engine and imaging system specifications for two-colorpyrometry.Engine Ricardo Proteus in optical configurationDirect injector Westport 1st Gen. HPDI*Fuel Pump diesel, Natural GasSpeed 600-1000 rpmImaging assembly Phantom + Doubler (See § 3.2)Exposure 8-32 µsec at f/5.6Frame rate 7,200-12,000 fps (0.5 CAD resolution)Spatial resolution 624×304 pixels, 0.47 mm/pixelNarrow band-pass filters 700 and 800 nm CWL; 10 nm FWHM* High Pressure Direct Injection (See § 2.2)Figure 5.5: Two-color pyrometry imaging system implementation on the op-tical engine.99not representative of modern common-rail diesel injection systems, operating atmuch higher rail pressures. Table 5.3 lists the specification of the considered fuel-ing and imaging parameters.Table 5.3: Operating parameters for diesel and PIDING combustion.CombustionModePilotFuelMainFuelSpeed[rpm]GIMEP[bar]Pin j[bar]Exposure[µsec]Diesel Diesel Diesel 600 1.8 200 8 at f/5.6PIDING Diesel NG 1000 10.5 180 32 at f/5.6As described in §3.3.3, a skip-firing procedure was used and the final firedcycle from each repetition was imaged. Here, 3 fired cycles are followed by 15motored cycles, and this set is repeated 15 times (restricted by camera buffer). In-take air heating (≈ 80◦C) was used to ensure ignition of the pilot diesel fuel. Theflat piston bowl and lower compression ratio are expected to lead to more pre-mixing for the PIDING operation, especially at the loads considered here. Suchpartially-premixed NG strategies have been noted elsewhere to result in significantreductions in engine-out PM [38, 99]. This represents an even more challengingcondition for pyrometric imaging, as the corresponding lower in-cylinder soot con-centrations will result in lower imaging signal intensities.Images were captured using the high-speed Phantom camera and the 60 mmlens coupled to the image doubler. Details of these components are presented inChapter 3 and the layout is illustrated in Figure 5.5. The use of flat and prism mir-rors in the image doubler design to project two combustion images onto the sensorprovides a compact and robust method to capture two images with one sensor. Thisresults in a lower signal attenuation compared to approaches using a beam splitter;however, it will introduce a parallax error, which must be addressed to ensure cor-related apparent temperatures are considered.5.1.3 System Calibration and Image ProcessingThe imaging system response was calibrated to relate the measured pixel intensi-ties, D (i.e., “counts”) to the apparent temperature in equation 2.4. Furthermore, astwo spectral intensities are required for each pixel position, a geometric calibration100Figure 5.6: Work flow to obtain temperature and KL distributions fromrecorded raw signal.was also implemented to ensure that the pixel spectral intensities are correlatedbetween the two images. This geometric calibration includes an affine transforma-tion, as well as static and dynamic parallax corrections. Finally, the SNR of theimages was improved through image filtering. Figure 5.6 shows work flow to eval-uate the temperature and KL for each image pair, beginning with the combustionradiation. The geometric corrections are applied to the recorded pixel intensitiesfrom individual cycles Dλ ,cyc prior to ensemble averaging to improve the SNR.Ensemble averaging to reduce noise was selected over ensemble averaging of Tand KL, as solutions (T, KL) are not available for each pixel for each image andcan result in unrepresentative mean values (see Appendix B). The sensor responsecalibration is applied to the ensemble averaged pixel intensities D′λ to evaluate theapparent temperature, after which KL and T are calculated for each pixel. In thefollowing, the effect of each of the geometric and sensor response calibration op-101erations are evaluated based on their impact on the calculated temperature and KLfactors for the diesel operating mode (see Table 5.3).Camera Sensor Response CalibrationThe instantaneous spectral apparent temperature Ta,λ (xi, j, t) is correlated to theimaging system output using the integrating sphere described in §3.2.2, wheredetails of camera sensor response calibration using the integrating sphere is alsopresented. As discussed, extrapolation to combustion relevant intensities is stillrequired is spite of the high intensity light source selected in this work. A highresolution calibration of the CMOS sensor showed that the camera chip does notprovide a linear response between the illuminating light and the detected charge, ashas been noted in literature [28, 59]. This effect was quantified and compensatedfor during the sensor response calibration process. An extrapolation approach sim-ilar to Jakob et al. is used [64], whereby the sensor response is linearized to fa-cilitate extrapolation. In addition to the non-linearities at high intensities noticedand considered in their work, non-linearities at lower intensities were captured andaccounted for. The geometrically corrected pixel intensities D′ are linearized us-ing a fourth order polynomial, based on the intensity calibration. The apparenttemperature of the combustion radiation is then:Ta,λ = β × (Xτ ×D′lin,λ )+ γ (5.1)where D′lin,λ is the linearized, ensemble averaged, corrected pixel intensity,β and γ are calibration coefficients, obtained from the sensor response calibrationprocess, correlating the pixel intensity and apparent temperature, and Xτ is the ratioof the exposure time used during the combustion imaging to the exposure time forcalibration. The exposure times are different during calibration and combustionimaging to ensure that the camera sensor is not saturated.Application of the response linearization affected both the calculated tempera-ture and KL factor, as shown in Figure 5.7. A more significant effect was noted forthe KL factor than for T. Both the temperature and the KL factor were most affectedalong the flame boundaries, where mean temperature and KL factor differences of> 8% (200 K) and > 60% are present, respectively, relative to an analysis without102Figure 5.7: Influence of linearized extrapolation on two-color pyrometry re-sults. Images shown for 9 CAD aTDC, diesel combustion. Geometriccalibrations were implemented, and temperature and KL were calculatedusing the hybrid algorithm.linearized extrapolation. The high relative differences for the KL factor around theedges of the flame is, in part, caused by the low absolute KL factor values resolvedwhen response linearization is used.Spatial non-uniformities in the intensity calibration can affect the local tem-perature and KL factors and may be caused by vignetting from optical components(e.g., image doubler, lens) and non-uniform CMOS sensor response [143]. To char-acterize these, the integrating sphere was translated while mounted on the engine,and the response was evaluated at several positions and used to correct the nominalintensity calibration. The effect of the non-uniform response on the temperatureand KL fields is shown in Figure 5.8, where a significant difference is noted at thebottom of the images. The average error over the entire region was 4% for tem-perature and 20% for KL factor. A later crank position image is considered here toshow the sensor response inhomogeneity near the piston bowl perimeter.103280024002000160015105010.[K]Offset[%]KLaccounting for spatialnon-uniformitiesestimation offset from spatialnon-uniformitiesFigure 5.8: Effect of spatial non-uniformities imaging system response(caused by sensor, optical elements, and vignetting). T and KL werecalculated using the hybrid algorithm. Images are for diesel combustionat 21 CAD aTDC.To maximize the detection envelope of the system, it was configured such thatthe camera sensor saturation limit is the same for both wavelengths. This is af-fected by the spectral response of the camera, spectral transmissivity of the opticalcomponents, and the spectral emissivity of the soot cloud. A neutral density filter(0.2 OD) was used to attenuate the 700 nm light, which resulted in sensor satura-tion for both wavelengths at approximately the same apparent temperature. Fig-ure 5.9 illustrates the detection envelope for the imaging system with and withoutmatched dynamic ranges. The detection envelope is bounded at lower tempera-tures by background noise and at higher temperatures by saturation of the imagingsensor. By matching the dynamic ranges of the two wavelengths, the temperaturedetection envelope is increased by ≈ 25% for temperatures at 0.5 < KL < 3 andby ≈ 30% for KL (at 2600 K). With the extended detection envelope, soot particlesin the 1720-2300 K temperature range at higher soot concentrations (KL = 3) and104Figure 5.9: Increase in detection envelope by the imaging system dynamicrange with matching aperture response.the 2050-3000 K temperature range at low soot concentrations (KL = 0.1). Lowertemperature soot particles (to the left of the marked region in Figure 5.9) cannot beresolved due to the pyrometry signal being too low for the current settings.Geometric Corrections and SNR EnhancementTo ensure that the apparent temperature (pixel) pairs correspond to the same phys-ical position in the cylinder, the recorded images are corrected using an imagetransformation based on the system geometry and parallax caused by the imagedoubler. To account for imperfect image corrections, filtering was implemented toimprove SNR.Parallax : To correct for parallax caused by the doubler, as well as other geo-metric differences between the images, a calibration target was positioned at thefiredeck plane (with cylinder head removed) and used to develop an image trans-formation matrix. Images of the parallax calibration target were acquired with allrelevant optical components (window, mirror, doubler, and lens) in the light path.Equation 5.2 describes the transformation matrix used to apply the affine and pro-105overlaid and transformed overlaid spatially correctedright imageleft imageAffine PorjectiveFigure 5.10: Parallax calibration target doubled image on the sensor (left),offset after affine transformation and the control point locations (cen-ter), and the fully registered doubled images (right).jective transformations:[u,v,w] = [x,y,1] ·sx cosθ sinθ px−sinθ sy cosθ pytx ty 1 (5.2)where s(x,y), t(x,y), p(x,y) are scaling, translation, and projective elements in x andy directions, respectively and θ is the rotation angle. After applying affine trans-formations (translation, rotation, and scaling), nine control points on each parallaxcalibration target (at 45◦ intervals and the center) were used to create eight cor-responding regions on the two images. The vertices of these regions, expressedin homogeneous coordinates (u,v,w), were used to fit a projective transformationmatrix mapping each region to its corresponding pair. The projective image trans-formation was performed on all eight regions independently and the image pieceswere then concatenated. The transformation matrix identified using this methodwas applied to every frame of a recorded image sequence. Figure 5.10 illustratesdifferent stages of the image registration process using schematic images of theparallax calibration target with exaggerated effects for more clarity. Affine trans-formation of the two distorted aperture images results in a doubled image and asuccessful registration of the two images is achieved through elimination of theperspective error using a projective transformation.Through the parallax correction, a significantly larger fraction of the image datacould be resolved into T and KL, as shown in Figure 5.11. In addition, this figurealso indicates the regions where no solution for equation 2.4 could be found, which106static parallaxadjustmentsdynamic parallaxadjustments280024002000160010. [K]KLno parallaxadjustmentsparallaxadjustedSoluon to Equaon 2 not possible13 CAD aTDC 23 CAD aTDCFigure 5.11: Impact of static and dynamic parallax corrections to increaseresolved pyrometric signal at different crank positions. Unresolvedregions are either due to Ta,700nm < Ta,800nm or Ta,700nm >> Ta,800nm.Linearized calibration extrapolation was implemented, and T and KLwere calculated using the hybrid algorithm.were found to coincide with conditions where: 1) Ta,700nm < Ta,800nm, which isnot possible based on Planck’s distribution, and 2) the difference between Ta,700nmand Ta,800nm was too large and resulted in calculated temperatures higher than theadiabatic flame temperature. Both of these conditions can be caused by imperfectimage registration or response calibration, as well as signal attenuation and wallreflections.Despite the improvements made by considering the parallax, there are still sig-107nificant unresolved regions, particularly late in the cycle. Consideration of parallaxeffects caused by the moving soot cloud, termed here dynamic parallax, enableda significant increase in the resolved T -KL space later in the cycle, as shown inFigure 5.11 (right). Dynamic parallax was corrected using an image transforma-tion matrix generated using the calibration target at firedeck (i.e., as above), as wellas at the piston crown position later in the cycle (23 CAD aTDC). For each crankposition the applied transformation matrix was an interpolation between the tworeference matrices, on the basis of the piston displacement.Binning/filtering : Pixel binning and spatial filtering can be used to improve theSNR of the unprocessed images [62], as well as to compensate for imperfect imageregistration and parallax correction. When binning is applied, the SNR is improvedby averaging the intensity of several pixels, at the expense of reduced resolution.Filtering (as opposed to binning) is a preferred approach for SNR improvement asit has less significant resolution impacts. Image filtering is applied through convo-lution of the image with a filter kernel, H. Figure 5.12 compares the temperaturefields evaluated using images with no SNR enhancement, 4×4 binning, a 4×4box-filter kernel, and a Gaussian filter kernel (γ = 1.5). The latter is commonlyreferred to as a Weierstrass transformation [157]. The total number of pixels witha recorded signal are indicated in Figure 5.12 to provide a characterization of theresolution. Also given is the fraction of the pixels for which a solution to equation2.4 was possible. The Weierstrass transformation resulted in the largest resolvedsignal fraction and was thus applied for the remainder of this work. For both fil-tering techniques (c and d), an increase in resolution is indicated, though this is,in part, due to pixels originally without a signal being assigned an intensity basedon neighboring pixels or having their unsolvable intensity pairs altered throughfiltering into resolvable values.The impacts of the individual refinements are presented above; however, theircombined effect is much more significant, particularly with regard to increases inthe fraction of the image data for which a solution to equation 2.4 is possible. Thefraction of the imaged data for which a solution could be evaluated is shown inFigure 5.13 for each of the proposed corrections individually, as well as the use ofall corrections and without any corrections. The uncorrected signal represents any1082800240020001600T [K]No binning/filtering Pixel binning (n=4) Box filter (n=4) ;78% of 15,590 pix 80% of 16,040 pix78% of 970 pix74% of 14,400 pixFigure 5.12: Comparison of pixel binning and filtering on calculated temper-ature fields. Image registration and non-linear response calibration areimplemented, and T and KL were calculated using the hybrid algo-rithm. Images shown for diesel combustion at 13 CAD aTDC.pixel with a recorded signal for both wavelengths, for which equation 2.4 could besolved, without applying any of the corrections. In this configuration, an averageof 38% of the pixels could be evaluated to give T and KL values. With the excep-tion of the parallax correction (55% resolved), other refinements do not result insignificant increases in the resolved signal area (hybrid algorithm 44%, linearizedresponse calibration 41%, and Weierstrass filtering 45% resolved). When thesecorrections are combined, a much more significant increase in the resolved sig-nal is possible (38% vs. 85% signal resolved). For the diesel operating conditionconsidered here, nearly all of the late-cycle signal is resolved which is relevant forrelating in-cylinder soot to exhaust stream measurements. Figure 5.13 also showsthe percentage difference in the spatially averaged two-color pyrometry tempera-ture and KL factor calculations with and without all the refinements. As shown inthis figure, while the temperature estimation difference remains within the ±10%band, differences of up to 80% was observed for KL factor estimations.1095 10 15 20 25 30Crank Angle [CAD aTDC]-20020406080100Before-After Difference [%]TemperatureKL factor10% error band102030405060708090100Resolved Signal Fraction[% available signal]102030405060708090100Available and ResolvedSignal [% image area]Available SignalNo CorrectionsAll CorrectionsParallaxWeierstrass FilterResponse CalibrationHybrid AlgorithmFigure 5.13: Effects of individual and combined corrections on the resolvedimage data, resolved data fraction, and spatially averaged T and KLfactor estimations for the diesel operating mode.1105.2 Sample ResultsThe imaging system and proposed refinements were evaluated using two combus-tion strategies: direct injected diesel and Pilot Ignited Direct Injection Natural Gas(PIDING) combustion (see Table 5.3). While numerous works have applied two-color pyrometry to diesel combustion, PIDING has only been evaluated using anintegral pyrometry system (i.e., not spatially resolved) [54]. As discussed in Chap-ter 2, PIDING is expected to be a challenging application for pyrometry in light ofthe lower PM concentrations and lower combustion temperatures associated withpartially-premixed natural gas combustion. Furthermore, the relatively low-loadand non-EGR conditions conserved here are expected to result in relatively lowsoot concentrations, even by PIDING standards, and thus add to the difficultyof pyrometry measurement. To demonstrate the efficacy of the proposed correc-tions, temperature and KL images are shown for diesel and PIDING combustionin Figures 5.14 and 5.15, respectively, with all corrections and with only affinegeometric corrections and matched dynamic ranges. Through the additional cor-rections (static and dynamic parallax, response non-uniformity, and spatial non-uniformities), a larger portion of the soot cloud could be resolved. The diesel andPIDING combustion modes are only discussed briefly below to demonstrate theutility of pyrometry enhancements and a much more thorough evaluation of thesoot processes is presented in the next chapter.The diesel combustion shown in Figure 5.14 is in agreement with the concep-tual mixing controlled combustion model presented by Dec [24], as well as two-color pyrometry results reported by Payri et al. [117]. In particular, a significantportion of the soot is generated through a mixing controlled flame established bythe fuel spray. During later stages of the cycle, the KL signal decreases in bothmagnitude and area due to soot oxidation and advection of the tail of the reactingjet. The net result is a ring-shaped soot cloud that remains at the bowl periphery forthe rest of the cycle, in part due to the low swirl in the considered engine. Althoughthe uncorrected two-color pyrometry image sequence indicates a similar behavior,the refinements result in a significantly higher fraction of the signal being resolved(see Figure 5.13), generally for regions with higher KL factors. This effect is ofparticular significance later in the cycle (e.g., after 20 CAD aTDC) where the re-111Figure 5.14: Temperature and KL distributions for diesel combustion (see Ta-ble 5.3) using proposed refinements developed in this work. Imageswith only affine corrections and matched dynamic ranges are shown toindicate the improvements due to the refinements.finements can increase robustness of the pyrometric method for weak signals andprovide insight into the late-cycle KL and temperature values, which are closelyrelated to the exhaust stream emissions [76].For PIDING combustion, shown in Figure 5.15, the KL factor is generallylower, relative to diesel combustion, as expected with natural gas as the primaryfuel. The high methane content and the low load, which leads to premixing ofthe natural gas before ignition, results in lower temperatures and less potential forsoot due to the low aromatic content of the fuel. In contrast to diesel combustion,the soot cloud is first detected at the piston bowl periphery and then propagatesinwards to form a more uniform region. This is followed by a reduction in the sootarea due to soot oxidation. While the uncorrected two-color pyrometry results forPIDING combustion indicate the location and timing of detectable soot, neither thesoot cloud size, nor its growth rate could be resolved when the corrections are not112Figure 5.15: Temperature and KL distributions for PIDING combustion (seeTable 5.3) using proposed refinements developed in this work. Imageswith only affine corrections and matched dynamic ranges are shown toindicate the improvements due to the refinements.applied.To quantify the impact of the corrections for PIDING, the resolved signal areaand resolved signal fraction are shown in Figure 5.16. This improvement is of par-ticular significance in the case of PIDING combustion, where the light intensitywas 75% lower than for diesel operation, based on the required camera sensor ex-posure to provide suitable dynamic range (see Table 5.3). For the majority of thecycle the resolved signal fraction is approximately doubled. While the late cycleperformance was increased, the low resolved fraction does indicate that this is achallenging application. It should be noted that there is little signal available af-ter ≈24 CAD. The percentage difference in the spatially averaged T and KL factorcalculations, with and without the proposed corrections, are also shown in Figure5.16, which indicates up to 50% error in KL factor estimations, while the tempera-ture estimation still remains within the ±10% difference band.113102030405060708090100Available and ResolvedSignal [% image area]Available SignalNo CorrectionsAll CorrectionsParallaxWeierstrass FilterResponse CalibrationHybrid Algorithm102030405060708090100Resolved Signal Fraction[% available signal]5 10 15 20 25 30Crank Angle [CAD aTDC]-40-200204060Before-After Difference [%]TemperatureKL factor10% error bandFigure 5.16: Effects of individual and combined corrections on the resolvedimage data, resolved data fraction, and spatially averaged T and KLfactor estimations for PIDING combustion.1145.3 Summary and ConclusionsThis chapter focused on identifying and assessing measures towards improved highspeed, two dimensional pyrometric imaging in a compression ignition engine. Thiswas carried out through revisiting implementation procedure, algorithm selection,imaging system calibration, and image post-processing. The refined technique wasused to consider direct injection diesel combustion and natural gas combustionin an optically accessible engine. It should be noted that the enhancements pro-posed in this work will not overcome the short-comings of the two-color pyrome-try method inherent with line-of-sight methods, nor will they address uncertaintiesdue to window contamination or wall reflections.Several improvements to the two-color pyrometry method were characterized:1. Use of a combined non-linear least square optimization algorithm (for highapparent temperature ratios) and Brent-Decker method (for lower appar-ent temperature ratios) reduced the computation time by 95% relative to aNewton-Raphson method, and resolved 99% of the measurement space with< 1% error (≈ 80% for Newton Raphson). The total computation time fora single test (≈ 107 analysis points) was close to 12 hours with the operat-ing system used in this work. In addition, use of this combined approachresulted in a more robust algorithm which allowed equation 2.4 to be solvedfor a greater range of apparent temperatures. This algorithm can be used forevaluation of images, or for the generation of look-up tables for subsequentprocessing of images.2. Static and dynamic parallax correction resulted in an increased fraction ofthe image being resolved (≈ 1.3− 2× increase). The former is a result ofthe imaging system used, and the latter due to the piston motion. Utilizingthese corrections, it is possible to utilize a compact and robust image doubler,simplifying the hardware and avoiding the SNR penalty associated