Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Characterization and system level study of air addition in a pilot ignited direct injection natural gas… Singh, Aditya Prakash 2019

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

Item Metadata

Download

Media
24-ubc_2019_may_singh_adityaprakash.pdf [ 5.71MB ]
Metadata
JSON: 24-1.0376211.json
JSON-LD: 24-1.0376211-ld.json
RDF/XML (Pretty): 24-1.0376211-rdf.xml
RDF/JSON: 24-1.0376211-rdf.json
Turtle: 24-1.0376211-turtle.txt
N-Triples: 24-1.0376211-rdf-ntriples.txt
Original Record: 24-1.0376211-source.json
Full Text
24-1.0376211-fulltext.txt
Citation
24-1.0376211.ris

Full Text

  i   CHARACTERIZATION AND SYSTEM LEVEL STUDY OF AIR ADDITION IN A PILOT IGNITED DIRECT INJECTION NATURAL GAS ENGINE  by  Aditya Prakash Singh (आदित्य प्रकाश द िंह)  B. Tech (Hons), Indian Institute of Technology Bombay, 2015  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Mechanical Engineering)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  January 2019  © Aditya Prakash Singh, 2019    ii   The following individuals certify that they have read, and recommend to the Faculty of Graduate and Postdoctoral Studies for acceptance, a thesis/dissertation entitled:  Characterization and system level study of air addition in a pilot ignited direct injection natural gas engine  submitted by Aditya Prakash Singh  in partial fulfilment of the requirements for the degree of Master of Applied Science in Mechanical Engineering  Examining Committee: Dr. Patrick Kirchen, Mechanical Engineering Supervisor  Dr. Steven Rogak, Mechanical Engineering Supervisory Committee Member   Supervisory Committee Member Dr. Gordon McTaggart-Cowan, Engine R&D Applications, Westport Fuel Systems Additional Examiner     Additional Supervisory Committee Members:  Supervisory Committee Member  Supervisory Committee Member  iii  Abstract The emissions from heavy duty engines must be reduced in light of their climate and health effects. Pilot-ignited direct injection natural gas engines (PIDING) allows cleaner combustion by using natural gas as the primary fuel instead of diesel. To achieve further emission reductions, the concept of air addition to natural gas, at various global equivalence ratio (Φ) and EGR rates was investigated in a heavy-duty mode of a PIDING engine. However, this concept requires compressed air over 300 bar, which may impose net compression work to the engine. Therefore, the system level implications were investigated by developing and characterizing a prototype compression system, and subsequently considering its compression work with the engine indicated efficiency. The investigations were carried out using a single cylinder research engine and an industrial reciprocating compressor.   Fuel dilution by air addition was demonstrated to effectively reduce emissions of PM, CO, and THC. Particulate matter (PM) reduced exponentially, resulting in more than an order of magnitude reduction. Similarly, carbon monoxide (CO) was also reduced albeit with lower magnitude. The total unburnt hydrocarbons (THC) reductions with air addition were significant only at high EGR (> 12.5%). However, air addition significantly increased NOx emissions (up to factor of 2.5); but increasing the EGR rates by 6.5%-point may compensate for this.  The net compression work was dependent on the engine operating conditions, number of compressor stages, and performance of the chosen compressor. Air addition resulted in indicated efficiency improvements on the order of 2.5%, which at high Φ were sufficient to compensate for the compression work of a three-stage reciprocating compressor. The same may be possible with even smaller and efficient 2-stage compressors.  An optimization study suggests that air addition would be most effective at high values of Φ, EGR rates, and air addition. As an in-cylinder strategy, it has the potential to be used in conjunction with ultra-high EGR, to significantly reduce both PM and NOx emissions, while avoiding the typical issues of ultra-high EGR: combustion instabilities and high THC and CO emissions.  iv  Lay Summary Heavy duty diesel engines are known for their efficiency and reliability but are a significant source of harmful air pollutants like NOx and particulate matter. To achieve better air quality via control of pollutants, emission regulations have become increasingly stringent, leading engine manufacturers to develop clean combustion technologies. Existing technology High Pressure Direct Injection (HPDI) by Westport fuel systems achieves relatively cleaner combustion by fuelling diesel engines with natural gas while retaining the fuel economy of diesel; however, it still requires complex and expensive after treatment systems to meet the regulations. This study investigates air addition to natural gas, a patented concept which has the potential to further reduce emissions in HPDI engines. The study also investigates the system implications of air addition by analyzing the parasitic power consumed from suitable compression systems required for this concept.      v  Preface This thesis presents the author’s contribution towards the characterization and system level assessment of air addition concept and its relevant compression system. Air addition is a patented concept from sponsor Westport Fuel Systems, and therefore the project was executed in collaboration with them. Lucien Halada, a visiting student from ETH Zurich, along with the support of retired technician, Robert Parry, had commissioned the first form of air addition system on the previous single cylinder research engine. Post the failure of the engine in Dec 2015, the engine was replaced by Robert Parry, and recommissioned by the joint efforts of the author, Pooyan Khierkhah, and Rene Zepeda over the course of a year. The author recommissioned the air addition system on the new engine, with changes introduced to enhance operator safety and access.  All data analysis, numerical models, response surface models, proposed optimum operation strategy, NOx control strategy, and conclusions presented on this thesis are the original work of the author, executed under the direct supervision of Dr. Patrick Kirchen. The methodology of experiments, and the test matrix designs were reviewed by the author’s supervisor- Dr. Patrick Kirchen, the engineers at Westport (Drs. McTaggart-Cowan, Jian Huang and Ashish Singh), and Dr. Steven Rogak.  The MASc candidate designed the test matrix, performed the experiments, and led the maintenance of engine and its subsystems for the air addition experiments (Chapter 4), which was the first experimental campaign on the new engine. The results and discussion in Section 4.3 and 4.5 were presented at the 2018 CICS spring meeting in the form of a conference paper titled “The Effect of Fuel-Air Dilution for varying EGR and Equivalence Ratio in a Direct Injection Natural Gas Engine” with Dr. Kirchen as the co-author. The lead author performed all written and analytical work with assistance from Dr. Kirchen.   The author ordered relevant instrumentation, conducted FMEA, designed operating procedures, and conducted relevant screening tests leading to the development of a functioning 2-stage prototype reciprocating compressor based on the 3-stage compressor. Riley Cahill collected the data required for the characterization of the 2-stage compressor. The methods of compressor friction estimation are the original work of the author, and its relevant experiments were conducted by the author.  vi  Table of Contents Abstract ......................................................................................................................................... iii Lay Summary ............................................................................................................................... iv Preface .............................................................................................................................................v Table of Contents ......................................................................................................................... vi List of Tables ................................................................................................................................ xi List of Figures ............................................................................................................................. xiii Nomenclature ............................................................................................................................ xvii Acknowledgements ......................................................................................................................xx Dedication ................................................................................................................................... xxi  Introduction..................................................................................................................1 1.1 CI Engines and Air Pollution .......................................................................................... 1 1.2 Diesel Pollutant Regulations ........................................................................................... 2 1.3 Emission Reduction in CI Engines ................................................................................. 3 1.3.1 Existing Systems and Strategies ................................................................................. 3 1.3.2 Natural Gas Fuelling: HPDI........................................................................................ 4 1.4 Air Addition in PIDING Engines ................................................................................... 5 1.5 Objectives ....................................................................................................................... 6 1.