with abeam splitter.3. Weierstrass spatial filtering improved the SNR and provided increased im-age resolution and resolved signal fraction, relative to pixel binning and boxfiltering, respectively. The increase in resolved signal fraction is attributed115to the increased signal to noise ratio for individual pixel, resulting in bettercorrelation between the two images.4. By matching the imaging dynamic range for each considered wavelength,the dynamic range of the camera is fully utilized for both wavelengths. Thisincreased the KL and T detection envelopes by ≈ 25% and ≈ 30%, respec-tively.Application of all of the refinements resulted in an increased resolved signalfraction for diesel and PIDING combustion strategies. For diesel combustion, atemperature and KL could be evaluated for > 80% of the image data, while forPIDING 40-80% of the image data could be resolved. It should be noted that 100%of the data cannot be resolved due to line-of-sight integration, wall reflections, andensemble averaging resulting in unrepresentative pixel intensities. The refinementsenabled the pyrometric method to be applied to PIDING combustion, which hada 75% lower signal intensity than the considered diesel combustion. In the nextchapter, the refinements presented here are used to assess the soot formation andoxidation in the low sooting PIDING combustion strategy under a wide range ofoperating conditions.116Chapter 6Optical Investigation of PIDINGCombustionThe improved two-color pyrometry method, as concluded in Chapter 5, can be im-plemented to study soot processes in low soot signal combustion strategies. Thepilot-ignited direct-injection natural gas combustion is a good example of suchlow soot signal combustion scenarios. As such, this chapter focuses on utiliz-ing simultaneous spatially and temporally resolved two-color pyrometry and OH*chemiluminescence measurements to better understand soot formation and oxida-tion processes of PIDIING combustion.6.1 Engine Operating Conditions and Analysis MetricsIn order to better understand the underlying fuel conversion mechanisms and iden-tify the soot formation and oxidation modes during typical PIDING combustionoperation, a range of engine operating conditions were considered. A “baseline”operating point, representative of a medium-load engine operating condition, wasselected at the center of the operating space. The pilot fuel and NG injection timingfor this operating point was selected such that NG ignition happens during NG fuelinjection, with the goal to examine whether a quasi-steady, non-premixed jet flamecorresponding to mid- and high-load diesel combustion operation can be achieved[126]. The baseline operating condition selected as such, served as a datum to117study the influence of the major fueling control parameters. Sweeps of three differ-ent control parameters, namely the NG injection pulse width (GPW), the relativetiming between the pilot diesel and NG injections (RIT), and the fuel injectionpressure (Pinj) was considered.The investigated parameter space was selected through measurements underthermodynamic engine configuration with steady state operating conditions. Thesemeasurements were used to identify the appropriate optical engine control settings,including pilot fuel and NG injection timing and duration and intake air conditions,as well as to evaluate air and fuel flow rates. Table 6.1 lists the operating conditionsinvestigated in this work. The selected operating conditions listed in this table werenot derived from or representative of any specific engine calibration.Table 6.1: Operating points specifications. cPSOI: commanded Pilot Start ofInjection; cGSOI: commanded Gas Start of Injection; PPW: Pilot PulseWidth; GPW: Gas Pulse Width; Pinj: fuel injection pressure; RIT: Rela-tive Injection Timing. The fuel system controls the natural gas rail pres-sure to a level ~10 bar below that of the diesel rail pressure.Point cPSOI cGSOI PPW GPW Pinj RIT GIMEP[◦ aTDC] [◦ aTDC] [ms] [ms] [bar] [◦] [bar]Baseline -18 -10 0.7 1.45 180 8 11.5Short GPW -18 -10 0.7 1.05 180 8 5.3Long GPW -18 -10 0.7 1.85 180 8 14.2Short RIT -11 -9 0.7 1.45 180 2 11.85Long RIT -24 -10 0.7 1.45 180 14 11.3Negative RIT -16 -22 0.7 1.45 180 -6 11.82High Pinj -16 -8 0.6 1.2 220 8 12.8Low Pinj -22 -14 1.0 2.1 140 8 11.2Diesel -18 -10 0.7 1.75 180 8 2.2As discussed in Chapter 2 (§ 2.1.3) OH* chemiluminescence signal can beused to identify the non-sooting flame area and high temperature reaction zones(local heat release), and it is also believed to be a good marker for soot oxidation[25, 94, 159]. Simultaneous two-color pyrometry and OH* chemiluminescencemeasurements provides more insight into in-cylinder soot analysis by considera-tion of these high temperature reaction zones and local heat release information inconjunction with soot formation and oxidation processes. The configuration of the118engine and the measurement setup is shown in Figure 6.1 and the imaging systemsettings are presented in Table 6.2. Additional details on the imaging system andtest facility is available in Chapter 3.To facilitate the imaging, the skip-firing procedure was used in which 3 firedcycles followed 17 motored cycles and only the final of the three fired cycles wasused for imaging purposes (see §3.3). The combustion images presented below areensemble averages of 15 cycles obtained using this skip-firing method. The lowcompression ratio (CR) required that the intake air be heated (~50-60◦C) to ensureignition of the pilot diesel fuel. Intake air pressure was adjusted to compensate forthe air density change due to the change in temperature.Table 6.2: Imaging system specificationsTwo-color Pyrometry OH* ChemiluminescenceCamera Phantom v7.1 CMOS Photron SA1 CMOSIntensifier - LaVision high-speed IROImage doubler LaVision -Gain - 70%Exposure/Gate 8-64 µsec @ f/5.6 10 µsec @ f/5.6Bit-depth 12 12Lens 60mm-f/2.8 Micro-Nikkor98mm-f/2.8 Cerco UVCamera Resolution 624x304 640x640Physical Resolution 0.47 mm/pixel 0.19 mm/pixelNarrow band-passfilters700 & 800 nm CWL 307 nm CWLfilter bandwidth 10 nm FWHM 20 nm FWHMFrame rate 12,000 fpsEngine Window 83 mm unblocked119Figure 6.1: Simultaneous two-color pyrometry and OH* chemiluminescenceimaging system6.1.1 Analysis MetricsThe analyses presented on in-cylinder soot processes in this chapter are basedon 2D and spatially averaged temporally resolved two-color pyrometry and OH*chemiluminescence, as well as HRR results, and also in correlation with pilot andNG injection pulses, presented in terms of estimated mass flow rates normalizedby the baseline value (more details in Appendix C). Mechanical injection delaywas estimated through illuminated combustion chamber optical measurements andneedle ramp-up and ramp-down timings were estimated in a previous work [37].The extent and mode of soot cloud development and combustion chamber cov-erage, temperature and KL factor levels, and late-cycle oxidation rates are the majorfactors that determine the engine-out soot. In order to better understand the pro-cesses controlling these determining factors, the rate of soot formation and oxida-120Figure 6.2: Net soot formation and oxidation for baseline PIDING operat-ing point and the corresponding detected soot formation and oxidationdurations.tion, as well as its correlation with instantaneous distribution of soot temperaturesand KL factors was investigated in the current analysis. A normalized net value forthe rate of soot formation was considered, which is described in equation 6.1. Thisprocess is illustrated for the pyrometry results from baseline operating condition inFigure 6.2.K˙Ln =[dKL/dθ∆θ]f orm(6.1)where ∆θ f orm is the crank position range from the first detected soot until thepeak KL and dKL/dθ is the mean production rate within ∆θ f orm range. The meanproduction rate was normalized by the formation duration in equation 6.1 in orderto isolate that effect in comparison of the results.To supplement the possible interpretations from pyrometric imaging results,the temperature and KL area fraction distributions, AT and AKL, were considered.At each crank position, the area fraction distribution is described using a “set”121(denoted by {}):Aξ (θ) ={n j =1Npix,totNpix,tot∑i=i( pixi |ξpixi > ξ j)}; j = 1, ...,m; (6.2)where ξ is the metric of interest (T or KL), Npix,tot is the total number of pixelswithin the bowl area, subscript j denotes the T or KL bin and m is the total numberof bins, and n j is the fraction of Npix,tot falling in bin j. The fractional area distribu-tions for all crank positions can be shown in a Cumulative Histogram Time Series(CHTS) diagram [125]. Figure 6.5, discussed in §6.2.1, shows an example of thearea fraction CHTS diagrams for the baseline operating condition. Area fractiondistributions were calculated for individual cycles and then ensemble averaged ateach crank position.6.2 Results and DiscussionThe focus of the current work was on the optical investigation of the described pa-rameter sweeps, specifically soot formation and oxidation processes. The study ofthe pilot and NG ignition and fuel conversion modes, using high-speed OH* chemi-luminescence and natural-luminosity imaging, was the focus of a parallel work inthe group [126]. This investigation provided an updated conceptual understandingof the governing mechanisms controlling the PIDING combustion.In order to facilitate the interpretation of the two-color pyrometry and OH*chemiluminescence results, in conjunction with the hypothesized processes, thissection first focuses on describing typical soot formation and oxidation trends forthe baseline operating point. This is followed by highlighting the effects of consid-ered parameter sweeps on the described trends and the correlations between variousdiagnostics during each parameter sweep. Based on these results, the next sectionattempts to make the link between the various parameter sweeps and help under-stand the soot formation and oxidation processes in accordance with the currentconceptual description of PIDING combustion modes.122Figure 6.3: Baseline PIDING operating point; Ignition and early stage com-bustion. Left: Overlaid OH* images of pilot and NG ignition from theircorresponding crank positions showing their spatial correlation; Center:OH* images of pilot ignition and NG ignition and ignited reaction zonegrowth; Right: HRR and spatially averaged OH* results6.2.1 Baseline Operating ConditionThe baseline operating condition was designed based on comparisons with typicalHRR response in a production engine. In order to achieve this, the combustioncontrol parameters, namely the fuel rail and intake boost pressure, commandedstart of pilot and NG injections and the relative injection timing between these twowere adjusted. The absolute value of these control parameters will not match thoseof a production engine due to the differences in the compression ratios, bowl ge-ometries and intake air temperatures. As a standard PIDING operating point, asmall amount of pilot diesel fuel is injected at 18 CAD bTDC, prior to injection ofthe primary natural gas fuel (10 CAD bTDC), to ignite and provide the activationenergy required to ignite the lower cetane natural gas. Figure 6.3 shows the OH*chemiluminescence images of the pilot and NG ignition as well as early combus-tion processes.The optical ignition timing and delay (θign and τign, respectively) were char-acterized based on the first indication of the OH*-CL signal, for pilot and NGfuels. Comparison of optical and thermodynamic ignition delay values for operat-ing points in Table 6.1 are available in Appendix D. OH* images of the baselinePIDING ignition phase shows that the pilot diesel fuel is ignited after estimated pi-lot end-of-injection (EOI), with 12.5◦ optical ignition delay. At 4 CAD bTDC, the123OH* chemiluminescence images in Figure 6.3 shows that pilot combustion createsa ring of “high reactivity” approximately half-way through the combustion cham-ber, i.e., a ring shaped high temperature reaction zone comprising ignited pilot fuelcombustion products and radicals, which matches the pilot HRR peak in phasing.The HPDI injector used in this work had 9 pilot and NG fuel injection holes, with20◦ interlace angle between each pair of holes. Thus, the natural gas is injectedin the space between ignited pilot kernels where it ignites on the jet axis along thesides of the quasi-steady gas jet as it passes through the high energy ring. Thisprocess is illustrated in the overlaid OH* chemiluminescence images from pilotand NG ignition in Figure 6.3 (-4 and 1 CAD aTDC, respectively), and was alsopredicted by numerical simulations [90]. Using a 0◦ interlace angle, as OH* imag-ing of NG ignition process showed [54], would cause the ignited pilot kernels toengulf the NG jet and ignite it at the head of the jets.After ignition, high temperature reaction zones develop in predominantly radialdirections on the NG jet axes (OH* chemiluminescence in Figure 6.3, 1-3 CADaTDC), with a calculated average reaction zone growth rate of approximately 10m/s [126]. No soot was resolved within the detection limit of the current imag-ing setup during the diesel pilot and primary NG combustion prior to the peakHRR. This is evident from Figure 6.4, which shows the process of soot forma-tion, soot cloud growth and oxidation for the baseline operating condition, throughspatially resolved and averaged pyrometry and OH* chemiluminescence, as wellas HRR results. This suggests combustion of a sufficiently mixed charge of pilotdiesel and NG during this phase. During this early stage of combustion, the meanOH* chemiluminescence signal follows HRR trend very closely and OH* signalis concentrated on the NG jet axes for NG ignition and early primary combustion.Filtered NL measurements (λ >340 nm) of the NG ignition and early combustionbefore the peak HRR by Hatzipanagiotou et al. [54], also show low NL signal,originated from the chemiluminescence emitted by CH*, C2*, CN* and CO2* rad-icals present in the same range of the spectrum, in addition to the soot luminescence[88].The mean imaging parameters (OH∗, KL, and T ) in Figure 6.4, and all subse-quent figures in this chapter are evaluated from the OH* images and calculated KLand T distributions as follows:124OH∗ =1Npix,tot∑pixOH∗i (6.3)KL =1Npix,tot∑pixKLi (6.4)T =1Npix,res∑pix,resTi (6.5)where subscript i denotes each pixel in processed images, Npix,tot is the totalnumber of pixels. For T only pixels for which a temperature can be solved (Npix,res)are considered. For OH∗ and KL all pixels are considered. T is plotted for crankpositions where KL > 0.01, which is ~5% of the peak KL for the baseline operatingcondition, given the low SNR and growing error bars outside this region (shadedarea around T and KL curves).As natural gas fuel conversion continues, the high temperature reaction zones,localized along the gas jets, reach the bowl wall region and PIDING combustionproceeds to the primary phase exerting heat at higher rates (Figure 6.3, 3 CAD aTDC).This is followed by the onset of soot signal detected at the points of NG jet im-pingement on the wall, forming small soot “pockets” on individual NG jet axesaround the periphery of the bowl wall. The detected onset of soot signal (6.4,4 CAD aTDC) temporally coincides with formation of toroidal shaped zone ofhigher temperature reactions from OH* chemiluminescence, with higher signallevels than off-axis area between NG jets.During the rest of NG injection, each initial soot pocket grows and merges intoneighbor pockets, forming a toroid adjacent to the wall (4-6.5 CAD aTDC). Thesoot cloud toroid then expands and progresses towards the center, as a result ofcontinued soot formation and reflected momentum of the NG jets impinging on thebowl wall. The latter process was discerned through tracking small high intensitysoot pockets at higher temporal resolution in the range of 7.5-12 CAD aTDC, pre-sented in Appendix E, and also through comparison of parameter sweep effects asdiscussed later in this chapter.Comparison of the temporal changes in intensity gradients of the 2D OH*chemiluminescence signal and KL factor distributions in Figure 6.4 (4-8.5 CAD aTDC),125Figure 6.4: Baseline PIDING operating point; NG combustion and soot for-mation and oxidation. Results are ensemble averaged from 15 cycles(see §3.3.3).126depicts signal interference for measurable OH* signals. This is shown more clearlyby the dash-line rings on OH* image frames within this range, marking the leadingedge of the growing soot cloud. During this process, the OH* signal on NG jetsoutside the expanding soot cloud toroid remains undisturbed, while portions of theOH* signal within the soot cloud are partially attenuated by this optically thickermedium. The higher OH* signal zones on the NG jet axes then partially recoverlater in the cycle (Figure 6.4, 10-12 CAD aTDC), as parts of the soot cloud becomeoptically thinner through soot oxidation, before the final decay in the OH* signalthrough the late-cycle soot oxidation stage. Alternatively, the recovery of the OH*signal might be due to production of more OH* radicals and stronger chemilumi-nescence signal, or a combination of these two processes. On the spatially averagedresults, this results in a local minimum, followed by a 2nd OH* peak due to partialsignal recovery. The OH* chemiluminescence signal attenuation by the opticallythick soot cloud should be accounted for during interpretation of the OH* imagingresults and, in some cases, explains discrepancies in spatially averaged results, asdiscussed later in this work.High temperature reactions on the NG jet axes, producing higher intensity OH*chemiluminescence relative to off-axis space between jets, are visible until shortlyafter NG EOI ([1 8.5] CAD aTDC in Figures 6.3 and 6.4). This is not overlapped byany detected soot particles (lie within the described dash-line rings zone), whichsuggests during PIDING baseline operating condition a portion of the direct in-jected natural gas undergoes a low-sooting non-premixed combustion mechanismlocalized along the gas jets.Following the soot processes later in the cycle, the pyrometry signal starts todecay and disappear after the mean KL peak, starting from the wall regions (Figure6.4, 18 CAD aTDC). This process is accompanied by higher intensity OH* signalon the periphery of the visible bowl area, denoting overlapping high temperaturereactions and late-cycle soot oxidation process. This is followed by a more ho-mogeneous OH* signal distribution and soot signal decay at a slower rate in thecentral bowl regions, suggesting a more uniform soot oxidation process towardsthe end of the cycle. The soot oxidation process at different concentration levelsfrom after the peak KL factor through the late-cycle stage is also evident from theCHTS diagrams of the baseline operating condition shown in Figure 6.5. Here,127Figure 6.5: Area fraction distribution of two-color pyrometry soot temper-atures and KL factors at each crank position (CHTS diagram) for thebaseline operating condition. Results are ensemble averaged from 15cycles (see §3.3.3).there is notably higher decay rates in KL factor levels compared to more steadytemperature levels denoting consistent exothermic reaction of oxidizing the dimin-ishing soot particles. The comparison of KL factor peaks at different levels in thisfigure and the mean KL factor curve in Figure 6.4 shows matched phasing as ex-pected. However, focusing on the processes after 20 CAD aTDC shows low meanKL factor values during the late-cycle oxidation stage and large resolved areas withpredominantly low KL factor values.It is worth noting that the soot cloud signal disappearance, in addition to thelate-cycle oxidation processes, could in-part be due to the soot particles coolinglater in the cycle during expansion and transitioning below the current two-colorpyrometry detection limit. The two-color pyrometry system detection envelopeand its optimized adjustments were discussed in §5.1.3. Figure 6.6 shows T -KL cumulative distribution from all optically recorded combustion cycles after10 CAD aTDC. The enclosing red lines in this figure shows the pyrometry de-tection limits, determined by the dynamic range of the camera (12 bits) and thecalibration range. Here, the pyrometry system detection range was optimized toavoid sensor saturation at higher intensities. High number of resolved values canbe observed across the lower detection limit at KL<0.25 and temperatures in the128range of 1700-2600 K, whereas limited number of higher KL points were observedat the lower temperature limit (~1600 K). This is an indication of oxidation being asignificant contributor to pyrometry signal decay, compared to the cooling effects,especially during the late cycle oxidation stage.Figure 6.6: Aggregate resolved temperature and KL factor cumulative distri-bution for the baseline operating condition after 10 CAD aTDC. Resultsare ensemble averaged from 15 cycles (see §3.3.3).6.2.2 Fuel mass (injection Duration) EffectThe primary fuel (i.e., natural gas) injection pulse width, GPW, is used in combina-tion with the fuel injection pressure to control the injected fuel mass and the engineload, similar to diesel engines. Injection of more natural gas is expected to increasethe engine load at the expense of higher soot formation. The imaging and integratedmeasurement results illustrated in Figures 6.7 and 6.8, respectively, show the effectof extending GPW on PIDING combustion. The commanded start of pilot and NGinjection, pilot injection duration, and the fuel injection pressure are kept constantto facilitate assessment of the GPW sweep effects. As a result, optical pilot and NGignition timings and delays remain the same among GPW sweep operating points129Figure 6.7: NG fuel mass effects at constant injection pressure; 2D OH*chemiluminescence and two-color pyrometry results. Results are en-semble averaged from 15 cycles (see §3.3.3).(Figure 6.7, Θign,pi = -1 and Θign,NG = 1 CAD aTDC, respectively), as do otherprocesses described by the OH*, pyrometry, and HRR results, until the NG EOIin the shortest GPW operating point (~4 CAD aTDC). The overall processes as in-ferred from the evolution of OH* and soot distributions and the HRR shapes aresimilar although, as expected, more soot was produced in response to increasingthe NG injection duration.For the shortest injection duration (GPW=1.05 ms), NG EOI occurs before theonset of soot signal was detected at the wall region. The soot cloud does not13016001800200022002400T [K][a.u.]-15 -10 -5 0 5 10 15 20 25 30 35 40 45 50 55 60CAD aTDC050100150200250300HRR [kJ/CAD.m3 ] GPW = 1.05 msGPW = 1.45 msGPW = 1.85 ms0.*Chem. [a.u.]Figure 6.8: NG fuel mass effects at constant injection pressure; Spatially av-eraged OH* chemiluminescence and two-color pyrometry, net soot for-mation and oxidation rates, estimated normalized fuel flow rates (m˙N),and HRR. More window contamination and ensemble averaging effects(15 cycles) are responsible for the lower peak OH* at GPW = 1.85 ms(before NG EOI for GPW = 1.45 ms)131progress towards the center and soot oxidation takes over soon after formationof the soot cloud toroid. High temperature reactions also remain concentrated atthe wall region after NG EOI during the late-cycle oxidation phase. Compared tothe baseline, this operating point resulted in a lower peak HRR and lower averagetemperature in the combustion chamber (Figure 6.8) and lower soot production aswell as higher OH* signal with minimal soot cloud attenuation (Figures 6.7 and6.8).0 10 20 30 40 50 60CAD aTDC00. Area Fraction00. Area Fraction0 10 20 30 40 50 60CAD aTDC0 10 20 30 40 50 60 70CAD aTDC1380154016901850201021702330248026402800T [K] [a.u.]GPW=1.05 ms GPW=1.45 ms GPW=1.85 msFigure 6.9: NG fuel mass effects at constant injection pressure; Two-colorpyrometry soot temperature and KL factor area fraction coverage at shortGPW (left), baseline (center), and long GPW (right). Results are ensem-ble averaged from 15 cycles (see §3.3.3).For GPW = 1.85 ms, the soot cloud eventually covers the entire bowl area and amore homogeneous temperature distribution is resolved during the late-cycle stage,which is also shown in Figure 6.9. More soot formation and higher optical thick-ness with longer injection duration is observed in the larger area fractions with highKL values in this figure, as well as the high KL peaks in the spatially averaged re-sults (Figure 6.8), in addition to the 2D results (Figure 6.7). More significant OH*132signal attenuation with increasing GPW manifests in lower spatially averaged OH*signal after 7 CAD aTDC (Figure 6.8), leading to no signal recovery and 2nd peakfor GPW = 1.85 ms. Before 7 CAD aTDC (baseline NG EOI), however, the lowerOH* peak for GPW = 1.85 ms in Figure 6.8 is due to window fouling and is anartifact of ensemble averaging.6.2.3 Relative Injection Timing EffectThe relative timing between the commanded start of pilot and NG injections is an-other important PIDING combustion control parameter that impacts the degree ofNG premixing [99] and consequently, affects the soot formation and oxidation pro-cesses. Figures 6.10 and 6.11 show the spatially resolved and averaged results onthe effect of changing relative injection timing (RIT) from the baseline value. ThePPW and GPW values are held constant and combustion phasing approximatelymatches for all these operating points.As the imaging and HRR results in Figures 6.10 and 6.11 reveals, the sweepof RIT has a notable impact on the combustion mode and soot processes, althoughthese points lead to roughly the same engine load (see Table 6.1). There is nosoot signal in the negative RIT case (i.e., injection of NG prior to the pilot fuelinjection) and less developed soot cloud with lower peak KL is observed for bothshort and long RIT operating conditions. Higher 2D OH* signal levels and peaksare recorded and HRR curves are significantly different (Figures 6.10 and 6.11).The retarded cPSOI for the shortest RIT (RIT=2 CAD), closer to TDC thanthe baseline case, leads to a shorter pilot ignition delay and less jet penetrationbefore ignition, and consequently formation of a smaller high reactivity ignitedpilot kernel ring (Figure 6.12-a), which is soon quenched by the concurrent NGinjection before growing further (Figure 6.10, 1-4 CAD aTDC). Throughout themeasurement campaign, the RIT=2 operating point is unique in that it is the onlycase where soot signal is detected during the pilot fuel combustion phase. Thisis illustrated more clearly in 2D pyrometry results at a higher temporal resolutionwithin the 1-4 CAD aTDC range in Figure 6.13. Despite the small area fractionscovered by the pilot soot pockets, as shown in Figure 6.14, zooming into theseregions shows a rather uniform distribution of resolved temperatures and KL factor133Figure 6.10: Relative injection timing effect; 2D OH* chemiluminescenceand two-color pyrometry results. Results are ensemble averaged from15 cycles (see §3.3.3).13416001800200022002400T [K][a.u.]-15 -10 -5 0 5 10 15 20 25 30 35 40CAD aTDC0100200300400HRR [kJ/CAD.m3 ]RIT = -6 CADRIT = 2 CADRIT = 8 CADRIT = 14 CAD0.*Chem. [a.u.]Figure 6.11: Relative injection timing effect; Spatially averaged OH* chemi-luminescence and two-color pyrometry, net soot formation and oxida-tion rates, estimated normalized fuel flow rates (m˙N), and HRR. Re-sults are ensemble averaged from 15 cycles (see §3.3.3).levels, indicating consistent signal with sufficient SNR in the ensemble averagedresults. The shorter pilot ignition delay and less pilot diesel mixing, as a resultof injecting into a hotter chamber environment, and concurrent injection of NG,displacing the air, are the probable causes for a locally richer pilot fuel conversionand detectable soot signal during the pilot combustion. Filtered NL images ofpilot combustion during PIDING under higher compression ratio of 17:1 (Figure2.9, -9.8 and -5.7 CAD aTDC) also show signal levels comparable to those of the135Figure 6.12: Relative injection timing effect on pilot combustion (OH*chemiluminescence images); (a) smaller high reactivity ignited pilotkernel ring at RIT=2◦ relative to baseline as a result of retarded cP-SOI, closer to TDC, and less jet penetration before ignition, and (b)Overlaid NG ignition (1.5◦ aTDC) and gamma enhanced ignited pi-lot kernels (-8◦ aTDC). Results are ensemble averaged from 15 cycles(see §3.3.3).Figure 6.13: High temporal resolution 2D OH* chemiluminescence and two-color pyrometry results for pilot combustion at RIT=2 CAD. Resultsare ensemble averaged from 15 cycles (see §3.3.3).primary combustion stage (>6.4 CAD aTDC), which suggests non-negligible sootsignal [54].Despite the shorter pilot ignition delay for RIT=2 CAD, optical NG ignitiondelay is longer than that of the baseline (τign,NG =12.5 vs. 10 CAD), as a resultof the retarded pilot injection. This is followed by a higher peak HRR and higherintensity OH* chemiluminescence ring close to the walls and mean OH* chemilu-minescence peak, compared to the baseline (Figures 6.10 and 6.11). Despite thesame NG injection duration, RIT=2 CAD leads to less soot cloud formation and136Figure 6.14: Relative injection timing effects; Two-color pyrometry soottemperature and KL factor area fraction coverage at short RIT (left),baseline (center), and long RIT (right). Results are ensemble averagedfrom 15 cycles (see §3.3.3).growth, with lower KL factor values, and minimal attenuation of the stronger OH*signal. Similar to the GPW=1.05 ms operating point, the OH* signal peak matchesin phasing with the peak KL factor.The pilot ignition delay is longer for the RIT=14 CAD operating condition rel-ative to baseline (τign,pi =17 vs. 13 CAD) and is followed by a lower pilot peakHRR and weak OH* chemiluminescence signal, which was only visible throughimage gamma corrections at the current settings (Figure 6.12-b). The extendedτign,pi did not appear to affect the τign,NG; nevertheless, NG ignition was more lo-calized on the gas jet axes. Overall, similar to the short RIT, this operating pointleads to a higher peak HRR, OH* signal at the wall region, and lower KL factorlevels and peak mean value. The comparison of area fraction distributions shownin Figure 6.14, however, shows more dominating higher temperature regions laterin the cycle as the RIT is decreased.137Early timing of the NG commanded start of injection for the negative RIToperating point (RIT=-6 CAD) results in a significantly longer NG ignition delay ofτign,NG =20 CAD. HRR curve is broadened and negligible soot signal was recordedthroughout the cycle rendering the soot temperature calculations not possible dueto the SNR being too low.Figure 6.15: Fuel injection pressure effect; 2D OH* chemiluminescence andtwo-color pyrometry results. Crank positions are synced relative to theonset of soot (aOS) signal and crank position aTDC for each operat-ing point is reported at the top-right of the OH* images. Results areensemble averaged from 15 cycles (see §3.3.3).1386.2.4 Injection Pressure EffectThe fuel injection pressure (Pin j), similar to diesel engines, is another key controlparameter to modify engine load and fundamental combustion processes in PID-ING combustion and, thus, has been a major focus in the PIDING combustionstudies to date [34, 98, 99]. The density of the injected fuel and its mass and mo-mentum flow rate increases with the upstream fuel rail pressure. The fuel systemcontrols the natural gas rail pressure (GRP) to a level ~10 bar below that of thediesel rail pressure (DRP). Figures 6.15 and 6.16 show the spatially resolved andaveraged results on the influence a of ~20% increase and decrease in the Pin j. Dura-tion and timing of the injection pulses where adjusted to maintain similar injectedfuel mass and combustion phasing for the considered Pin j sweep operating points.As such, in order to facilitate the comparison, the crank positions in the 2D resultspresented in Figure 6.15 are relative to the onset of soot signal (aOS).Decreasing Pin j to 140 bar spread the onset of soot signal region upstream onthe NG jets to also include regions away from points of impingement on the bowlwall. This is in contrast with other operating points listed in Table 6.1. This on-set of soot signal is followed by a higher KL factor distributions at the bowl walland along the gas jets (Figure 6.15, 3-9.5 CAD aOS), higher mean KL peak and alower peak HRR and mean soot temperature (6.16). In contrast to the long injec-tion case (GPW=1.85 ms, Figure 6.7), the high KL factor region for the low Pin joperating point in Figure 6.15 remains concentrated at the proximity of the bowlwall throughout the cycle.13916001800200022002400T [K][a.u.]-15 -10 -5 0 5 10 15 20 25 30 35 40 45CAD aTDC050100150200250300HRR [kJ/CAD.m3 ]Pinj = 140 barPinj = 180 barPinj = 220 bar0.*Chem. [a.u.]Figure 6.16: Fuel injection pressure effect; Spatially averaged OH* chemilu-minescence and two-color pyrometry, net soot formation and oxidationrates, estimated normalized fuel flow rates (m˙N), and HRR. Results areensemble averaged from 15 cycles (see §3.3.3).Similar to the Pin j=140 bar operating point, increasing Pin j to 220 bar also led tohigher peak KL factor than the baseline case with traces of higher soot signal alongthe NG jet axes (Figure 6.15, 4.5-7 CAD aOS). However, the Pin j=220 bar also in-creased the peaks for HRR and corresponding OH*, for both pilot and NG com-bustion (Figure 6.16, -3 and 5 CAD aTDC, respectively), relative to the baselinecondition, and resulted in higher engine load. High Pin j and baseline points other-wise show similar onset of soot signal locus, estimated temperature at the peak KL,and OH* chemiluminescence signal levels and attenuation from soot cloud.1400 10 20 30 40 50CAD aTDC00. Area Fraction00. Area Fraction0 10 20 30 40 50CAD aTDC0 10 20 30 40 50 60CAD aTDC1380154016901850201021702330248026402800T [K] [a.u.]Pinj=140 bar Pinj=180 bar Pinj=220 barFigure 6.17: Fuel injection pressure effects; Two-color pyrometry soot tem-perature and KL factor area fraction coverage at low Pin j (left), baseline(center), and high Pin j (right). Results are ensemble averaged from 15cycles (see §3.3.3).6.2.5 Updated Understanding of Soot Processes during PIDINGCombustionCombustion mode analysis of the current investigation campaign identified fivemajor stages to provide a conceptual description for the PIDING combustion pro-cesses, for a positive RIT value [126]. Figure 6.18 illustrates these through rep-resentative OH* chemiluminescence snapshots linked to corresponding sectionsof the HRR history for the baseline operating point described in Table 6.1. Ac-cording to the proposed conceptual description, PIDING combustion comprises(in chronological order): 1- Auto-ignition of a low-momentum pilot diesel “puff”jets away from the bowl wall; 2- ignition of the NG jets passing through the pilotreaction zones containing high temperature combustion products and radicals; 3-combustion of the partially-premixed NG at the wall region; 4- Non-premixed com-bustion with continuation of NG injection during and after the partially-premixed141fuel conversion stage; and 5: late-cycle oxidation of the remaining reactants andincomplete combustion products, after NG EOI.The reaction zone growth in a predominantly radial direction during the NGignition stage (Figure 6.18, stage 2) was attributed to NG fuel conversion alongthe gas jets, with a low HRR, and momentum of the fuel jets. In a previous work,local fuel concentration measurements showed evidence of methane concentrationincrease at the wall region before fuel was oxidized [156]. This suggests that anunreacted fraction of the injected NG penetrates beyond the ignited pilot kernelsand reaches the wall region, where it partially premixes with air while imping-ing on the bowl walls. Numerical simulation of equivalence ratio distributionsduring PIDING NG injection also showed NG jet wall impingement, under sim-ilar operating conditions [54]. Once the reacting NG reaches the bowl wall, thepartially-premixed charge is converted rapidly, which leads to a sharp rise in theheat release rate and formation of a toroid of higher temperature reactions at thewall region. The transition to the partially-premixed combustion was also charac-terized by reaction zone growth rates of more than 40 m/s [126]. Presence of atransitional combustion regime between stages 3 and 4 was acknowledged, wheredistinguishing between the simultaneous combustion modes was not possible.142Figure 6.18: HRR and OH* chemiluminescence images of combustion re-action zones through a typical PIDING combustion event. Reprintedfrom Rochussen et al.[126], with the permission from publisherThroughout the parameter sweep campaign presented in this work, there ap-pears to be a consistent correlation between the partially-premixed HRR peak andonset of soot signal detected at the points of NG jet impingement on the wall. Fig-ure 6.19 illustrates the correlation between the peak HRR and onset of detectablesoot signal crank positions. As pointed out before, the Pin j=140 bar operating pointshowed an earlier onset of soot signal, detected along the gas jets, and beforethe HRR peak and was an exception to this trend. Combustion of the partially-premixed charge at the wall region brings about higher fuel conversion and oxi-dation (oxidizer combustion) rates. Furthermore, continued injection of NG intothese high temperature reaction zones and air displacement also contribute to makethe wall regions locally richer zones. This is often accompanied by accumulationof low concentration soot formed on the NG jets, which are below the detectionlimit, especially for longer injection durations. Altogether, these processes result143-2.5 0 2.5 5 7.5 10Onset of soot[CAD aTDC]-2.502.557.510Peak AHRR[CAD aTDC]Short GPWBLLong GPWShort RITLong RITLow PinjHigh Pinjy =1.0x +0.2R2 =0.96Figure 6.19: Correlation between the peak HRR and detected onset of sootsignalin higher local soot formation rates and explain the matched phasing with the onsetof soot signal detected at the points of NG jet impingement on the wall. A simi-lar correlation was reported in a diesel engine study by Bobba et al. [14], wherethe soot mass in the cylinder started to become significant enough to be capturedoptically roughly at the peak premixed HRR. Prior to the described onset of sootsignal at the wall region, ignition and combustion of the low carbon:energy ratiogaseous NG on the free injected jets, does not lead to enough soot production tothe detection limits of the current imaging system.The peak KL factor timing for all operating conditions investigated in this workshows a correlation with the NG EOI and the resulting abrupt change in the rate ofavailable mixture, i.e., converted vs. introduced NG in the reaction zone. Figure6.20 illustrates this correlation across all operating points in the current campaign.During the GPW sweep, in each case the peak crank position is retarded with in-creasing injection duration, such that it still lies at a consistent shift from NG EOI.The RIT or Pin j effects did not affect the correlation between KL peak and NGEOI, despite altering fuel conversion and soot formation and oxidation trends, as1440 2.5 5 7.5 10 12.5 15NG-EOI[CAD aTDC]02.557.51012.515Peak KL[CAD aTDC]Short GPWBLLong GPWShort RITLong RITLow PinjHigh Pinjy =1.2x +1.2R2 =0.96Figure 6.20: Correlation between peak KL factor and NG EOIhighlighted in the previous section. These observations suggest that the NG EOI isthe determining factor in establishing the extent of soot formation and the peak KLphasing. As such, this crank position marks the end of high soot formation ratesand the dominance of the soot oxidation processes, in response to the end of NGsupply into the reaction zone.The peak HRR and onset of soot signal also corresponds to the 1st OH* chemi-luminescence peak, from the combustion of the partially-premixed charge, wherethere is a temporary drop in the signal, supporting the hypothesis presented on OH*signal attenuation from the developing optically thick soot cloud. For cases withminimal OH* signal attenuation, such as the GPW=1.05 ms and the RIT=2 CADoperating points, the undisturbed OH* curve peak matches the peak KL factor in-stead. This further supports the discussion on transition to oxidation dominatedstage and more clearly marks the peak of the soot oxidation process followed byrapid decrease in KL factor values at more steady temperature levels (e.g. Figure6.14, first column).These results reveal that there is an important correlation between PIDINGsoot formation and oxidation processes and the relative timing between the fuel145injection profile and the peak HRR. Based on these correlations, the extensionof the NG injection beyond the peak HRR and into the reaction zones leads toincreasingly mixing-rate limited fuel conversion [126], and higher soot forma-tion, optically thicker soot cloud development, and higher peak KL factors. TheGPW=1.85 ms and Pin j=140 bar operating points in Figures 6.8 and 6.16, respec-tively, illustrate these phenomena. Operating points with the lowest KL factor peaksand distributions (RIT=-6 and 2 CAD in Figure 6.11 and GPW=1.05 ms in Figure6.8), have their NG EOI advanced relative to the HRR peak. This revelation wasalso the basis for choosing the normalized net rates as described in §6.1.1. Since byextension of NG injection beyond the peak HRR, soot formation rate would con-tinuously increase, normalizing the net rates with the formation/oxidation durationcould isolate this effect and help assess the influence of the considered parametersweeps more independently.Continuation of NG injection during and beyond the partially-premixed HRRpeak, in addition to increasing local soot formation at the bowl wall, contributesto the soot cloud growth and progression towards the center through the reflectedmomentum of the NG jets impinging on the bowl wall. This phenomenon canbe inferred from examining injection duration effects and also comparison of theGPW=1.85 ms vs. Pin j=140 bar operating points (discussed in the next section),in addition to tracing small high intensity soot pocket trajectories, as describedbefore. In what follows the influence of considered parameter sweeps are analyzedin accordance with the soot formation and oxidation mechanisms described above.NG Fuel Mass (Injection Duration)With the extended NG injection duration in GPW=1.85 ms operating point, whilethe commanded injection settings and Pin j are held constant, the described sootcloud growth mechanisms are extended. As a result, soot cloud develops furthertowards the center and covers a higher percentage of the combustion chamber, andthe soot signal is extended until later in the cycle. While the mean KL factor peaksand trends are similar between this and the Pin j=140 bar operating points (Figures6.8 and 6.16), the soot cloud toroid in Pin j=140 bar remains more concentrated atthe bowl wall region throughout the cycle (Figure 6.21), despite the longer injection146duration (GPW=2.1 ms), as a result of the lower momentum of injected NG jets.This is illustrated more clearly in Figure 6.21, which shows time series of theradial distribution of the soot KL factor values. At each radial position, the meanKL factor value over a thin ring element (KLmean,R) was considered to generatethe radial distribution vector. In this figure, GPW=1.85 ms operating condition,after formation of high KL factor values at the bowl wall region (5-10 CAD aOS),shows high KL factor concentrations half-way into the bowl (10-25 CAD aOS).Pin j= 140 bar case, in comparison, shows high KL factor values at Rbowl earlierafter onset of soot signal and the KL factor concentration region remains at the wallregion.The distribution contours of Figure 6.21 are overlaid by the concentration cen-troid radius histories, described in equation 6.6. Here, GPW=1.85 ms case showstransition of the centroid radius back to half-way into the bowl at 15 CAD aOS,while at Pin j= 140 bar the centroid radius tends to remain further away from thecenter and shows a plateau around 15 CAD aOS.RC =∑Rbowli=0 (KLmean,Ri×Ri)∑Rbowli=0 (KLmean,Ri)(6.6)Conversely, for GPW=1.05 ms, with the early end of injection and lack of thedescribed reflected NG jet momentum effects, soot cloud does not progress towardsthe center. Here, NG EOI is notably advanced relative to the partially-premixedHRR peak and fuel is majorly injected during τign,NG (needle closing starts at ~0.5,Θign,NG =1, and ΘpeakHRR =4 CAD aTDC). As a result, this operating point leadsto predominantly partially-premixed fuel conversion, with higher OH* signal andless soot formation at the wall and lower peak KL factor. The lower NG inputbefore ignition of the partially-premixed charge also led to a lower HRR peak andmean soot temperature for GPW=1.05 ms operating point.The mean soot temperature for GPW=1.85 ms is also lower than that of thebaseline. The lower rate energy release within the rich mixture with injection ofmore fuel into the reaction zone causes an extended combustion phasing, smallerarea fractions at high temperatures and larger low temperature regions (Figure 6.9),and a lower averaged temperature estimation. As previously reported elsewhere[108, 162], it is possible that the decreased area fraction of the high temperature147Figure 6.21: Radial KL factor distribution time series for GPW=1.85 ms andPin j= 140 bar operating conditions (see Table 6.1); the overlaid solidblack lines mark the corresponding concentration centroid radius (seeequation 6.6). The dotted line on Pin j= 140 bar contour map is the pro-jection of GPW=1.85 ms concentration centroid radii for comparisonregions is due to the shrouding from the high optical thickness, low temperaturesoot cloud, which introduces a bias in pyrometry temperature estimates.Compared to the fraction distributions for the baseline GPW, the longer injec-tion duration results in Figure 6.9 (right column) show a smaller difference in thedecay rates for KL and temperature levels of 2000 K and below, after the meanKL peak (12.5 CAD aTDC, 6.8). Although there is an even steadier temperaturedistribution evolution at the high temperature levels (>2100 K) compared to thebaseline case, these regions only constitute a small fraction of the more developedsoot cloud (Figure 6.9, bottom-right, 20-40 CAD aTDC). This ultimately leads to alower soot oxidation rates, supported by early disappearance of OH* signal relativeto the soot cloud signal (Figure 6.7, 21-31 CAD aTDC), and higher ranges of KLfactors within the larger soot cloud and extended soot signal (Figures 6.7 and 6.9).Fuel injection pressureThe increase in Pin j of NG and the consequent increased fuel jet momentum flowrate was shown elsewhere to increase the NG jet penetration and mixing rates fora given charge density [113]. Also, higher rates of heat release during the mix-148ing controlled combustion stage of PIDING with higher Pin j was attributed to theincreased fuel-oxidizer mixing rates [126].For Pin j=140 bar operating point, the lower fuel jet momentum flow rate and theconsequent lower mixing rates result in richer charge mixture around jet axes. Theresulting richer zone at the bowl wall region makes NG conversion more bound bymixing, leading to a broadened HRR curve with a lower peak [126]. This leadsto higher soot formation and onset of soot being detected also on the NG jet axesand before reaching the bowl wall, unlike any other cases studied in the currentcampaign. Followed by more soot formation at the wall with the continued fuelinjection after the peak HRR, and with accumulation of more soot formed on thejet axes, a higher optical thickness soot cloud toroid was produced (2D KL). Thepoor charge mixing at the wall region and lower rate of heat release also results ina lower two-color pyrometry temperature. Study of fuel injection pressure effectsin multi-cylinder PIDING engines suggested that insufficient injection pressureresulted in “under-penetration” of the NG jets in the combustion chamber and poormixing of the gas with air, slow combustion and ultimately low thermal efficiencyand higher soot emissions [34, 98].At Pin j=220 bar, the higher peak HRR relative to the baseline, caused by thehigher fuel jet momentum and enhanced mixing, is also followed by a higher KLfactor peak. The higher KL values recorded along the gas jets (Figure 6.15, 10-12 CAD aTDC), mark the higher mass flow rate and higher local net soot formationrates. The effect of injected fuel mass flow rate on soot formation becomes moreclear through comparison of the normalized net soot formation rates at differentinjection pressures in Figure 6.22. Here, the combined effects of the increased rateof reactions, due to enhanced mixing, and increased mass flow rate at a higher Pin jlead to a higher normalized net formation rate. As illustrated in Figure 6.22, thenet soot formation rate, normalized over the entire span of the formation process,decreases with reducing Pin j, despite the higher peak KL factor. Comparison ofarea fraction distributions in Figure 6.17 shows a shift in the peaks of temperaturelevels towards the rising edge and onset of detectable soot signal with increasingPin j from 140 to 220 bar, which is consistent with expected higher rate of reactionsas discussed above. A similar normalized net soot formation rate was estimatedby increasing injection duration from 1.45 (baseline) to 1.85 ms (GPW sweep) as149GPW Pinj0123456Normalized Net SootFormation Rate [a.u.]10-3short/lowBaselinelong/highFigure 6.22: Normalized net soot formation rates for GPW and Pin jsweeps. A similar normalized net soot formation rate is estimated atGPW=1.85 ms with extending soot cloud growth mechanisms undersimilar Pin j and injection timing. Increasing Pin j, and the consequentmass and momentum flow rates lead to higher normalized net soot for-mation rates.similar soot cloud growth mechanisms are extended.In addition to the normalized net formation rates, reducing the Pin j from 220 to140 bar, also decreases soot cloud growth rates towards the center. The mean sootcloud growth rate was evaluated based on the crank position range from onset ofsoot signal at the wall until soot cloud growth to ~30% of the visible bowl radius.Comparison of mean soot cloud growth rates at different injection pressures are il-lustrated in Figure 6.23 and more details regarding the analysis metric is availablein Appendix F. At Pin j=140 bar, mean growth rate is reduced to 48 m/s (less thanhalf the rate at Pin j=220 bar), and the higher optical thickness region as a result re-mains concentrated at the proximity of the wall, which leads to more heterogeneityin distribution of KL factors over the bowl area. This phenomenon confirms thedescribed reflected jet momentum effects after impinging on the bowl walls (see150140 180 220P inj [bar]406080100120Mean Growth Rate [m/s]Figure 6.23: The influence of Pin j sweep on mean soot cloud growth rate.The rates are evaluated based on 2D pyrometry results (See AppendixF for details of analysis metric).Figure 6.21). At Pin j=220 bar, the initial soot pockets quickly merge and form thesoot cloud toroid, and develop towards the center at a higher rate. The higher mo-mentum flow rate causes more rapid cloud development following reflection offthe wall as well as further progression of the soot cloud towards the bowl center,despite the shorter relative timing between the NG EOI and peak HRR with Pin j.During the soot oxidation phase, the lower momentum flow rate and higherpeak KL factor of the Pin j=140 bar point lead to a lower oxidation rate compared tothe baseline. Despite the enhanced mixing at 220 bar Pin j, the normalized net sootoxidation rate is similar to that of the Pin j=140 bar case. This phenomenon can alsobe inferred from the comparison of the CHTS diagrams in Figure 6.17, where after~20 CAD aTDC temperature and KL levels decay with similar rates for the low andhigh Pin j cases, unlike the baseline operating point. In addition, noticeable lowtemperature area fractions for these two cases also indicate lower soot oxidationrate compared to the baseline point.Relative Injection TimingThe longer NG ignition delay for RIT=2 CAD allows for more charge mixing,especially close to the walls, and creates a more partially-premixed NG combustion151[126]. The delayed combustion of the partially-premixed charge led to NG EOIprior to the peak HRR, similar to the GPW=1.05 ms operating point. Combinedwith the combustion of the remainder of the diesel fuel, this results in the highestpeak HRR and OH* chemiluminescence signal in the current campaign. The highintensity OH* chemiluminescence ring indicates higher temperature reactions, andpresumably higher NOx formation [34, 100]. As a result of the more premixednature of this operating point, and despite the same injection duration, it leads tosignificantly less soot cloud formation and growth with lower KL factor values.Thus, there is minimal attenuation in the stronger OH* chemiluminescence signaland no temporary drop in the OH* chemiluminescence curve (8.5-15 CAD aTDCin Figures 6.10 and 6.11), whose peak matches in phasing with the peak KL factorinstead. This is similar to GPW=1.05 ms case, as the other predominantly partially-premixed operating point.Contrary to the RIT=2 CAD case, the advanced commanded pilot injection forRIT=14 CAD and injecting into a colder environment, resulted in a more homo-geneous reactive region from lean dispersed pilot diesel combustion with a longerτign,pi, a lower pilot peak HRR and OH* signal. Despite the unaffected τign,NG, inthe absence of stratified pilot kernels the NG ignition was more localized on the jetaxes and showed a slower growth, especially in peripheral directions from the jets.The less peripherally expanding high temperature reactions on the jet axes conse-quently lead to more partial premixing. Ultimately, similar to the RIT=2 CAD, thiscreates a higher partially-premixed peak HRR, higher temperature reactions andhigher OH* signal ring at the wall, and retarded onset of soot signal with lower KLfactor values and peak, compared to the baseline.The influence of premixing on the KL formation can also be elucidated by con-sidering the soot formation delay, i.e., the time from the start of NG injection untilsoot is first detected (τsoot =Θsoot−cGSOI). The retarded onset of soot signal rel-ative to cGSOI (increased τsoot) as a result of moving RIT in either direction fromthe baseline value and higher fraction of the partially-premixed charge, is shown inFigure 6.24. As can be seen in this figure, despite the same NG injection durationsand pressures and similar engine loads, manipulation of the RIT towards a morepartially-premixed charge, delays the onset of soot formation processes (and peakHRR as described above), which in turn leads to a lower peak KL factor. As dis-152Figure 6.24: The effect of relative timing between onset of soot signal andcGSOI (τsoot = Θsoot − cGSOI): (a) τsoot comparison for all operatingpoints in Table 6.1. “BL” and “Dis” labels denote baseline and dieseloperating points, respectively; (b) lower KL factor peak at longer τsootas a result of more partial premixing with RIT sweep and increasingKL factor peak at consistent τsoot for GPW sweep.cussed before, modifications in duration of NG injection changes peak KL, withoutaffecting the onset of soot signal, and hence τsoot . However, it is evident from Fig-ure 6.24-a that discussed correlation does not hold for the Pin j sweep points, due tothe changes in mixing rates as a result of concurrent changes in injection timings,durations, and mass and momentum flow rates. At Pin j=220 bar, despite the similarτsoot to the baseline, a higher KL peak is obtained with the higher mass flow rateand at Pin j=220 bar the combination of poor mixing and low rates of heat release,153and earlier cGSOI led to a high KL peak with an extended τsoot relative to baseline.As expected, the τsoot in diesel operating condition is much shorter (Figure 6.24-a).This is discussed in more detail later in this chapter.In comparison with RIT=2 CAD, however, the NG injection profile for RIT=14CAD is closer to the peak HRR, which as described above, could explain the simi-lar 2D soot cloud evolution, KL factor peaks, and OH* chemiluminescence curves(Figures 6.10 and 6.11). The higher oxidation rates can also be inferred from thearea fraction distributions shown in Figure 6.14, where more high temperature re-gions are evident through the late-cycle stages as the RIT is decreased. In this fig-ure, CHTS diagrams for the baseline and long RIT are only different at the lowertemperatures, given the overall higher soot formation in the baseline.Comparison of the net soot formation and oxidation rates in Figures 6.8, 6.11,and 6.16 supports the HRR-injection pulse timing correlations and extent of partial-premixing effects discussed above. To facilitate the comparison, the mean net pro-duction rates were evaluated within the ∆θ f orm range (i.e.,[dKL/dθ]f orm) for theconsidered parameter sweeps and is shown in Figure 6.25. RIT=2 and 14 CADpoints, similar to the KL factor peaks, both lead to a lower mean net soot formationrate as a result of more partial-premixing. Increasing NG injection duration andinjecting into a successively richer reaction zone consistently increases the net for-mation rate and results in a non-linearly increasing KL factor peak (Figure 6.24-b).Comparison of the net soot formation rates from the Pin j sweep, unlike the nor-malized net soot formation results, shows a higher rate for Pin j=140 and 220 baroperating points.By transitioning to negative RIT values, a different combustion mode, qual-itatively similar to classic dual-fuel combustion, is achieved [126]. Significantlyadvanced NG injection, before injection of the pilot fuel, results in a longer NGignition delay (τign,NG =20 CAD). As a result of the more homogeneous chargedistribution, leaner NG auto-ignition sites, and flame propagation, NG conversionenergy release increases (and decays) at a slower rate, leading to a broadened HRRcurve (extended heat release at a lower rate). The locally leaner NG combustionprocess and delayed energy release and high temperature reactions relative to EOIresult in negligible soot signal throughout the cycle. As a result, the mean soottemperature and KL calculations have significant uncertainty due to the weak py-154Figure 6.25: The influence of considered parameter sweeps on mean net sootformation rates. Lower soot formation rates are obtained as a resultof more premixing at RIT=2 and 14 CAD and increasing GPW consis-tently increases the formation rates.rometry signal and low SNR.Summary of PIDING in-cylinder soot processesBased on the discussed correlations and investigated parameter sweep effects, in-cylinder soot processes during a typical PIDING combustion event can be de-scribed through four stages as follows. The summary of these processes is illus-trated in Figure 6.26 through HRR and mean KL factor results from the baselineand representative KL factor distributions.1. Pilot combustion and NG ignition (no detectable soot): Combustion of asmall mass diesel pilot “puff” jet, sufficiently mixed with air during the pilotignition delay, creates a high reactivity ring approximately half way throughthe combustion chamber, without soot formation to detectable levels. Thisis followed by ignition of the injected NG along the gas jets passing throughthe high reactivity pilot ring. Lower soot production characteristics of theNG as the primary fuel result in negligible soot signal during this stage,below the detection limits. Retarded injection timing of the pilot diesel fueland injecting into a hotter combustion chamber closer to TDC resulted in155detection of soot also during pilot combustion 6.13.2. Partially-premixed NG combustion and soot formation: Some unreactedfraction of the injected NG penetrates past the pilot high reactivity ring andto the bowl wall, where it mixes with air. Ignition of this partially-premixedcharge around the periphery and the corresponding peak HRR coincides withthe onset of soot signal at the points of NG jet impingement on the bowl wall,forming small “pockets” of soot. The higher fuel conversion and oxidationrates within and continued injection of NG into the reaction zone and air dis-placement, as well as accumulation of low concentration soot formed on theNG jets are the contributing factors for the timing and locus of the onset ofsoot signal. Manipulation of relative pilot and NG injection timings towardsa more partially-premixed combustion increased the onset of soot signal de-lay relative to the start of NG injection and led to a lower KL factor peakand less soot cloud growth. The locus of the detected onset of soot movesupstream along the gas jets at lower Pin j, due to the lower jet momentum andpoor mixing.3. Soot cloud growth: With continuation of soot formation mechanisms, ini-tially formed soot pockets grow and merge to form a toroidal soot cloud.The soot cloud toroid grows towards the center of the bowl following moresoot formation and advection from NG jet momentum reflected off the wall.The extent of soot cloud growth towards the center and soot concentrationsclosely depend on the relative timing between the NG injection profile andpartially-premixed HRR peak, as well as Pin j. Further NG injection beyondthe peak HRR and higher Pin j lead to increased peak KL factors and bowlarea coverage. Combination of a lower Pin j and extended NG injection be-yond HRR, however, still leads to high KL factors and soot cloud growth,but with a more heterogeneous KL factor distribution. The normalized netrates of soot formation also increase with Pin j. Numerical simulations byHatzipanagiotou et al. showed high equivalence ratios on the NG jets andsuggested mixture formations resembling that of a classic diesel injectionwith rich zones (λ << 0.5). Nevertheless, during this stage, high temper-ature reactions on quasi-steady lifted jet flames outside the growing soot156cloud indicates low-sooting non-premixed combustion on the NG jets, giventhe lower soot production propensity of NG.4. End of Injection and soot oxidation: After NG EOI and the peak KL fac-tor, with higher oxidation rates more concentrated around the periphery ofthe bowl, soot oxidation becomes the dominating process. As such, the sootcloud starts diminishing around the edges and disappears latest in the centralregion of the bowl. This is followed by the diminishing of the more homoge-neous OH* chemiluminescence signal during the late-cycle oxidation stage.The soot oxidation rates majorly depended on the KL factor peak and dis-tribution. The longest injection duration showed the lowest normalized netsoot formation rates and more soot cloud signal throughout the late-cycle ox-idation stage, followed by Pin j=140 and 220 bar operating points. Increasedmomentum flow rates at Pin j=220 bar did not show signs of enhanced sootoxidation after NG EOI.Figure 6.26: Soot processes during a typical PIDING combustion event. Theillustration is based on the baseline operating condition and imagesshow KL distribution at select crank positions.-157It is worth noting that the above describes the statistically representative in-cylinder soot formation and oxidation processes, based on the ensemble averagedresults. Deviations from these general behaviors in individual cycles, e.g., sporad-ically detected soot signal during pilot combustion or extended signal later in thecycle than described above, has been noted.PIDING vs. Conventional Diesel CombustionThe typical PIDING combustion soot formation and oxidation behavior describedabove is substantially different from that of the conventional diesel combustion,although both ultimately lead to similar efficiencies [112]. Figure 6.27 presents acomparison between the baseline PIDING combustion and examples of diesel com-bustion NL and two-color pyrometry imaging results. The NL images by Singhet al. [136] were captured using a 10-bit CMOS camera with exposures set tohigher intensities of diesel soot particles. Figure 6.27 also illustrates the resultsfrom a low pressure, double pulse (pilot and primary injections, both through thediesel holes), diesel operating condition, which was carried out using the currentHPDI injector (see Table 6.1). The HPDI injector is designed to inject only enoughdiesel to provide pilot ignition and is not meant to be used for conventional diesel-only combustion. Furthermore, the injection pressure is much lower than moderncommon-rail diesel engines, which will impact the results. It is worth noting that,although consistent trends are observed among the cases illustrated in Figure 6.27,the engine specifications and fueling settings are different and the absolute resultsare not directly comparable. Despite the varying operating conditions, the dieselcombustion in all cases similarly show higher temperature sooty mixing-controlledstanding primary diesel flames. Soot formation and oxidation is observed on in-jected diesel jets, showing higher KL levels/incandescence signal with the localizedfuel conversion process, consistent with Dec’s conceptual model [24].When compared to NG ignition in PIDING, primary diesel ignition occurs witha much shorter optical ignition delay and closer to the injector, well within the pilothigh reactivity ring. This is accompanied by much earlier soot detection along thegas jets, with the current imaging system settings. The high temperature reactionzones (OH* signal) and strong soot cloud signal (KL distributions) continue to158Figure 6.27: Baseline PIDING vs. Diesel combustion; 2D OH* chemilumi-nescence and two-color pyrometry results from this work, NL imagingresults from Singh et al. [136], and two-color pyrometry results fromMancaruso et al.[94] and Lee et al. [84]. Numbers at the top-right ofimage frames indicate CAD aTDC.159Figure 6.28: Aggregate resolved temperature and KL factor cumulative dis-tribution for baseline PIDING (top) and conventional diesel (bottom)operating conditions160overlap throughout the entire cycle, remaining localized on the jet axis, consistentwith Dec’s description of OH* radicals creating an envelope around the mixingcontrolled flame. The close matching between OH* signal and soot clouds is stillobserved during late-cycle oxidation process in diesel combustion, concentratedaround the bowl walls. This emphasizes the much more localized phenomena,differing significantly from that of the PIDING combustion, where soot signal isnot detected until the combustion of the partially-premixed charge close to thewall and the peak HRR timing. The processes remain substantially different afterthe detected soot signal in PIDING, as low-sooting non-premixed NG conversionalong the gas jets continues, while a toroidal soot cloud is formed at the wall region.The soot cloud growth stage, however, is in part mixing-rate limited [126], similarto the diesel combustion processes. Finally, a more uniform late-cycle oxidationstage is observed for the PIDING combustion across the bowl area, unlike theoverlapping soot oxidation regions (OH* and KL regions) in diesel combustion.These two fundamentally different engine operating conditions ultimately leadto substantially different soot concentration distributions. This is illustrated in Fig-ure 6.28, which shows T -KL cumulative distribution throughout the cycle from alloptically recorded combustion cycles in each case. The enclosing red lines in eachcase shows the two-color pyrometry detection limits, determined by the dynamicrange of the camera and the calibration range. As discussed in the previous chapter,particular attention was paid to maximizing the detection limit in consideration ofthe expected limited pyrometry signal in PIDING combustion[74]. The imagingsystem in both cases was optimized for the primary combustion stage, specificallyto capture majority of the signal while avoiding saturation and data loss. As such,the high population area for both combustion strategies is located well within thedetection limit. Comparison of the baseline PIDING and diesel combustion tem-perature and KL factor distributions in Figure 6.28 shows significantly less sootformation and slightly lower temperature combustion for PIDING operation strat-egy, relative to the very low load diesel point considered in this work.1616.3 Concluding RemarksOptical investigation of in-cylinder soot processes during the PIDING CI engineoperating strategy was executed for the first time, using simultaneous spatially andtemporally resolved two-color pyrometry and OH* chemiluminescence measure-ments. Typical soot formation and oxidation processes under a standard baselineoperating condition were analyzed. This was followed by parameter sweeps of NGinjection pulse width, relative injection timing between pilot and NG fuels, andfuel injection pressure.Throughout the investigated campaign, consistent correlations were observedbetween the partially-premixed HRR peak and onset of detectable soot signal. PeakKL factor also demonstrated a close correlation with an offset from NG EOI acrossthe campaign, regardless of the commanded start of injection settings and injectiondurations, the consequent degree of partial premixing, or NG jet momentum flowrates. These correlations suggest that the relative timing between the injectionprofile and peak HRR play a significant role in the soot processes during PIDINGcombustion. Extending NG injection beyond the partially-premixed peak HRRand introducing more NG into the reaction zone results in increasingly mixing-ratelimited combustion mode and substantially increases soot formation and peak KLfactor levels. The principal conclusions on the analysis results can be listed asfollows:• No soot signal was detected for a typical PIDING combustion prior to thepeak HRR, due to sufficiently mixed pilot diesel combustion and ignition ofthe lower soot production propensity NG. Once the ignited charge reachesthe bowl wall, NG is converted through both low-sooting non-premixed com-bustion along the gas jets and partially-premixed combustion of the fuelinjected during the NG ignition delay. Coincident with combustion of thepartially-premixed charge, soot was formed in the reaction zone at the pointsof NG jet impingement on the wall. From there, soot pockets then forma toroid and grow towards the center following more soot formation andadvection from reactive NG jets reflected off the wall. With oxidation hap-pening more extensively on the periphery, the soot cloud starts diminishingaround the edges and disappears latest in the central region of the bowl. This162is followed by the diminishing of the more homogeneous OH* chemilumi-nescence signal during the late-cycle oxidation stage.• Increasing NG injection duration resulted in further soot cloud growth withhigher optical thickness and extended soot signal and more significant OH*signal attenuation. Minimal net soot formation during the short GPW operat-ing point showed negligible OH* signal attenuation and strong OH* chemi-luminescence was recorded at the wall region. Although the onset of thedetectable soot signal and peak HRR timing remained the same, the highamount of soot formed within a rich mixture for the long GPW operatingcondition led to a much slower normalized net soot oxidation rate.• Reducing or increasing the relative injection timing value from the baselineresulted in more partial premixing, higher peak HRR and OH* chemilu-minescence signal at the wall region and lower peak KL factor. The shortRIT in particular led to much higher partially-premixed peak HRR and OH*chemiluminescence signal peak, with no sign of attenuation from the lessdeveloped optically thin soot cloud.• Increasing fuel injection pressure introduced a trade off between higher mo-mentum flow rate effects, i.e., enhanced mixing during injection and due tostronger reflected momentum off the walls after impingement, and highermass flow rate effects, i.e., richer NG jets and higher local soot formationrates. As such, a higher Pin j increased the rate of reactions and soot cloudgrowth, peak HRR, and also the normalized net soot formation rate, leadingto a slightly higher KL factor peak, as a result of enhanced mixing, but highermass flow rate. With the lower injection pressure, the lower momentum flowrate and mixing leads to earlier onset of detectable soot signal on the NGjet axes before they reach the bowl wall, followed by more soot formationwithin the rich mixture with the extended NG injection duration. These ef-fects combine to produce a much higher KL peak in the soot cloud that isformed at a lower normalized net formation rate and remains more concen-trated at the wall region due to the weaker advection forces in reflection afterjet impingement.163The presented optical investigation of PIDING combustion in this chapter pro-vided insight into the typical soot formation and oxidation modes, and the effectof GPW, Pin j, and RIT control parameters on these processes. Most significantly,as described by the effects of the relative timing between NG injection profile andpeak HRR, these control parameters alter the fraction of partially-premixed versusmixing-controlled energy release. This in turn, determines the rate and extent ofsoot formation and oxidation and, ultimately, is expected to affect engine out soot,through manipulation of injected fuel mass, momentum flow rate, and residencetime.This survey, however, was conducted using a limited number of PIDING op-erating space points, within inherent restrictions of an optical engine, with a lowercompression ratio, and line-of-sight two-color pyrometry measurements. Designedbased on the current findings, a more comprehensive investigation of the afore-mentioned control parameter effects in a thermodynamic engine is introduced anddiscussed in the next chapter.164Chapter 7Optical Probe SignalCharacterizationThe study of in-cylinder processes through spatially resolved optical measurementsoffers a substantial depth of information on the involved combustion mechanismsand soot formation and oxidation modes, as demonstrated in the previous chapter.However, as discussed in §2.1.1, the optical engine configuration necessities mod-ifications to the combustion chamber geometry and engine operating conditions,which could alter the in-cylinder processes relative to a thermodynamic “all-metal”engine. As a less processes-altering approach, the optical probe design describedin §3.2.5 can be used for optical measurements from the combustion chamber ofan all-metal engine, without imposing optical engine restrictions, albeit withoutproviding spatial attributes of the collected signal.Therefore, in order to extend the understating afforded by the 2D measure-ments, and to study the in-cylinder processes under more realistic conditions ofthe all-metal engine, this chapter focuses on simultaneous 2D and 0D optical mea-surements of the PIDING combustion, performed under the optical engine con-figuration. This is then followed by discussion of 0D two-color pyrometry mea-surements in thermodynamic engine, to analyze the effects of different engine con-figurations. Better characterized optical probe measurements of in-cylinder soottemperature and concentrations, and chemiluminescence, in conjunction with HRRresults, would set the framework to develop a thermo-optical analysis tool-kit, to165study combustion in-cylinder processes under more realistic operating conditions.7.1 2D vs. 0D two-color pyrometry measurementsThe measurement characteristics of the optical probe and 2D measurement systemswere described in Chapter 3. To facilitate characterization of the optical probesignal, the analysis of in-cylinder soot processes based on 0D optical probe mea-surements was considered in correlation with spatially resolved and averaged 2Dtwo-color pyrometry. Figure 7.1 shows the optical engine and measurement systemlayout used for this analysis.Figure 7.1: 2D and 0D Two-color pyrometry measurement system on opticalengine configuration.The two-color pyrometry method described in Chapter 2 is a line-of-sight mea-surement technique. As described by the Radiative Transfer Equation (equation7.1), the incremental changes and ultimately recorded spectral radiant intensity sig-nal on any directional pathway (line-of-sight) is the aggregate result of the combi-nation of light emission from hot soot parcels at each point, and partial attenuationof the light from previous points on the pathway through absorption and scatter-166ing, and light scattered into the considered pathway from any incoming directionalpathways.dIλ (S, t)dS= κλ Iλ ,b(S, t)−κλ Iλ (S, t)−σS,λ Iλ (S, t)+14pi∫Γi=4piσS,λ Iλ (S, t)Φλ (Γi)dΓi(7.1)where S is the line-of-sight path, κλ is the absorption coefficient (see §2.1.4),σS,λ is the scattering coefficient, and Φλ is the scattering phase function, and Γis the solid angle. The RTE is the conservation of radiative energy through anabsorbing, emitting, and scattering medium along a path (S) for a given wavelength.The scattering effects are assumed negligible for the size of particles pertinent totwo-color pyrometry measurements of soot clouds [58] (3rd and 4th terms on right-hand-side of equation 7.1). The light emission and absorption effects are illustratedin Figure 7.2.Figure 7.2: Line-of-sight measurement effects in a partially transparent sootcloud. The incandescent light emission and illuminating light rays fromand to individual particles occur at all directions and here shown onlyon the optical path for simplicity.Therefore, changing the point of view to the soot cloud in the combustionchamber might affect the detected light intensity (i.e., two-color pyrometry rawsignal), due to the asymmetries in the soot particle distributions in the soot cloud.167In addition to the described point of view effects, the field-of-view (FOV) inthe considered optical measurement systems in this analysis, as illustrated in Fig-ure 7.1, also have significantly different geometries. The cylindrical FOV of the2D measurement system collects an array of (approximately1) parallel directionalspectral radiant intensities on the CMOS sensor of the high-speed camera. Theoptical probe, conversely, collects the incoming light within its conical FOV, usingit sapphire rod lens, onto a small surface area of an optical fiber, which leads to asingle value spectral radiant intensity reading.In what follows, the influence of this inherently different measurement charac-teristic is examined and an approach to reflect it into comparison of the results isintroduced. The introduced method is then evaluated through analysis of a rangeof PIDING combustion operating conditions.7.1.1 The FOV effectsThe differences in measurement principles of the optical probe and 2D measure-ment systems can render the direct comparison of the results more challenging.Figure 7.3 compares probe 0D two-color pyrometry results and the mean tempera-ture and KL factor values over the bowl area from the 2D measurements (describedby equation 6.3). Here, the baseline and long GPW PIDING operating conditionsare shown (see Table 6.1), as is a double NG pulse PIDING operating condition,with high amount of soot production (here referred to as the long split GPW operat-ing condition). Table 7.1 presents the specifications for these operating conditions.The long split GPW operating point is not to be mistaken for the late post injection(LPI) strategy, which uses a smaller mass second injection of NG later in the cycle,calibrated (relative timing and duration) to enhance late cycle soot oxidation andreduce engine-out soot [39]. The dwell between EOI of the 1st and the GSOI of the2nd NG injection result in significant soot and ample two-color pyrometry signal,which makes it suitable for the purpose of the current analysis.Comparison of the results in Figure 7.3, in general, shows a close estimationof the temperature and KL factor values and trends, considering the differences inFOV geometries and dimensional differences. In both set of measurements, the1In the case of two-color pyrometry measurements via image doubler the collected columns oflight make a ~3◦ angle168Table 7.1: Operating points discussed in Figure 7.3; RIT = 8 CAD, Pin j=180 bar, speed=1000 rpm for all points.Point cPSOI cGSOI1 cGSOI2 PPW GPW1 GPW2 GIMEP[◦ aTDC] [◦ aTDC] [◦ aTDC] [ms] [ms] [ms] [bar]Baseline -18 -10 - 0.7 1.45 - 11.5Long GPW -18 -10 - 0.7 1.85 - 14.2Long splitGPW-20 -12 3 0.7 1.25 0.85 14.1Figure 7.3: 0D probe vs. mean 2D two-color pyrometry results (equation 6.3)for the operating points listed in Table 7.1.two KL factor peaks at long split GPW, resulting from the two closely coupled NGinjection pulses [39], are resolved. The 0D and mean 2D measurements both cap-tured the higher KL factor peaks for the long split GPW and long GPW operatingconditions, and the value of the peaks in all the cases are in close agreement be-tween the two set of measurements. Furthermore, the crank positions of the KLfactor peaks and the rising edge of the curves, between the optical and 2D mea-surements, are relatively close (<5 CAD offset).The 0D probe results in Figure 7.3 indicate more significant cycle-to-cyclevariability, as marked by the shaded regions, especially for the baseline and longGPW cases. This is explained by the more localized measurements within theprobe FOV, compared to the entire bowl area available for spatial averaging in themean 2D results, and is relevant for heterogeneous combustion processes under169consideration.Despite the general agreement between the 2D and 0D results, consistent dis-crepancies are noticed in Figure 7.3. The mean temperature estimates from the 2Dmeasurements are up to 50 K lower than those from the 0D probe measurements,before the KL factor peaks. Furthermore, the rising edge of KL curves and thepeaks are delayed in 0D results and more fluctuations are observed in the tail-endof those results. The 2nd NG injection in long split GPW is not as clearly capturedin mean 2D temperature results and there is a rather consistent decay in calculatedtemperature results from the optical probe measurements.7.1.2 Weighted function averagingThe observed discrepancies described above, are in part an artifact of the morelocalized nature of the optical probe measurements. Accounting for the FOV dif-ferences will allow more direct comparison of the 2D and 0D results. While, this isnot completely achievable solely based on the available optical results, an approachis proposed towards this goal and to provide a closer comparison between the twomethods.The approach is based on generation of a dynamic weighted averaging functionfor the 2D results, as illustrated in 7.4, knowing the relative distances and cosineangles between the probe sapphire lens and the window normal vectors. The opticalprobe FOV (F(u)) is modelled using a general 3D right circular cone equation:F(u) = (u.d)2− (d .d)(u.u)cos2 θ (7.2)Where θ is half of the aperture angle (25◦), u = (x,y,z) is the position vector,apex is at the origin, and axis is parallel to the unit vector d = (dx,dy,dz). Ateach crank position, the intersection of the window surface and the probe FOVforms an ellipse whose boundary is described by equation 7.3, where hw f denotesthe instantaneous distance between window surface and the firedeck. As such, aregion-of-interest (ROI) mask is created at each crank position, which focuses theaveraging on the instantaneous projection of the conical FOV of the probe ontothe quartz window. At each pixel coordinate within the ROI mask, the fraction ofthe column height above it (window to firedeck) that lies within the angled conical170FOV is used as a weight factor for that pixel coordinate, at that crank position(Figure 7.4 and equation 7.4). Repeating the described procedure for every pixelwithin the ROI mask, results in the final dynamic weighted averaging functionmatrix. The weighted spatially averaged (WSA) 2D results as such, could eliminatesome of the discussed localized measurement effects.Ayy2+2dy(hw f dz+dxx)y+Azh2w f +2hw f dzdxx+Axx2A(x,y,z) = (d2(x,y,z)− cos2 θ)(7.3)WSA(x,y,θ) =hFOV (x,y,θ)hw f (θ)(7.4)Figure 7.4: Generation of the weighted spatially averaged ROI for 2D two-color pyrometry results based and projection of optical probe conicalFOV onto the quartz window surface. The fraction of column withinthe inclined probe FOV at each coordinate is used as the weight factorfor the ROI mask. The ROI mask geometry changes corresponding tooptical piston movement within the cycle.171Figure 7.5: WSA ROI mask at three crank positions and the correspondingoptical probe cone projection cross-sections. The surrounding ring onthe WSA cones depicts the intersection with the quartz window surfaceat each crank position.Figure 7.5 shows snapshots of the dynamic weighted averaging function atsample crank positions. The WSA 2D results, however, rely on a rather homo-geneous angular distribution of the soot cloud signal, at least within the FOV ofthe two measurement systems, in order to provide a good agreement with the 0Dprobe measurements. This premise is not necessarily always true, especially undercertain engine operating conditions which lead to high heterogeneity in the results,as will be discussed later in this chapter.With the optical test repeatability measures put in place (see §3.3.2), the stand-off distance between the 2D imaging systems and the firedeck is constant amongdifferent measurements. However, there is slight variations in the injector coordi-nates on the image sensor, which is recorded during two-color pyrometry calibra-tion. Therefore, once the dynamic weighted averaging function matrix was gener-ated for the baseline operating condition, translation of the matrix by the center-to-center distance vector (injector position between the baseline and operating pointunder consideration) makes it applicable to the new 2D results.172The order in which the image processing stages, from the individual cycles ofthe raw image sequences to the WSA curves, is implemented can also affect the re-sults. The major stages involved are the ensemble averaging of the recorded cycles,the two-color pyrometry algorithm (see Chapter 5), and weighted spatially averag-ing procedure described here. Different permutations of these stages are discussedin more detail in Appendix G. Based on that analysis, this chapter presents WSAresults from two different approaches: 1) Application of the WSA to the pyrom-etry results, and 2) Ensemble average of the single cycle WSA KL and T results.The former approach is spatially averaging of the pyrometry results as presented inChapter 6 and here is referred to as the WSASNR, while the latter provides a con-ventional ensemble average for comparison with 0D probe results and is referredto as the WSAEns.The WSASNR and WSAEns approaches are implemented on the operating con-ditions listed in Table 7.1 and the results are compared to their 0D probe coun-terparts in Figure 7.6. The WSA results show better agreement with the probetemperature measurements, compared to the mean 2D results. In all cases, theWSA temperature trends closely follow 0D results during the soot formation stage(before peak KL) until shortly after the peak KL factor. Also similar to the proberesults, the 2nd NG injection effects in long split GPW are evident in WSA results(KL and T in 7-22 CAD aTDC range). Later in the cycle and during the soot oxi-dation stage, the WSA results stabilize and deviate from those of the optical probe,which continue to decay. In addition, WSA KL factor results more closely followthe 0D results, especially during the soot oxidation stage (after KL factor peak) forthe GPW=1.85 ms and long split GPW operating points. The onset of soot signaland soot formation rates, identified by the rising edge of the KL factor results, moreclosely match in phasing between the WSA and probe results (<1.5 CAD).The comparison of the WSASNR and WSAEns results in Figure 7.6 suggeststhat the ensemble averaging of the raw signal for the WSASNR approach leads tolower KL factor values, relative to the WSAEns approach.Thus, the observed improvements in correlations between the WSA and 0D re-sults demonstrates the influence of accounting for FOV geometry differences. Therising edge of the KL factor and detected onset of soot is delayed for 0D probedue to delay in the soot cloud growth until it reaches the probe FOV. This can be173Figure 7.6: 0D probe vs. WSA 2D two-color pyrometry results for the oper-ating points listed in Table 7.1.inferred from the comparison of the overlay of weighted mask on KL factor distri-butions for the baseline condition in Figure 7.7-a with KL curves in Figure 7.6 (5.5-6.5 aTDC). In addition, the positioning of the probe closer to the wall region, wherehigher temperature reactions from combustion of the partially-premixed charge oc-cur and remain concentrated until late in the cycle (see §6.2.5), explains the closertemperature estimations with the WSA results (OH* results in Figure 7.7-a; notethat the ROI is more heavily weighted here).Similarly, crank positions after the KL factor peak in the GPW=1.85 ms oper-ating condition reveals that with transition of the higher optical thickness regiontowards the center (reflected momentum effects, see §6.2.5), the mean 2D resultsshow a more rapid decay in the KL signal, while the 0D and WSA results capturemore soot within the “FOV” (Figure 7.7-b and middle column in Figure 7.6).Soot formation in the high temperature reaction zone during the mixing-ratelimited fuel conversion processes after the peak HRR [128] and formation of higherKL factor, lower temperature soot pockets, results in a decay in estimated tem-perature values starting from onset of soot signal (coinciding the peak HRR) un-til shortly after the peak KL factor. This phenomenon is illustrated more clearlyin comparison of the T -KL cumulative distribution diagrams2 in the range of 5-10 CAD aTDC for the baseline operating condition, shown in Figure 7.8. The high2See §6.2.1 for details regarding the metric174Figure 7.7: Probe FOV geometry effect on WSA results: (a) 2D OH* chemi-luminescence and KL factor results for the baseline, (b) 2D KL factorresults for the GPW = 1.85 ms operating point.temperature region (yellow color denotes highest counts) moves from 2300-2500K at 5 CAD aTDC to 2100-2300 K at 8 CAD aTDC with notable population ataround 2000-2200 K. Consequently, the spatially averaged temperature values willdecrease between 5 to 8 CAD aTDC. This, however, does not explain the discrep-ancy between temperature estimates from WSA and 0D results later in the cycle,which is investigated later in this section.As discussed above, the WSA results explains the 0D pyrometry temperaturedecay before and shortly after the KL factor. In order to investigate the continuedtemperature decay in probe results and the discrepancy between 2D and 0D tem-perature estimations later in the cycle, the apparent temperature data for the twomeasurement techniques and the ideal gas temperature histories are considered.For the WSA results, the weighted function was applied to each pair of instan-taneous apparent temperature images from individual cycles and the results wereensemble averaged for each wavelength (Ta,WSA at 700 and 800 nm). The results175Figure 7.8: Two-color pyrometry temperature and KL factor populationmaps for the baseline operating condition at 5 and 8 CAD aTDC.are shown in Figure 7.9.Comparison of the apparent temperatures in Figure 7.9-a, similarly to the cal-culated temperature results, shows a departure in the WSA and 0D results later inthe cycle. The timing at which this deviation in temperature estimations occursappears close to the peak of the ideal gas temperature estimation, especially forthe baseline and long GPW operating conditions. Therefore, it can be inferred thatthe continuous decay in probe temperature results, in part, originates from the suf-ficiently optically thick soot cloud moving away and out of the probe FOV, withdownward piston motion, and is not entirely an artifact of the solution algorithmand low SNR later in the cycle. The low SNR however, ultimately causes thehigher uncertainty temperature and KL calculations later in the cycle (after 22 and45 CAD aTDC for baseline and GPW=1.85 ms cases, respectively, in Figure 7.9,with widening error-band probe apparent temperatures). Although the hypothe-sized cloud movement away from the firedeck cannot be illustrated via the avail-able imaging system, the diminishing probe FOV coverage in Figure 7.9-b sup-ports the loss of signal in the conical FOV. The 0D two-color pyrometry KL factorcalculations of PIDING combustion by Hatzipanagiotou et al.[54], under similaroperating conditions, also shows the similar late-cycle KL behavior, i.e., decreas-ing oxidation slope with higher fluctuations in the results (peremptory temperatureresults was not reported). Similar temperature decay behavior was observed for176Figure 7.9: (a) 0D probe vs. WSA 2D two-color pyrometry results, and theircorresponding apparent temperatures (Ta), for the operating points listedin Table 7.1, (b) 2D KL factor results vs. probe FOV coverage duringsoot oxidation stage in the baseline operating point.alternative weighted spatially averaging approaches later during the low-SNR late-cycle soot oxidation stage (WEP and WPE results in Appendix G).7.2 Thermodynamic vs. optical engine configurationsFollowing the characterization measures described in the previous section, the op-tical probe was used for 0D two-color pyrometry measurements of PIDING com-bustion in thermodynamic engine configuration. The primary goal of these mea-surements was to compare the in-cylinder processes in the thermodynamic enginewith the toroidal bowl design (as opposed to the transparent flat cylindrical bowlof the optical engine) and without constraints of optical measurements (e.g., skip-firing and load limits), as inferred from the optical measurements. In the process,the possibility of extending the spatially resolved measurements findings to theprocesses occurring in a thermodynamic engine was assessed. Table 7.2 providesthe specifications for the operating conditions considered for this investigation.177Table 7.2: Operating conditions investigated under thermodynamic andoptical engine configurations using 0D optical probe measurements;Engine speed was 1000 rpm for all points.Point Configuration cPSOI cGSOI PPW GPW GIMEP CA50[◦ aTDC] [◦ aTDC] [ms] [ms] [bar] [◦ aTDC]Baseline*Thermo. -16 -10 0.55 1.45 11.8 6.7Optical -18 -10 0.7 1.45 10.8 7.9Long RITThermo. -22 -10 0.55 1.55 12.3 6.6Optical -24 -10 0.7 1.45 11.3 7.4Pin j=220 barThermo. -15 -8.5 0.45 1.25 13.1 6.8Optical -16 -8 0.6 1.2 12.8 9.0Pin j=140 barThermo. -21 -13.5 0.7 2.2 12.3 7.2Optical -22 -14 1.0 2.1 11.2 7.9* Pin j=180 bar; RIT= 8◦It should be noted that due to technical difficulties in the available engine facil-ity, performing optimized replicas (based on CA50, load, HRR shape) of the opticalcampaign operating points discussed in the previous chapter (see Table 6.1), underthe thermodynamic engine configuration was not possible in the time frame of thiswork. Nevertheless, the presented available pairs of operating conditions in Table7.2 suggest relatively similar in-cylinder phenomena, inferred from general HRRshapes, engine loads and combustion phasing, and therefore, still could be usedto compare the 0D two-color pyrometry results under the two engine configura-tions. Figure 7.10 illustrates this comparison for the baseline operating condition.The HRR and two-color pyrometry results for the thermodynamic measurements inthis and all subsequent figures are the ensemble average of 110 cycles, and the tem-peratures are plotted only in the range where the cycle-to-cycle standard deviationis less than 150 K (for both engine configuration results).Here, with the toroidal bowl design and higher compression ratio in the ther-modynamic configuration, a lower HRR peak is observed for the thermodynamicconfiguration, which suggests less partially premixing at the bowl wall (shorter ig-nition delay and lower injection rate). The combustion chamber reaches a higher178Figure 7.10: 0D probe two-color pyrometry results (top), and HRR results(bottom) for the baseline operating condition under optical and ther-modynamic engine configurations.temperature in the thermodynamic engine, due to the higher compression ratio aswell as continuous firing (no skip-firing cycles) and different heat transfer charac-teristics of the two engine configurations (further discussed bellow). This results inhigher pyrometry temperatures and advanced NG ignition (shorter τNG), inferredfrom the advanced rising edge of HRR at ~0 CAD aTDC and advanced peak. Anearlier rise in the KL curve (and first appearance of temperature signal) is observedas a result. The strong correlation between the onset of KL and peak HRR tim-ing was shown in Figure 6.19. A higher fraction of the NG mass being convertedthrough a more mixing-rate limited process leads to a slightly higher peak KL, aswell as later in the cycle (figure 7.10, 15-25 CAD aTDC). However, the higher py-179rometry temperature within the same range suggests higher soot oxidation rates,which eventually leads to similarly low KL factor values after 25 CAD aTDC. Sim-ilar decay in the pyrometry temperature values is observed for both engine configu-rations, which suggests similar contributing factors, explained in §7.1.2, affect thesoot cloud temperature in the combustion chamber of the thermodynamic engine.In addition to the higher compression ratio and continuous firing, lower radia-tion heat losses in the all-metal engine in the absence of the quartz window, couldbe a contributing factor for the higher estimated temperatures. A simplified radi-ation heat loss model for comparison of the two engine configuration suggested a45% lower heat loss in the all-metal engine. Details of this analysis is presentedin Appendix H. However, the quartz window also leads to lower conductive heatlosses compared to the metal piston crown. Investigation of the effect of the lowerconductivity of the optical window is reported to significantly affect the heat re-lease rate characteristics [5]. Considering these opposing effects and limited avail-able information in this regard, it would be difficult to confidently comment on theeffect of total heat transfer loss through the quartz window.It should also be noted that light reflection off the metal surfaces, higher inthe absence of the transparent quartz window in the all-metal engine, might beconsidered as a contributing factor to a higher pyrometry temperature calculation.However, the wall reflection was shown to be more significant when two-colorpyrometry measurement wavelengths in the IR range is utilized [96]. Under lowsoot concentration diesel operating conditions, that investigation showed temper-ature and KL errors of less than 1 and 10%, respectively, for visible wavelengths,whereas errors of less than 8 and 50% in temperature and KL estimates were ob-served for IR measurements.Comparison of other operating points in Table 7.2 is presented in Figure 7.11and shows similar behaviors in the HRR and two-color pyrometry results. In allcases, the combustion phasing and peak HRR crank positions are advanced inthermodynamic measurements, despite the similar NG cGSOI, which as explainedabove leads to an earlier onset of soot signal, suggested by the advanced pyrome-try KL factor and temperature curves. The pyrometry temperature values are alsohigher under thermodynamic configuration and show similar decay to those of theoptical engine.180Figure 7.11: 0D probe two-color pyrometry results (top row), and HRR re-sults (bottom row) for: (a) Pin j=220 bar, (b) Pin j=140 bar, and (c) longRIT operating conditions under optical and thermodynamic engineconfigurations described in Table 7.2.The pilot combustion HRR results in Figure 7.11 are advanced for the ther-modynamic engine configuration, due to the described combustion chamber tem-perature effects in combination with more advanced cPSOI (Pin j=140 bar and longRIT cases) or higher mass and momentum flow rate (in the case of Pin j=220 bar),relative to the baseline. It is worth noting that the injected pilot mass might also belower in the thermodynamic engine due to higher compression ratio. At Pin j=220 bar,the more significant pilot combustion HRR peak leads to small traces of soot for-mation detected in the ensemble average KL factor results. However, the signalwas still too weak for a reliable temperature estimation within the selected un-certainly limit (standard deviation<150 K). In all three cases in Figure 7.11 thepartially-premixed peak HRR is lower in the thermodynamic engine, despite thehigher engine loads. Similar to the baseline, the higher fraction of NG conversionthrough the more mixing-rate limited processes leads to a higher peak KL factorfor Pin j=220 bar and long RIT operating conditions. However, the Pin j=140 baroperating point, which showed significant heterogeneity in soot concentration dis-tributions in the 2D optical pyrometry results, shows a lower KL factor peak and181more extended signal under thermodynamic configuration. This is attributed tothe toroidal bowl design, which prevents accumulation of rich regions at the wallregion through better guiding of the lower momentum NG jets and enhanced mix-ing after impingement on the walls. Nevertheless, the higher temperature suggestshigher late-cycle soot oxidation rates, which leads to similar KL factor values after40 CAD aTDC.Overall, a good agreement between the thermodynamic and optical results, interms of describing the influence of fueling parameters on soot trends, is observedacross the operating points in Table 7.2. NG EOI prior to the peak HRR at longRIT shows a predominantly partially-premixed fuel conversion and low soot signalwithin a limited crank position range; extended injection of NG beyond the peakHRR, with a lower momentum flow rate, leads to higher KL factor peak relative tobaseline and a more heterogeneous soot cloud distribution, inferred from the shapeof the KL factor curves; and at higher Pin j (220 bar), higher mass flow rate NGjets and soot formation rates lead to a higher KL factor peak relative to baseline.Similar to the Pin j sweep analysis in the optical engine configuration (Chapter 6),Pin j=220 bar here also leads to a higher engine load. Therefore, direct comparisonof in-cylinder soot processes to describe the previously observed lower engine-outsoot emissions at increased Pin j, where the load was held constant, [34, 98] was notpossible.7.3 Concluding Remarks0D optical probe two-color pyrometry measurements under optical and thermody-namic engine configurations were considered in this chapter. A simplified weightedspatially averaging (WSA) technique for 2D measurements was introduced to ac-count for FOV differences during optical engine measurements. WSA results im-proved the agreement between 2D and 0D results and explained the discrepanciesin the two signals originating from the more localized nature of the probe light col-lection technique. As such, more insight into temperature and KL estimates fromthe optical probe was provided. The higher pyrometry temperature estimates fromoptical probe were attributed to more heavily weighted light collection from re-gions closer to the walls, where higher concentration of high temperature reactions182were previously observed. Similarly, delayed soot signal detection was explainedby the soot cloud formation and development until it reaches the optical probeFOV. An initial temperature drop was explained by WSA results and correspondedto continued NG injection and formation of higher optical thickness soot cloud andlower heat release during more mixing-rate limited fuel conversion. However, theconsistent temperature decay in probe results later in the cycle could not be repro-duced by the proposed WSA technique and was in part attributed to the soot signalleaving the probe FOV with higher soot oxidation rates in those regions.0D optical probe pyrometry measurements from optical and thermodynamicengines showed good agreement in the results, considering the inherent differencesbetween the two engine configuration designs and operating procedures. This af-forded interpretation of the thermodynamic engine in-cylinder soot processes ina similar fashion to the optical engine configuration. However, the higher com-pression ratios and continuous firing under thermodynamic engine configuration,consistently resulted in shorter ignition delays, higher pyrometry temperature es-timations and advanced the onset of soot processes, and higher engine load withgenerally a lower partially-premixed HRR peak, followed by slightly higher peakKL factor. Nevertheless, the higher temperature values later in the cycle and thetoroidal bowl geometry appear to compensate for the higher peak and result insimilar KL factor values during the late-cycle oxidation stage.183Chapter 8Conclusions and Future WorkThe presented work focused on developing optical characterization techniques tostudy in-cylinder processes in cleaner combustion strategies, leading to relativelylower yet non-negligible soot formation, such as those involving natural gas. Specif-ically, conventional diesel-ignited dual-fuel (DIDF) combustion of NG and pilot-ignited direct-injected natural gas (PIDING) combustion strategies were opticallyinvestigated through natural luminosity, two-color pyrometry, and OH* chemilu-minescence measurements. In this final chapter the major findings from variousinvestigations carried out in this work are presented and some prospects of futureinvestigations in continuation are presented.8.1 Summary of the Significant FindingsOptical investigations pursued in this work were primarily executed in an ex-perimental facility that supported spatially resolved optical measurements via aBowditch piston arrangement in a 2-litre, single-cylinder research engine. Thecomplete optical measurement system is illustrated in Figure 8.1, with all opticaldiagnostics utilized in different chapters of this work. The development of the 2Doptical measurement systems were accomplished in this work and the 0D opticalprobe design was the focus of a previous work [156].In what follows, a summary of the significant results and novel contributionsto the topic from individual chapters of the current work is presented.184Figure 8.1: 2D and 0D natural luminosity/two-color pyrometry and OH*chemiluminescence measurement system on optical engine configura-tion.8.1.1 Optical Investigation of DIDF combustionA simultaneous high-speed natural luminosity (NL) and OH* chemiluminescenceimaging system was developed to assess the influence of fulling control parame-ters on the fuel conversion processes in DIDF combustion. These measurementsrevealed the potential to manipulate combustion of premixed natural gas (ignitionsites locations and reaction zone growth mechanisms) through controlling pilot in-jection pressure Pin j and relative diesel-CH4 ratios, which were noted as the keydual-fuel operating control metrics. At higher Pin j the auto-ignition sites wereconcentrated around the piston bowl periphery and the reaction zone propagatedtowards the center of the bowl. Concentration of high temperature reaction zones185Figure 8.2: Conceptual effect of pilot fuel injection pressure on reaction zonegrowth mechanism (left: high pressure; right: low pressure) on DF com-bustion modearound the edges promotes more effective fuel conversion at these critical, wall-influenced regions. Reducing Pin j caused ignition to initiate in the vicinity of thepilot fuel jet structures and resulted in a more heterogeneous fuel conversion pro-cess with regions of intense natural luminosity, attributed to soot. The concep-tual representation of the described Pin j effects is presented in Figure 8.2, whichis based on OH* images from representative high and low pilot injection pres-sure operating points. An increase in diesel fraction in the charge resulted in amore aggressive combustion event, due to a larger fraction of energy released in apremixed auto-ignition event, coupled with a decrease in the fraction of the com-bustion chamber with significant OH* or NL light emission, indicating incompletefuel conversion in these regions.8.1.2 Two-Color Pyrometry Method ImprovementThe two-color pyrometry technique, being an unstimulated method relying on nat-ural light emissions from target species, is susceptible to low signal-to-noise ratioeffects imposed by the lower soot emissions of the cleaner combustion strategies.As such, the performance of the pyrometric method was enhanced to facilitatequantitative soot analysis in the low soot signal conditions, such as those of theDIDF and PIDING combustion strategies. This was achieved through modifica-tions to the numerical algorithm, increasing the detection envelope during mea-surements, performing perspective adjustments, accounting for non-uniform and186Figure 8.3: Two-color pyrometry method enhanced performance depicted asimprovements in the resolved signal fraction throughout the ensembleaveraged cycle (left) and at a sample crank position for PIDING com-bustion (right)non-linear system response, as well as localized signal-to-noise ratio enhancementthrough image filtering. Figure 8.3 compares the resolved signal fraction fromsample diesel and PIDING combustion operating conditions and sample PIDINGpyrometric results with and without the proposed enhancements. The improvedpyrometric method provided a nearly 40% increase in the resolved signal fraction.The achieved improvements are of particular significance for the PIDING combus-tion strategy, which had a 75% lower signal intensity than the considered dieselcombustion. Furthermore, the improved method affords pyrometric investigationof other low soot signal strategies, such as LTC, as well as late-cycle soot oxidationprocesses in various combustion strategies, with proper adjustments to the detec-tion envelope. The latter is particularly important for providing the link betweenin-cylinder and engine-out soot trends.1878.1.3 Optical Investigation of PIDING CombustionThe enhanced two-color pyrometry method was implemented simultaneously withhigh-speed OH* chemiluminescence imaging to PIDIING combustion in order tobetter understand in-cylinder soot formation and oxidation processes. The resultsrevealed that a standard PIDING operation can be characterized as low-sooting,non-premixed combustion of the natural gas (NG) along the gas jets and of apartially-premixed reaction zone near the wall. The first detectable soot is observedat the points of NG jet impingement on the wall, in the partially-premixed reactionzone. This results in formation of a soot cloud at the wall region which then growstowards the center due to continued soot formation and reflected momentum of theNG jets impinging on the bowl wall. With higher temperature reactions remainingmore concentrated closer to the wall, the soot cloud signal decay initiates in thoseregions, due to to higher oxidation rates, and disappears latest in the central bowlregions. These primary stages are summarized in Figure 8.4 for a typical PIDINGcombustion event.The relative timing between NG injection pulse and the peak HRR, and Pin j,showed strong influence in the rate and extent of soot formation and peak concen-tration levels. Fuel injection timings that led to an end-of-injection of NG prior tothe peak HRR resulted in predominantly partially-premixed fuel conversion withlower KL and smaller soot cloud, while with increased injection durations continu-ing after the peak HRR, fuel conversion was more mixing-rate limited and resultedin higher soot formation and peak concentration levels and extended growth andlate-cycle signal.8.1.4 Thermo-Optical Analysis Tool-KitWith the developed imaging diagnostics, the final chapter focused on characteriz-ing the signal from optical probe based 0D measurements. The rugged probe de-sign affords optical measurements from all-metal engine configurations and couldbe used to extend the findings of 2D measurements to 0D optical measurementsfrom an all-metal engine under higher loads. These operating conditions would bemore relevant to real-world on-road operating conditions.Comparisons between 2D and 0D two-color pyrometry measurements were188Figure 8.4: The four principal stages of soot processes during a typical PID-ING combustion event: 1-pilot combustion and NG ignition, 2-partially-premixed NG combustion and soot formation, 3-soot cloud growth, and4-oxidation; The illustration is based on the baseline operating condi-tion (see Table 6.1) and images show KL distribution at select crankpositions.made under optical engine configuration in order to better understand and inter-pret the 0D results. A novel weighted spatially averaging (WSA) technique forthe 2D measurements was introduced to account for field-of-view (FOV) geometrydifferences between the two measurement approaches. The 0D and WSA resultsshowed reasonable agreement and explained some of the discrepancies between the0D and mean 2D results. Figure 8.5 illustrates these comparisons for select operat-ing conditions along with the weighted spatially averaging mask for a sample crankposition. 0D pyrometry temperature estimates before peak soot concentration arehigher in response to the light collection more heavily from regions closer to thewalls where higher concentration of high temperature reactions was evident in 2Dmeasurements. The indication of onset of soot processes is also slightly delayed in0D measurements; As illustrated by the 2D results, detectable soot signal is formedat the proximity of the wall and outside probe FOV, and the delay is in response tothe soot cloud growth until signal reaches probe FOV. Furthermore, KL values in189Figure 8.5: WSA mask at 25 CAD aTDC (top) and comparison of the WSA2D and 0D optical measurements for select operating points (bottom,see Table 7.1); The surrounding ring on the WSA cone depicts the in-tersection with the quartz window surface at this crank position.0D measurements are higher after the peak KL in response to transition of the highconcentration soot cloud towards the center due to reflected momentum of NG jetsafter impingement on the walls. The consistent decay in 0D temperatures and devi-ation from WSA results was partially explained by soot cloud movement and sootleaving probe FOV, as well as the decreasing gas temperature due to downwardpiston motion. However, further investigation of the this phenomenon was deemednecessary to provide a more clear understanding.0D pyrometry measurements were carried out under thermodynamic and op-tical engine configurations in order to evaluate the effects of the two inherentlydifferent combustion environments and assess the extensibility of the 2D measure-ments findings to in-cylinder processes of a thermodynamic engine. 0D pyrometrymeasurements of similar operating conditions in the all-metal engine showed simi-lar temperature and soot concentration trends to those from optical engine. As such,190these results led to similar conclusions regarding soot in-cylinder processes, de-spite the influences from geometric and operational differences, namely the highercompression ratio, toroidal bowl geometry, and continuous firing. Higher com-pression ratios and continuous firing led to higher pyrometry temperatures in ther-modynamic engine and an earlier onset of soot processes in response to advancedpeak HRR due to the shorter ignition delays. Slightly lower peak HRR and higherpeak soot concentrations were obtained due to lower partially-premixed fraction,although higher temperatures and toroidal bowl geometry, led to signs of enhancedlate-cycle soot oxidation, which resulted in similar soot concentration signal.8.1.5 Concluding RemarksSummaries of the major results from different chapters of this work was presentedabove. To summarize the work as whole, the principal novel contributions can belisted as follows:1. Enhancement of the two-color pyrometry technique to extend its applicabil-ity for modern cleaner combustion strategies with less naturally emitted PMsignal.2. Present a conceptual understanding of the in-cylinder soot formation and ox-idation processes during a standard PIDING combustion event and elucidatethe influence of fueling control parameters on PIDING in-cylinder processes.3. Characterization of 0D signal from an optical probe design, in accordancewith the detailed information provided in a 2D optical measurements, toextend optical results interpretations to more-representative higher load op-erating conditions in all-metal engines.With the provided insight into in-cylinder fuel conversion and soot formationand oxidation processes and better characterization of the 0D measurements, thecurrent study sets the groundwork for extensive investigation of combustion in-cylinder processes under more realistic higher load operating conditions of DIDFand PIDING combustion strategies. The analysis methods and introduced metrics,however, are not restricted to NG combustion and can be implemented in assess-ment of other combustion scenarios.191Further characterization of the optical probe signal could be possible throughadditional investigations and analyses discussed in the next section, which wouldcomplete the set of tools for characterization of in-cylinder processes relevant tocleaner combustion strategies. The characterized optical probe measurements, inconjunction with HRR results and engine boundary conditions and exhaust emis-sion measurements, would constitute a valuable and effective “Thermo-Optical”analysis toolkit to provide more insight into combustion in-cylinder processes andoptimize the operating protocols and future designs of the combustion strategy un-der investigation. As such, the modern cleaner CI combustion strategies couldconform to ever-increasing stringent emission regulations, while maintaining highperformance and efficiencies resembling those of their diesel engine ancestors.8.2 Future WorkIn addition to the provided insight into in-cylinder combustion phenomena, theinvestigation results presented in this work also identified several areas potentiallyworthwhile to further investigate.8.2.1 2D optical DIDF combustion analysisOptical investigation of the fueling control parameter effects in DIDF combustiondemonstrated potential to significantly affect fuel conversion processes and moti-vated more extensive assessment of such effects on known DIDF combustion draw-backs. Simultaneous 2D OH* chemiluminescence and natural luminosity mea-surements of DIDF combustion identified areas of localized high intensity naturalluminosity without overlapping high intensity OH* chemiluminescence signal, forhigher diesel pilot fraction in the charge or at lower pilot injection pressures. Al-though engine-out soot emissions of DIDF combustion is generally not significant,pyrometric imaging measurements of such operating conditions could provide abetter understanding of the influence of these key fueling control parameters onin-cylinder processes and is not reported in literature.1928.2.2 2D optical PIDING combustion analysisStudy of soot formation and oxidation processes during PIDING combustion overa range of operating conditions identified several characteristic behaviors. Inves-tigation of the NG injection duration effects suggested that a shorter NG injec-tion duration and end-of-injection (EOI) before the peak HRR, leads to a predom-inantly partially-premixed combustion, with a lower peak HRR and soot signallevel, which remains concentrated at wall. Increasing the injection duration, pastthe peak HRR crank position, increased the peak HRR and the normalized netsoot formation rates to a point, beyond which both these metrics remained con-stant. It would be interesting to investigate the hypothesis that a “maximum partial-premixedness” metric at a certain injection pressure and start-of-injection timingscould be considered, after which the same soot formation mechanisms, with thesame normalized rates, are only extended to reach higher concentration peaks.Investigating the fuel injection pressure effects also showed a higher soot con-centration peak at higher injection pressure, which was attributed to higher massflow rate and richer charge mixture around NG jet axes, and traces of enhancedlate-cycle soot oxidation relative to the baseline operating condition was not ob-served. However, the tested higher fuel injection pressure operating condition alsoled to a higher engine load with a slightly different combustion phasing. Mea-surements from more closely relevant “high injection pressure” cases would benecessary to optically investigate the in-cylinder soot processes leading to the pre-viously demonstrated soot emission advantages of operating under higher injectionpressures.As part of the analyses for PIDING soot processes characterization in thiswork, a basic image thresholding approach was implemented for tracking highconcentration soot pockets to illustrate NG jet reflection and soot cloud advectioneffects. A properly developed image feature tracking algorithm, similar to com-bustion image velocimetry [27], could be applied to the provided higher spatialand temporal resolution two-color pyrometry results so as to provide a quantitativemetric for soot cloud dynamics within the combustion chamber.Correlations between temporal evolution of the soot cloud and OH* chemi-luminescence signal depicts OH* signal attenuation by the growing soot cloud.193Knowing the spatial distribution of the soot concentration parameter (KL factor)and making simplifying assumptions to the radiative transfer equation 7.1, the fea-sibility of correcting the OH* signal could be assessed. If possible to achieve, theresult of this analysis would extend the crank position range for which the compar-ison of relative OH* signal magnitudes could be practical.8.2.3 Exhaust Gas recirculation (EGR)A commonly used approach for additional control on pollutant emissions from CIengines is intake charge dilution through implementing EGR [72, 100]. Applica-tion of EGR on PIDING combustion, for instance, reduced the NOx emissions,albeit at the cost of increased hydrocarbons, CO, and PM emissions at higher EGRlevels. Nevertheless, manipulation of fueling control parameters was shown to beable to mitigate these side effects [100]. In addition to engine-out PM, reintro-ducing exhaust gases and oxygen dilution, and the consequent lower in-cylindertemperatures, are expected to significantly affect in-cylinder soot processes. Theengine facility used in the current work, similar to most other optical engines, didnot allow EGR. Further development of the engine facility to implement EGR in thethermodynamic and optical engine configurations would be useful to characterizethe imposed modifications to the in-cylinder processes. Under the optical engineconfiguration, introducing a simulated EGR mixture of CO2 and N2 is a commonapproach to mimic the EGR effects in all-metal engines [54]. However, it has alsobeen shown that the simulated EGR gas composition, in particular fraction of themajor species, i.e., N2, CO2, H2O, CO and uHC, could influence in-cylinder pro-cesses [23]. Considering the complexities associated with imaging diagnostics ofEGR effects, optical probe pyrometric and chemiluminescence measurements (us-ing the third fibre-optic channel) would be particularly valuable to provide moreinsight into the altered in-cylinder processes.8.2.4 Optical probe 0D signal characterizationAs discussed in Chapter 7, several measures can be considered towards an im-proved interpretation of the 0D signal provided by the optical probe. In this work,the two-color pyrometry results obtained from 0D optical probe measurements194were in part characterized through simultaneous 2D and 0D measurements usingthe optical engine configuration and application of a FOV based weighted averag-ing function (WSA). The proposed averaging method, however, poses caveats as itassumes a homogeneous soot cloud. As such, the pyrometric temperature decay in0D probe measurements could not be explained by the proposed approach. Severalinvestigative paths could be considered for further exploration of the 0D opticalprobe signal behavior, such as computational fluid dynamic and Monte Carlo sim-ulations, or adding a third 0D pyrometry signal channel.Reactive computational fluid dynamics (CFD) simulations of the in-cylindercombustion phenomena is a valuable characterization tool which has demonstratedits utility in numerous studies [54, 90, 114]. The estimated three-dimensional dis-tributions of various species and their properties, including PM and temperaturevalues, could be used to fill the gap of the lost information due to the line-of-sight (LOS) nature of the two-color pyrometry measurements. After introducingassumptions for optical properties of the soot particle distribution, the vertical pro-jection of the calculated cell spectral incandescence signals on the quartz windowcould be estimated. Comparison of the projected intensity distribution to those ofthe 2D optical measurements could be used to assess the LOS effects and improvethe correlation between 2D and 0D measurements for better understanding of theoptical probe signal.Alternatively, as a higher fidelity investigation approach, 3D Monte Carlo raytracing simulations based on the reactive CFD results could be considered. To per-form those simulations, instantaneous gaseous flame and PM distributions fromnumerical simulations and modelled geometry of combustion chamber at the cor-responding crank position are used as inputs to the Monte Carlo simulation algo-rithm for the 3D participating medium. Rays of particle emission intensities arethen tracked while undergoing repeated absorption, transmission, reflection, scat-tering and re-emission in interaction with suspended particles and the boundaries.The simulated light collection from modelled “camera views” for the probe con-ical FOV and quartz window would then provide a metric for comparison of therecorded signal through the two measurement systems.From an experimental two-color pyrometry measurement standpoint, 0D op-tical measurements within a narrow spectral pass-band around a third wavelength195could be considered. Introducing the third signal channel could potentially helpfurther characterization of the 0D signal in comparison of the results acquired fromthree separate pairs of spectral radiant intensities. Specifically, it would be interest-ing to assess the feasibility of isolating the effects of signal escaping probe FOV vs.high uncertainly late cycle signal, given different wavelength sensitivities towardslow SNR signal. Once the optical probe signal behavior understanding is improvedthrough these comparisons, the third optical probe channel can be used for 0D OH*or CH* chemiluminescence measurements as an additional investigative metric forin-cylinder processes within thermodynamic engine configuration.In comparison of the 0D two-color pyrometry measurements from thermody-namic and optical engines, certain differences in the results could be attributed tothe more optimized toroidal piston crown geometry in the thermodynamic engine.As an example, a lower soot concentration peak was obtained at a lower fuel in-jection pressures for PIDING combustion, which suggested enhanced mixing withbetter guided NG jets, preventing accumulation of high concentration soot cloud atthe proximity of the bowl walls. This could be addressed through substituting theflat quartz window for a toroidal shaped one. This more representative window ge-ometry introduces complexities for analysis and interpretation of 2D results and ne-cessitates additional image-preprocessing stages, e.g., ray-tracing algorithms [79],to correct for image distortion due to curved window surfaces. Nevertheless, 0Dprobe measurements would not encounter these challenges and the more repre-sentative geometry affords closer comparison of the thermodynamic and opticalengine results.8.2.5 Linking in-cylinder to engine-out soot measurementsThe ultimate goal of the research and development efforts on combustion diagnos-tics is reducing engine-out emissions while maintaining or improving thermal andfuel efficiencies. Application of the developed tools in this work was only focusedon understanding the primary in-cylinder soot formation and oxidation processesof PIDING combustion in an optical engine, as well as characterizing the opticalprobe signal for further investigation of those processes under more realistic op-erating conditions in all-metal engines. Hence, the imaging system settings were196optimized to capture spatially resolved and integrated signals throughout a com-bustion event with high SNR, while avoiding sensor saturation. Given the limiteddynamic range of these sensors, lower incandescence and chemiluminescence sig-nals during the late-cycle oxidation stage was inevitably lost to low SNRs. Dedi-cated optical measurements focusing at late-cycle soot oxidation events would benecessary in order to make the link to engine-out soot emissions.In addition to the conventional, cycle cumulative, engine-out soot measure-ment apparatus, a Fast Exhaust Nephelometer design is being developed within thegroup [73], which affords cycle resolved exhaust stream soot measurements. Cor-relating dedicated pyrometric measurements of the late-cycle in-cylinder events toengine-out soot measurements, on a cycle-by-cycle level, isolating the correspond-ing variability effects, could be substantial in providing insight into soot emissioncharacteristics of different combustion strategies working under various operatingconditions.197Bibliography[1] V. Aesoy and H. Valland. Hot surface assisted compression ignition ofnatural gas in a direct injection diesel engine. SAE, (960767), 1996. →page 7[2] E. E. Agency. Air pollution statistics: Emissions of air pollutants, eu-28,1990-2016, 2017. → page 11[3] P. Aleiferis, Y. Hardalupas, A. Taylor, K. Ishii, and Y. Urata. Flamechemiluminescence studies of cyclic combustion variations and air-to-fuelratio of the reacting mixture in a lean-burn stratified-charge spark-ignitionengine. Combustion and Flame, 136:72–90, 2004. → pages 22, 25[4] C. Arcoumanis, P. Cutter, D. Whitelaw, and B. Flannery. Heat TransferProcesses in Diesel Engines. Chemical Engineering Research and Design,76(2):124–132, 1998. → page 30[5] U. Aronsson, C. Chartier, U. Horn, O. Andersson, B. Johansson, andR. Egnell. Heat release comparison between optical and all-metal hsdidiesel engines. SAE, (2008-01-1062), 2008. → pages 18, 19, 46, 180[6] B. Ayoola, R. Balachandran, J. Frankc, E. Mastorakos, and C. Kaminski.Spatially resolved heat release rate measurements in turbulent premixedflames. Combustion and Flame, 144:1–16, 2006. → pages 22, 23, 24, 25[7] O. Badr, G. Karim, and B. Liu. An examination of the flame spread limitsin a dual fuel engine. Applied Thermal Engineering, 19(10):1071 – 1080,1999. → page 37[8] H. Baker and E. Ryder. Temperature Measurement in Engineering. JohnWiley, New York, 1961. → page 30[9] J. Ballester and T. Garcı´a-Armingol. Diagnostic techniques for themonitoring and control of practical flames. Progress in Energy andCombustion Science, 36(4):375 – 411, 2010. → pages 22, 23, 25198[10] C. R. F. Barbosa, M. R. da Silva, and S. W. Celere. The performance of anotto cycle bi-fuel engine with natural gas direct injection. SAE, (962381),1996. → page 7[11] M. Bell, D. Davis, and T. Fletcher. A retrospective assessment of mortalityfrom the london smog episode of 1952: the role of influenza and pollution.Environmental Health Perspectives, 112:6–8, 2004. → page 8[12] D. Bertsch. Experimentelle Untersuchungen zum Einflußgemischbildungsseitiger Maßnahmen auf den Zu¨ndprozeß, Verbrennungund Schadstoffbildung an einem optisch zuga¨nglichen DE-Dieselmotor.Phd dissertation, University of Karlsruhe, 1999. → page 24[13] N. J. Birger and S. N. Rogak. Flow characteristics of a gas-blast fuelinjector for direct-injection compression-ignition engines. SAE,(2009-01-1857), 2009. → page 7[14] M. K. Bobba, C. L. Genzale, and M. P. B. Musculus. Effect of ignitiondelay on in-cylinder soot characteristics of a heavy duty diesel engineoperating at low temperature conditions. SAE International Journal ofEngines, 2(1):911–924, 2009. → pages 35, 144[15] M. A. Boumehdi, S. A. Stepanyan, P. Desgroux, G. Vanhove, and S. M.Starikovskaia. Ignition of methane- and n-butane-containing mixtures athigh pressures by pulsed nanosecond discharge. Combustion and Flame,162(4):1336 – 1349, 2015. → page 7[16] F. W. Bowditch. A new tool for combustion research a quartz pistonengine. SAE, (610002), 1961. → pages 16, 18[17] R. P. Brent. Algorithms for Function Minimization without Derivatives.Prentice-Hall, Englewood Cliffs, NJ, 1973. → page 95[18] A. P. Carlucci, D. Laforgia, R. Saracino, and G. Toto. Study of combustiondevelopment in methane-diesel dual fuel engines, based on the analysis ofin-cylinder luminance. SAE, (2010-01-1297), 2010. → pages 18, 38, 82[19] F. Catapano, S. Di Iorio, P. Sementa, and B. M. Vaglieco. Characterizationof ch4 and ch4/h2 mixtures combustion in a small displacement opticalengine. SAE International Journal of Fuels and Lubricants, 6(1):24–33,2013. → pages 7, 22, 23, 24, 25[20] Y. Cengel and M. Boles. Thermodynamics: An Engineering Approach.McGraw-Hill, Boston, 6th ed. edition, 2001. → page 5199[21] H. M. Cho and B. Q. He. Combustion and emission characteristics of alean burn natural gas engine. International Journal of AutomotiveTechnology, 9(4):415–422, 2008. → page 7[22] J. Chong, A. Tsolakis, S. Gill, K. Theinnoi, and S. Golunski. Enhancing theno2/nox ratio in compression ignition engines by hydrogen and reformatecombustion, for improved aftertreatment performance. InternationalJournal of Hydrogen Energy, 35(16):8723 – 8732, 2010. → page 9[23] W. F. Colban, D. Kim, P. C. Miles, S. Oh, R. Opat, R. Krieger, D. E. Foster,R. P. Durrett, A. Manuel, and D. Gonzalez. A detailed comparison ofemissions and combustion performance between optical and metalsingle-cylinder diesel engines at low temperature combustion conditions.SAE International Journal of Engines, pages 1–15, 2008. → pages 46, 194[24] J. E. Dec. A Conceptual Model of DI Diesel Combustion Based onLaser-Sheet Imaging. SAE, (970873), 1997. → pages 111, 158[25] J. E. Dec and C. Espey. Chemiluminescence Imaging of Autoignition in aDI Diesel Engine. SAE, (982685), 1998. → pages 24, 118[26] T. J. Dekker. Finding a Zero by Means of Successive Linear Interpolation.In B. Dejon and P. Henrici, editors, Constructive Aspects of theFundamental Theorem of Algebra. Wiley-Interscience, London, 1969. →page 95[27] H. Dembinski. In-cylinder Flow Characterisation of Heavy Duty DieselEngines Using Combustion Image Velocimetry. Phd dissertation, KTHRoyal Institute of Technology, 2013. → pages 19, 193[28] J. M. Densmore, M. M. Biss, K. L. McNesby, and B. E. Homan.High-speed digital color imaging pyrometry. Appl. Opt., 50(17):2659–2665, 2011. → page 102[29] S. Di Iorio, P. Sementa, and B. M. Vaglieco. Experimental Investigation ofa Methane-Gasoline Dual-Fuel Combustion in a Small DisplacementOptical Engine. SAE, (2013-24-0046), 2013. → pages 20, 21, 23, 31[30] S. di Stasio and P. Massoli. Influence of the Soot Property Uncertainties inTemperature and Volume-Fraction Measurements by Two-ColourPyrometry. Measurement Science and Technology, 5(12):1453–1465,1999. → page 31200[31] D. Dockery, C. r. Pope, X. Xu, J. Spengler, J. Ware, M. Fay, B. Ferris, andF. Speizer. An association between air pollution and mortality in six u.s.cities. New England Journal of Medicie, 329(24):1753 – 1759, 1993. →page 10[32] Y. Dong, C. M. Vagelopoulos, G. R. Spedding, and F. N. Egolfopoulos.Measurement of laminar flame speeds through digital particle imagevelocimetry: Mixtures of methane and ethane with hydrogen, oxygen,nitrogen, and helium. Proceedings of the Combustion Institute, 29(2):1419– 1426, 2002. → page 39[33] N. Dronniou, J. Kashdan, B. Lecointe, K. Sauve, and D. Soleri. Opticalinvestigation of dual-fuel cng/diesel combustion strategies to reduce co2emissions. SAE International Journal of Engines, 7(2):873–887, 2014. →pages 38, 39, 40, 44, 74, 80, 90[34] S. Dumitrescu, P. G. Hill, G. Li, and P. Ouellette. Effects of injectionchanges on efficiency and emissions of a diesel engine fueled by directinjection of natural gas. SAE, (2000-01-1805), 2000. → pages41, 42, 43, 139, 149, 152, 182[35] N. Englert. Fine particles and human health—a review of epidemiologicalstudies. Toxicology Letters, 149(1):235 – 242, 2004. Proceedings ofEUROTOX 2003. The XLI European Congress of Toxicology. Science forSafety. → pages 6, 10[36] J. F. Epperson. An Introduction to Numerical Methods and Analysis. JohnWiley & Sons, Inc., Hoboken, New Jersey, 2nd editio edition, 2013. →page 95[37] E. Faghani, P. Kirchen, and S. N. Rogak. Application of Fuel MomentumMeasurement Device for Direct Injection Natural Gas Engines. SAE,(2015-01-0915), 2015. → pages 120, 220[38] E. Faghani, P. Kheirkhah, C. W. Mabson, G. McTaggart-Cowan,P. Kirchen, and S. Rogak. Effect of Injection Strategies on Emissions froma Pilot-Ignited Direct-Injection Natural-Gas Engine- Part II: SlightlyPremixed Combustion. SAE, (2017-01-0763), 2017. → pages 41, 100[39] E. Faghani, P. Kheirkhah, C. W. Mabson, G. McTaggart-Cowan,P. Kirchen, and S. Rogak. Effect of injection strategies on emissions from apilot-ignited direct-injection natural-gas engine- part i: Late post injection.SAE, (2017-01-0774):1–16, 2017. → pages 168, 169201[40] T. H. Fereja, A. Hymete, and T. Gunasekaran. A Recent Review onChemiluminescence Reaction, Principle and Application onPharmaceutical Analysis. ISRN Spectroscopy, pages 1–12, 2013. → page22[41] C. Ferguson, D. Tree, D. DeWitt, and S. Wahiduzzaman. Design,Calibration and Error Analysis of Instruments for Heat TransferMeasurements in Internal Combustion Engines. ASME reprint fromdevelopments in experimental techniques, 71, 1987. → pages 28, 32[42] R.-C. Florea. Optical Ethanol and n-heptane Dual-Fuel Partially PremixedCombustion in a Reciprocating Engine. PhD thesis, Wayne StateUniversity, 2012. → page 40[43] O. for Economic Co-Operation and D. (OECD). Strategies to reducegreenhouse gas emissions from road transport: analytical methods, 2002.→ page 10[44] P. Forster, V. Ramaswamy, P. Artaxo, T. Berntsen, R. Betts, D. Fahey,J. Haywood, J. Lean, D. Lowe, G. Myhre, J. Nganga, R. Prinn, G. Raga,M. Schulz, and R. V. Dorland. Changes in Atmospheric Constituents and inRadiative Forcing. Climate Change 2007, chapter 1.2. CambridgeUniversity Press, 2008. → page 37[45] N. L. Garland and D. R. Crosley. On the collisional quenching ofelectronically excited oh, nh and ch in flames. Symposium (International)on Combustion, 21(1):1693 – 1702, 1988. Twenty-First Symposuim(International on Combustion). → page 24[46] A. Gaydon. The Spectroscopy of Flames. Chapman and Hall, London, 2nded. edition, 1974. → pages 20, 22, 23, 24, 25[47] A. Gaydon and H. Wolfhard. Flames, Their Structure, Radiation andTemperature. Chapman and Hall, London, 1978. → page 22[48] B. Grandin, I. Denbratt, J. Bood, C. Brackmann, and P.-E. Bengtsson. Theeffect of knock on the heat transfer in an si engine: Thermal boundary layerinvestigation using cars temperature measurements and heat fluxmeasurements. SAE, (2000-01-2831), 2000. → page 4[49] B. Group. Statistical review of world energy, 2018. → page 6202[50] S. B. Gupta, B. P. Bihari, M. S. Biruduganti, R. R. Sekar, and J. Zigan. Onuse of co2∗ chemiluminescence for combustion metrics in natural gas firedreciprocating engines. Proceedings of the Combustion Institute, 33(2):3131– 3139, 2011. → pages 3, 22[51] D. Guyot, F. Guethe, B. Schuermans, A. Lacarelle, and C. Paschereit.Ch∗/oh∗ chemiluminescence response of an atmospheric premixed flameunder varying operating conditions. In ASME Turbo Expo 2010: Power forLand, Sea, and Air, Glasgow, UK, 2010. Volume 2: Combustion, Fuels andEmissions, Parts A and B. → page 22[52] Y. Han, K. Lee, and K. Min. A study on the measurement of temperatureand soot in a constant-volume chamber and a visualized diesel engine usingthe two-color method. Journal of Mechanical Science and Technology, 23(11):3114–3123, 2010. → pages 30, 57, 58[53] Y. Hardalupas and M. Orain. Local measurements of the time-dependentheat release rate and equivalence ratio using chemiluminescent emissionfrom a flame. Combustion and Flame, 139:188–207, 2004. → pages22, 23, 25[54] A. Hatzipanagiotou, F. Marko, G. Koenig, C. Krueger, P. Wenzel, andT. Koch. Numerical and optical analysis of heterogeneous gas combustionwith diesel pilot ignition in a commercial vehicle engine. InternationalJournal of Engine Research, 19(1):109–119, 2018. → pages21, 41, 42, 45, 111, 124, 136, 142, 176, 194, 195[55] B. Higgins, M. McQuay, F. Lacas, J. Rolon, N. Darabiha, and S. Candel.Systematic measurements of oh chemiluminescence for fuel-lean,high-pressure, premixed, laminar flames. Fuel, 80(1):67 – 74, 2001. →page 25[56] S. K. Hoekman and C. Robbins. Review of the effects of biodiesel on noxemissions. Fuel Processing Technology, 96:237 – 249, 2012. → pages9, 10[57] H. G. Hottel and F. P. Broughton. Determination of True Temperature andTotal Radiation from Luminous Gas Flames. Ind. Eng. Chem. Anal., 4(2):166–175, 1932. → page 28[58] J. Howell, R. Siegel, and P. Menguc. Thermal Radiation Heat Transfer.CRC Press, Taylor and Francis, Boca Raton, 5th ed. edition, 2010. → page167203[59] Q. Huang, F. Wang, J. Yan, and Y. Chi. Simultaneous estimation of the 3-dsoot temperature and volume fraction distributions in asymmetric flamesusing high-speed stereoscopic images. Appl. Opt., 51(15):2968–2978,2012. → page 102[60] E. Huestis, P. A. Erickson, and M. P. B. Musculus. In-cylinder and exhaustsoot in low-temperature combustion using a wide-range of egr in aheavy-duty diesel engine. SAE, (2007-01-4017), 2007. → page 35[61] Y. Ikeda, J. Kojima, T. Nakajima, F. Akamatsu, and M. Katsuki.Measurement of the local flamefront structure of turbulent premixed flamesby local chemiluminescence. Proceedings of the Combustion Institute, 28(1):343 – 350, 2000. → pages 22, 23[62] T. Ito and T. Kitamura. Effects of Flame Lift-Off and Flame Temperatureon Soot Formation in Oxygenated Fuel Sprays. SAE, (2003-01-0073),2003. → pages 16, 30, 108[63] M. Jakob, T. Hu¨lser, A. Janssen, P. Adomeit, S. Pischinger, andG. Gru¨nefeld. Simultaneous high-speed visualization of soot luminosityand oh∗ chemiluminescence of alternative-fuel combustion in a hsdi dieselengine under realistic operating conditions. Combustion and Flame, 159(7):2516 – 2529, 2012. → page 25[64] M. Jakob, P. Lehnen, P. Adomeit, and S. Pischinger. Development andApplication of a Stereoscopic 3-Color High-Speed Ratio-Pyrometry.Combustion and Flame, 161(11):2825–2841, 2014. → pages30, 57, 58, 102[65] W. Jing, W. Roberts, and T. Fang. Comparison of Soot Formation ForDiesel and Jet-A in a Constant Volume Combustion Chamber Using Two-Color Pyrometry. SAE, (2014-01-1251), 2014. → page 30[66] R. Johnson. The structure and origin of the swan band spectrum of carbon.Philosophical Transactions of the Royal Society A, 226:157–230, 1927. →pages 3, 23[67] T. Kamimoto and Y. Murayama. Re-examination of the emissivity of dieselflames. International Journal of Engine Research, 12(6):580–600, 2011.→ page 28[68] T. Kamimoto, N. Uchida, T. Aizawa, K. Kondo, and T. Kuboyama. Dieselflame imaging and quantitative analysis of in-cylinder soot oxidation.204International Journal of Engine Research, 18(5-6):422–435, 2017. →pages 31, 33, 35, 36[69] G. Karim. Combustion in gas fueled compression: Ignition engines of thedual fuel type. Journal of Engineering for Gas Turbines and Power, 125(3):827–836, 2003. → pages 37, 43, 83, 88, 90[70] G. A. Karim and I. Wierzba. Comparative studies of methane and propaneas fuels for spark ignition and compression ignition engines. SAE,(831196):676–688, 1983. → page 7[71] S. Karnani and D. Dunn-Rankin. Visualizing ch∗ chemiluminescence insooting flames. Combustion and Flame, 160(10):2275 – 2278, 2013. →page 25[72] J. T. Kashdan and B. Thirouard. A comparison of combustion andemissions behaviour in optical and metal single-cylinder diesel engines.SAE International Journal of Engines, 2(1):1857–1872, 2009. → pages18, 19, 46, 194[73] P. Kheirkhah, P. Kirchen, and S. Rogak. Fast exhaust nephelometer (fen):A new instrument for measuring cycle-resolved engine particulateemission. SAE, (2016-01-2329), 2016. → page 197[74] M. Khosravi and P. Kirchen. Refinement of the Two-Color Method forApplication in a Direct Injection Diesel and Natural Gas CompressionIgnition Engine. Proc IMechE Part D: J Automobile Engineering, 2019. →page 161[75] M. Khosravi, J. Rochussen, J. Yeo, P. Kirchen, G. McTaggart-Cowan, andN. Wu. Effect of Fuelling Control Parameters on CombustionCharacteristics of Diesel-Ignited Natural Gas Dual-Fuel Combustion in anOptical Engine. In ASME 2016 Internal Combustion Engine Fall TechnicalConference, ICEF 2016, Greenville, SC, 2016. → pages 21, 40, 73[76] P. Kirchen, P. Obrecht, K. Boulouchos, and A. Bertola. Exhaust-stream andin-cylinder measurements and analysis of the soot emissions from acommon rail diesel engine using two fuels. ASME. J. Eng. Gas TurbinesPower, 132(11):1128041–1128048, 2010. → page 112[77] J. Kojima, Y. Ikeda, and T. Nakajima. Spatially resolved measurement ofoh∗, ch∗, and c2∗ chemiluminescence in the reaction zone of laminarmethane/air premixed flames. Proceedings of the Combustion Institute, 28(2):1757 – 1764, 2000. → pages 23, 25205[78] H. Kosaka, T. Aizawa, and T. Kamimoto. Two-Dimensional Imaging ofIgnition and Soot Formation Processes in a Diesel Flame. InternationalJournal of Engine Research, 6(1):21–42, 2005. → page 27[79] K. Koyanagi, H. O¨ing, G. Renner, and R. Maly. Optimizing commonrail-injection by optical diagnostics in a transparent production type dieselengine. SAE, (1999-01-3646), 1999. → pages 24, 46, 196[80] J. Kubesh and D. D. Brehob. Analysis of knock in a dual-fuel engine. SAE,(922367), 1992. → pages 7, 40[81] N. Ladommatos and H. Zhao. A Guide to Measurement of FlameTemperature and Soot Concentration in Diesel Engines Using theTwo-Colour Method Part I: Principles. SAE, (941956), 1994. → page 31[82] P. J. Langhorne. Reheat buzz: an acoustically coupled combustioninstability. part 1. experiment. Journal of Fluid Mechanics, 193:417–443,1988. → page 23[83] A. Larsson. Optical Studies in a DI Diesel Engine. SAE, (1999-01-3650),1999. → page 30[84] J. Lee, H. Oh, and C. Bae. Combustion process of jp-8 and fossil dieselfuel in a heavy duty diesel engine using two-color thermometry. Fuel, 102:264 – 273, 2012. Special Section: ACS Clean Coal. → page 159[85] J. G. Lee, K. Kim, and D. Santavicca. Measurement of equivalence ratiofluctuation and its effect on heat release during unstable combustion.Proceedings of the Combustion Institute, 28(1):415 – 421, 2000. → page22[86] T. Lee, J. Park, S. Kwon, J. Lee, and J. Kim. Variability in operation-basednox emission factors with different test routes, and its effects on thereal-driving emissions of light diesel vehicles. Science of the TotalEnvironment, 461-462:377 – 385, 2013. → page 9[87] W. Lee, C. E. Solbrig, T. A. Litzinger, and R. J. Santoro. Planar laser lightscattering for the in-cylinder study of soot in a diesel engine. SAE,(902125), 1990. → page 26[88] C. A. Leermakers and M. P. Musculus. In-Cylinder Soot Precursor Growthin a Low-Temperature Combustion Diesel Engine: Laser-InducedFluorescence of Polycyclic Aromatic Hydrocarbons. Proceedings of theCombustion Institute, 35(3):3079–3086, 2015. → pages 26, 27, 124206[89] M. D. Leo, A. Saveliev, L. A. Kennedy, and S. A. Zelepouga. Oh and chluminescence in opposed flow methane oxy-flames. Combustion andFlame, 149(4):435 – 447, 2007. → page 23[90] G. Li, P. Ouellette, S. Dumitrescu, and P. G. Hill. Optimization study ofpilot-ignited natural gas direct-injection in diesel engines. SAE,(1999-01-3556), 1999. → pages 41, 124, 195[91] Z. Liu and G. A. Karim. The ignition delay period in dual fuel engines.SAE, (950466), 1995. → page 43[92] H. Ma, R. Stevens, and R. Stone. In-Cylinder Temperature Estimation froman Optical Spray-Guided DISI Engine with Color-Ratio Pyrometry (CRP).SAE, (2006-01-1198), 2006. → pages 30, 31, 34, 57, 58[93] E. Mancaruso and B. M. Vaglieco. Uv-visible spectroscopic measurementsof dual-fuel pcci engine. SAE International Journal of Fuels andLubricants, 4(2):271–281, 2011. → pages 40, 81[94] E. Mancaruso, B. Vaglieco, and C. Ciaravino. Combustion Analysis in anOptical Diesel Engine Operating with Low Compression Ratio andBiodiesel Fuels. SAE, (2010-01-0865), 2010. → pages 24, 118, 159[95] Y. Matsui, T. Kamimoto, and S. Matsuoka. A Study on the Time and SpaceResolved Measurement of Flame Temperature and Soot Concentration in aD. I. Diesel Engine by the Two-Color Method. SAE, (790491), 1979. →page 31[96] Y. Matsui, T. Kamimoto, and S. Matsuoka. A Study on the Application ofthe Two-Color Method to the Measurement of Flame Temperature and SootConcentration in Diesel Engines, 1980. → pages 28, 30, 31, 180[97] G. McTaggart-Cowan. Pollutant formation in a gaseous-fuelled, directinjection engine. Phd dissertation, University of British Columbia, 2006.→ pages 6, 9, 10, 41[98] G. McTaggart-Cowan, K. Mann, J. Huang, A. Singh, B. Patychuk, Z. X.Zheng, and S. Munshi. Direct injection of natural gas at up to 600 bar in apilot-ignited heavy-duty engine. SAE International Journal of Engines, 8(3):981–996, 2015. → pages 41, 42, 139, 149, 182[99] G. P. McTaggart-Cowan, W. K. Bushe, S. N. Rogak, P. G. Hill, and S. R.Munshi. Injection Parameter Effects on a Direct Injected, Pilot Ignited,207Heavy Duty Natural Gas Engine with EGR. SAE, (2003-01-3089), 2003.→ pages 41, 100, 133, 139[100] G. P. McTaggart-Cowan, W. K. Bushe, S. N. Rogak, P. G. Hill, and S. R.Munshi. PM and NOx reduction by injection parameter alterations in adirect injected, pilot ignited, heavy duty natural gas engine with EGR atvarious operating conditions. SAE, (2005-01-1733), 2005. → pages41, 152, 194[101] S. S. Merola and B. M. Vaglieco. Optical investigations of fuel depositionburning in ported fuel injection (PFI) spark-ignition (SI) engine. Energy,34(12):2108–2115, 2009. → pages 22, 23, 24, 25[102] T. I. Mohamad, A. Yusoff, S. Abdullah, M. Jermy, M. Harrison, and H. H.Geok. The combustion and performance of a converted direct injectioncompressed natural gas engine using spark plug fuel injector. SAE,(2010-32-0078), 2010. → page 7[103] I. Mudway and F. Kelly. Ozone and the lung: a sensitive issue. MolecularAspects of Medicine, 21(1):1 – 48, 2000. → page 10[104] C. J. Mueller and G. C. Martin. Effects of oxygenated compounds oncombustion and soot evolution in a di diesel engine:broadband naturalluminosity imaging. SAE, (2002-01-1631):1–20, 2002. → pages 20, 21[105] U. Muller. Reduzierter Reaktionsmechanismen fur die Zundung vonn-Heptan und iso-Oktan unter motorrelevanten Bedingungen. Phddissertation, RWTH-Aachen, 1993. → page 24[106] T. Muruganandam, B.-H. Kim, M. Morrell, V. Nori, M. Patel, B. Romig,and J. Seitzman. Optical equivalence ratio sensors for gas turbinecombustors. Proceedings of the Combustion Institute, 30(1):1601 – 1609,2005. → page 22[107] M. P. Musculus, J. E. Dec, and D. R. Tree. Effects of Fuel Parameters andDiffusion Flame Lift-Off on Soot Formation in a Heavy-Duty DI DieselEngine. SAE, (2002-01-0889), 2002. → page 27[108] M. P. Musculus, S. Singh, and R. D. Reitz. Gradient Effects on Two-ColorSoot Optical Pyrometry in a Heavy-Duty DI Diesel Engine. Combustionand Flame, 153(1-2):216–227, 2008. → pages 21, 30, 31, 45, 147208[109] M. P. B. Musculus. Multiple simultaneous optical diagnostic imaging ofearly-injection low-temperature combustion in a heavy-duty diesel engine.SAE, (2006-01-0079), 2006. → page 20[110] M. Nesbit, M. Fergusson, A. Colsa, J. Ohlendorf, C. Hayes, K. Paquel, andJ.-P. Schweitzer. Comparative study on the differences between the eu andus legislation on emissions in the automotive sector, 2016. → page 11[111] J. Nocedal and S. J. Wright. Numerical Optimization. Springer Series inOperations Research and Financial Engineering. Springer, New York, 2nded. edition, 2006. → page 95[112] P. Ouelette, D. Goudie, and G. McTaggart-Cowan. Progress in thedevelopment of natural gas high pressure direct injection for Euro VIheavy-duty trucks. In J. Liebl and C. Beidl, editors, InternationalerMotorenkongress 2016: Mit Konferenz Nfz-Motorentechnologie, pages591–607. Springer Fachmedien Wiesbaden, Wiesbaden, 2016. → pages8, 41, 43, 45, 158[113] P. Ouellette and P. Hill. Turbulent transient gas injections. Journal ofFluids Engineering, 122(4):743–752, 2000. → page 148[114] P. Ouellette, P. L. Mtui, and P. G. Hill. Numerical simulations of directlyinjected natural gas and pilot diesel fuel in a two-stroke compressionignition engine. SAE, (981400), 1998. → pages 41, 195[115] R. Papagiannakis and D. Hountalas. Experimental investigation concerningthe effect of natural gas percentage on performance and emissions of a didual fuel diesel engine. Applied Thermal Engineering, 23(3):353 – 365,2003. → page 43[116] B. D. Patychuk. Particulate matter emission characterization from anatural-gas high-pressure direct-injection engine. PhD thesis, Universityof British Columbia, 2013. → page 7[117] F. Payri, J. V. Pastor, J. M. Garcı´a, and J. M. Pastor. Contribution to theApplication of Two-Colour Imaging to Diesel Combustion. MeasurementScience and Technology, 18(8):2579–2598, 2007. → pages30, 31, 57, 58, 111[118] R. Pearce and A. Gaydon. Identification of Molecular spectra. Routledgepublishers, London, 4th ed. edition, 1976. → page 23209[119] J. A. Pinson, T. Ni, and T. A. Litzinger. Quantitative Imaging Study of theEffects of Intake Air Temperature on Soot Evolution in anOptically-Accessible D.I. Diesel Engine. SAE, (942044), 1994. → page 27[120] C. Potera. Air pollution: Salt mist is the right seasoning for ozone.Environmental Health Perspectives, 116:A288, 2008. → page 8[121] M. Powell. A Fortran Subroutine for Solving Systems of NonlinearAlgebraic Equations. In P. Rabinowitz, editor, Numerical methods fornonlinear algebraic equations, page 199. Gordon and Breach SciencePublishers, 1970. → page 95[122] W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery.Numerical Recipes in FORTRAN : the Art of Scientific Computing.Cambridge University Press, New York, 1993. → page 95[123] R. Price, I. Hurle, and T. Sugden. Optical studies of the generation of noisein turbulent flames. Symposium (International) on Combustion, 12(1):1093– 1102, 1969. → page 23[124] J. Rochussen. Thermodynamic and Optical Investigation of theCombustion Mechanisms of Diesel-Ignited Dual-Fuel Natural GasCombustion. Masters dissertation, University of British Columbia, 2015.→ pages 37, 50, 71[125] J. Rochussen and P. Kirchen. Characterization of reaction zone growth inan optically accessible heavy-duty diesel/methane dual-fuel engine.International Journal of Engine Research, 0(0):1468087418756538, 2018.→ page 122[126] J. Rochussen and P. Kirchen. Parametric study of pilot-igniteddirect-injection natural gas combustion in an optically-accessibleheavy-duty engine. International Journal of Engine Research, 0(0):1–20,2019. → pages 117, 122, 124, 141, 142, 143, 146, 149, 152, 154, 161, 221[127] J. Rochussen, J. Yeo, and P. Kirchen. Effect of fueling control parameterson combustion and emissions characteristics of diesel-ignited methanedual-fuel combustion. SAE, (2016-01-0792), 2016. → pages37, 73, 74, 75, 79, 83, 84, 90[128] J. Rochussen, J. Son, J. Yeo, M. Khosravi, P. Kirchen, andG. McTaggart-Cowan. Development of a Research-Oriented CylinderHead with Modular Injector Mounting and Access for Multiple In-CylinderDiagnostics. SAE, (2017-24-0044), 2017. → pages 63, 64, 174210[129] A. M. Rothrock. Photographic study of combustion incompression-ignition engine. SAE, (340094):203–211, 1934. → page 20[130] B. Sahoo, N. Sahoo, and U. Saha. Effect of engine parameters and type ofgaseous fuel on the performance of dual-fuel gas diesel engines—a criticalreview. Renewable and Sustainable Energy Reviews, 13(6):1151 – 1184,2009. → pages 36, 37[131] J.-M. SAMANIEGO, F. N. EGOLFOPOULOS, and C. T. BOWMAN.Co2∗ chemiluminescence in premixed flames. Combustion Science andTechnology, 109(1-6):183–203, 1995. → page 23[132] A. Sarvi, J. Lyyra¨nen, J. Jokiniemi, and R. Zevenhoven. Particulateemissions from large-scale medium-speed diesel engines: 1. particle sizedistribution. Fuel Processing Technology, 92(10):1855 – 1861, 2011. →pages 9, 10[133] S. Schlatter, B. Schneider, Y. Wright, and K. Boulouchos. Experimentalstudy of ignition and combustion characteristics of a diesel pilot spray in alean premixed methane/air charge using a rapid compression expansionmachine. SAE, (2012-01-0825), 2012. → pages16, 34, 38, 39, 44, 86, 89, 90, 91[134] J. Seinfeld and S. Pandis. Atmospheric Chemistry and Physics: From AirPollution to Climate Change. Wiley, 2011. → page 9[135] J. S. Shakal and J. K. Martin. Imaging and Spatially Resolve Two-ColorTemperature Measurements Through a Coherent Fiberoptic: Observationof Auxiliary Fuel Injection Effects on Combustion in a Two-Stroke DIDiesel. SAE, (940903), 1994. → page 18[136] S. Singh, R. D. Reitz, and M. P. Musculus. 2-Color ThermometryExperiments and High-Speed Imaging of Multi-Mode Diesel EngineCombustion. SAE, (2005-01-3842), 2005. → pages21, 30, 33, 34, 45, 158, 159[137] H. A. Stadler, M. Fo¨rster, and R. Kneer. Experimental investigation ofch-chemiluminescence in an oxyfuel flame. In 3rd European CombustionMeeting : ECM 2007, Chania, Crete, Greece 11 - 13 April 2007 ;proceedings / Mediterranean Agronomic Institute of Chania. CombustionInstitute Greek Section, Chania, 2007. MAICh. → pages 22, 23211[138] A. Starikovskiy and N. Aleksandrov. Plasma-assisted ignition andcombustion. Progress in Energy and Combustion Science, 39(1):61 – 110,2013. → page 7[139] R. R. Steeper and E. J. Stevens. Characterization of combustion, pistontemperatures, fuel sprays, and fuel-air mixing in a disi optical engine. SAE,(2000-01-2900), 2000. → page 46[140] B. D. Stojkovic, T. D. Fansler, M. C. Drake, and V. Sick. High-speedimaging of OH∗ and soot temperature and concentration in astratified-charge direct-injection gasoline engine. Proceedings of theCombustion Institute, 30 II(2):2657–2665, 2005. → pages 21, 22, 23[141] K. I. Svensson, A. J. Mackrory, M. J. Richards, and D. R. Tree. Calibrationof an RGB, CCD Camera and Interpretation of its Two-Color Images forKL and Temperature. SAE, (2005-01-0648), 2005. → page 31[142] M. Taschek, P. Koch, J. Egermann, and A. Leipertz. Simultaneous opticaldiagnostics of hsdi diesel combustion processes. SAE, (2005-01-3845):1–13, 2005. → pages 19, 21, 24[143] H. Tian. Noise Analysis in CMOS Image Sensors. Phd dissertation,Stanford University, 2000. → page 103[144] C. Tighe, M. Twigg, A. Hayhurst, and J. Dennis. The kinetics of oxidationof diesel soots by no2. Combustion and Flame, 159(1):77 – 90, 2012. →page 9[145] C. Tornatore, S. Merola, and P. Sementa. Combustion Process Investigationin a Small SI Engine using Optical Diagnostics. SAE, (2010-01-2262),2010. → pages 22, 23[146] D. R. Tree and J. E. Dec. Extinction Measurements of in-Cylinder SootDeposition in a Heavy-Duty DI Diesel Engine. SAE, (2001-01-1296),2001. → pages 26, 27[147] D. R. Tree and K. I. Svensson. Soot Processes in Compression IgnitionEngines. Progress in Energy and Combustion Science, 33(3):272–309,2007. → pages 26, 27, 28, 29, 31, 32[148] K. Walsh, M. Long, M. Tanoff, and M. Smooke. Experimental andcomputational study of ch∗, ch∗, and oh∗ in an axisymmetric laminardiffusion flame. Symposium (International) on Combustion, 27(1):615 –212623, 1998. Twenty-Seventh Sysposium (International) on CombustionVolume One. → page 24[149] M. Walsh. Mobile Source Related Air Pollution: Effects on Health and theEnvironment, pages 803–809. Elsevier, 12 2011. → page 10[150] F. Wang, J. B. Liu, J. Sinibaldi, C. Brophy, A. Kuthi, C. Jiang, P. Ronney,and M. A. Gundersen. Transient plasma ignition of quiescent and flowingair/fuel mixtures. IEEE Transactions on Plasma Science, 33(2):844–849,2005. → page 7[151] X. Wang, D. Westerdahl, J. Hu, Y. Wu, H. Yin, X. Pan, and K. M. Zhang.On-road diesel vehicle emission factors for nitrogen oxides and blackcarbon in two chinese cities. Atmospheric Environment, 46:45 – 55, 2012.→ page 9[152] J. Warnatz, U. Maas, and R. Dibble. Combustion; Physical and ChemicalFundamentals, Modeling and Simulation, Experiments, PollutantFormation. Springer, Berlin, Germany, 2006. → page 7[153] C. S. Weaver and S. H. Turner. Dual fuel natural gas/diesel engines:Technology, performance, and emissions. SAE, (940548), 1994. → pages7, 36, 40[154] L. Wei and P. Geng. A review on natural gas/diesel dual fuel combustion,emissions and performance. Fuel Processing Technology, 142:264 – 278,2016. → pages 36, 43[155] J. Yan and G. L. Borman. Analysis and in-cylinder measurement ofparticulate radiant emissions and temperature in a direct injection dieselengine. SAE, (881315), 1988. → pages 17, 64, 65[156] J. Yeo. Development and application of in-cylinder fuel concentration andpyrometry optical diagnostic tools in diesel-ignited dual-fuel natural gasengines. Masters dissertation, University of British Columbia, 2017. →pages 49, 50, 63, 64, 65, 66, 67, 142, 184[157] A. I. Zayed. Weierstrass Transform. In Handbook of function andgeneralized function transformations, chapter 18, page 643. CRC Press,Boca Raton, FL, 1996. → page 108[158] K. Zha, R.-C. Florea, and M. Jansons. Soot Evolution With CyclicCrank-Angle-Resolved Two-Color Thermometry in an Optical Diesel213Engine Fueled With Biodiesel Blend and ULSD. Journal of Engineeringfor Gas Turbines and Power-Transactions of the Asme, 134(9):1–7, 2012.→ pages 30, 31, 35, 36, 57, 58[159] J. Zhang, W. Jing, and T. Fang. High speed imaging of OH∗chemiluminescence and natural luminosity of low temperature diesel spraycombustion. Fuel, 99:226–234, 2012. → pages 16, 24, 118[160] J. Zhang, W. Jing, W. L. Roberts, and T. Fang. Soot Temperature and KLFactor for Biodiesel and Diesel Spray Combustion in a Constant VolumeCombustion Chamber. Applied Energy, 107:52–65, 2013. → pages30, 31, 57, 58[161] Q. Zhang, Y. Gong, Q. Guo, X. Song, and G. Yu. Experimental study onch∗ chemiluminescence characteristics of impinging flames in an opposedmulti-burner gasifier. AIChE Journal, 63(6):2007–2018, 2017. → page 25[162] H. Zhao and N. Ladommatos. Optical Diagnostics for Soot andTemperature Measurement in Diesel Engines. Progress in Energy andCombustion Science, 24(3):221–255, 1998. → pages 26, 28, 30, 31, 147[163] P. Zoldak, J. John Joseph, W. Shelley, J. Johnson, and J. Naber.Characterization of partially stratified direct injection of natural gas forspark-ignited engines. SAE, (2015-01-0937), 2015. → page 7214Appendix AHigh-Speed Imaging SystemConsiderationsThe design of the CMOS (complementary metal-oxide-semiconductor) sensors,with each photodiode being paired with an amplifier, making each pixel a separatelight detection unit, allows high frame rates of 4,800 and 5,400 frames-per-second(fps) at the full resolution for the Phantom V7.1 and Photron SA-1 cameras used inthis work, respectively. This is in contrast with the CCD (charged-coupled device)sensors, where one or more amplifiers are located at the edges of the sensor to readout the pixel charges successively, making the process slower and more energydemanding.The combination of the camera sensor resolution and the focal length of the ob-jective lens, determines the stand-off distance of the camera from the combustionchamber to maximize the recorded image resolution and spatial information. Themaximum frame rate of the camera, on the other hand, is determined by the ad-justable pixel resolution. Therefore, the final resolution of the camera sensor, andconsequently the recorded image, are reduced to allow recording images at higherframe rates than the maximum values at the full resolution and reach higher tem-poral resolutions of 0.25-0.5 CAD/frame. Managing the amount of light reachingthe camera sensor is the other important camera setting and the only way to phys-ically control the SNR (as opposed to the image post-processing considerations).It is worth noting here that the commonly used ISO setting in photography as a215means to adjust the “sensitivity” of the the sensor is merely adjusting the dynamicrange of the output image through applying a post-senor gain to the signal from thesensor. The sensitivity of the sensor is constant and determined by the quantum ef-ficiency of the CMOS photodiode arrays. The amount of light reaching the camerasensor is controlled by the camera exposure duration and the lens aperture. Theexposure duration in scientific high-speed CMOS cameras, such as the ones usedin this work, is tightly controlled by the high shutter rates of ~1 µsec, affordedby the on-chip global electronic shutter design. This shutter rate also defines thelower limit for the possible exposure duration settings. The upper limit for expo-sure duration is determined by the image recording frame rate (i.e., the intervalof two consecutive image frames), which can be reached at lower light intensitycases, leaving the lens aperture the only sensor exposure control factor.To present an estimate of the range of light intensities to expect on the imagingsystem, a natural luminosity imaging (with no spectral/neutral density filters) of alow-load diesel combustion, using the HPDI injector, nearly saturated the sensorof the Phantom camera (ISO: 4,800) at 10 µsec exposure duration and f/5.6 lensaperture setting.Images of a rapid phenomena captured by the high-speed cameras are usuallystored on the internal memory of the camera, referred to as the camera buffer, lim-iting the number of image frames that can be recorded for the incident. As a result,partitioning capabilities are considered in the camera system design to optimizethe use of the limited camera buffer and minimize “empty” frames captured beforeand after the phenomena of interest. The V7.1 Phantom camera used in the currentwork provided up to 15 partitions, accommodating image sequences from 15 in-dividual combustion cycles to examine cycle-to-cycle variability and for ensembleaveraging purposes.216Appendix BEnsemble Averaging of Resultsvs. Raw signalThe spatially resolved two-color pyrometry results presented in this work wereacquired from application of the two-color pyrometry algorithm on the ensembleaverage raw image frames at each crank position (SNR enhancement approach).As depicted in Figure B.1, ensemble averaging of the two-color pyrometry re-sults instead would lead to inaccurate temperature estimations due to cycle-to-cyclevariability effects on distribution of the soot cloud signal and “0” values. The un-resolved regions within individually resolved 2D two-color pyrometry results iscaused by missing incandescent soot cloud signal, low SNR signal below the detec-tion limit of the current imaging system, method imperfections, or a combinationof these factors.217Figure B.1: Effect of ensemble averaging on the two-color pyrometry tem-perature distributions instead of using it as a SNR enhancement ap-proach for the raw data218Appendix CPIDING Fuel Injection ProfileThe estimated fuel injection profiles for PIDING combustion are presented in Fig-ure C.1. The pulse magnitudes are presented in terms of estimated mass flow ratesnormalized by the baseline value (see Table 6.1).219Figure C.1: Estimated PIDING fuel injection profile. Mechanical injectiondelay was estimated through illuminated combustion chamber opticalmeasurements and needle ramp-up and ramp-down timings were esti-mated in a previous work [37]220Appendix DOptical and ThermodynamicIgnition DelaysThe optical and thermodynamic ignition delay (τign) values for the operating pointsdescribed in Table 6.1 can be compared in Table D.1. The optical ignition delayswere evaluated in this work based on the first indication of the OH*-CL signal in2D measurements. A single frame delay was considered for all optical ignitiondelay values in order to ensure distinguishing between the OH*-CL signal frompilot combustion vs. NG ignition. The absolute values of the optical ignition delayis dependant on the optical imaging settings. The thermodynamic ignition delaysfor the considered operating conditions were evaluated in conceptual combustionmode analysis of the same campaign [126]. Comparison of the ignition delay val-ues in this table shows consistent influence trends for the considered PIDING fuel-ing parameter sweeps.221Table D.1: Comparison of optical and thermodynamic ignition delays for op-erating points described in Table 6.1PointPilot τign [CAD] NG τign [CAD]Optical Thermo. Optical Thermo.Baseline 13 8.2 10 7.0Short GPW 13 8.2 10 7.0Long GPW 13 8.2 10 7.0Short RIT 11.5 6.8 12.5 9.2Long RIT 16 10.9 10.5 6.9Negative RIT 13 8.5 19-22 19.8High Pinj 12 7.0 10 5.7Low Pinj 15 10.3 11.5 8.7222Appendix ENG Jet Momentum ReflectionEffectsFollowing the formation and merging of the initial soot pockets to form a toroidalsoot cloud, it grows towards the central bowl region (see §6.2.1). Small high con-centration soot pockets, depicted as black regions in KL factor distributions of Fig-ure E.1, can be traced during this process. This phenomenon reveals that advectioneffects due to the reflected NG jets momentum after impingement is a contribut-ing factor in soot cloud growth process, in addition to the continued formation andaccumulation effects.223Figure E.1: Soot cloud advection effects with reflected NG jet momentumafter impingement onto the walls. Image frames show KL factor distri-bution from the baseline operating condition described in Table 6.1224Appendix FMean Soot Cloud Growth Rate2D two-color pyrometry temperature distributions, shown in Figure F.1, were usedto evaluate soot cloud growth rates. The mean soot cloud growth rate was calcu-lated based on 1000 rpm engine speed and the crank position after onset of sootsignal (CAD aOS) where the soot cloud front (RSCF ) reaches a marked (imaginary)ring with a radius equal to 30% of the visible bowl radius (83 mm). Note that at0 CAD aOS initial soot cloud pockets are adjacent to the wall (RSCF,0 = Rbowl)Figure F.1: The influence of Pin j on 2D two-color pyrometry temperature dis-tribution. The crank positions are relative to onset of soot signal and theblack rings mark 30% visible bowl radius (see Table 6.1 for operatingpoint specifications)225Appendix GEnsemble Averaging of Resultsvs. Raw signalAlternative approaches towards application of the weighted spatial averaging func-tion on the 2D two-color pyrometry results are illustrated in Figure G.1. Thecorresponding WSA results from these approaches for the long GPW operatingcondition described in Table 7.1 are compared in Figure G.2. Temperature val-ues reported are restricted to the range where the mean KL factor is higher than0.025, which is ~5% of the peak KL value. It is clear from this figure that the EWPapproach leads to incorrect temperature and KL factor results. The WPE, WEP,EPW approaches lead to very similar results before late cycles stage at >40 CADaTDC. Similar to the probe results, WPE and WEP approaches lead show a dropin temperature values during the late-cycle stage and the KL signal does not reduceto near zero values as it does for the PWE and EPW approaches. The KL factorvalues for this and other investigated operating conditions show closer agreementwith the results acquired using the PWE approach.226Figure G.1: Alternative WSA procedures flowchart0 10 20 30 40 50 60Crank Position [CAD aTDC]60080010001200140016001800200022002400Temp [K] factorProbe Mean 2D WSAEPW WSAPWE WSAWPE WSAWEP WSAEWPFigure G.2: Alternative WSA procedures; Sample results on long GPWoperating point in Table 7.1227Appendix HRadiation Heat Loss throughQuartz WindowIn the absence of the quartz window in the all-metal engine, lower radiation heatlosses can be thought as a contributing factor for higher estimated in-cylinder tem-peratures. A simplified radiation heat loss model for comparison of the two engineconfigurations is shown in Figure H.1, which suggests a 45% lower heat loss in theall-metal engine.Figure H.1: A simplified radiation heat loss model for comparison of the twoengine configurations228


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items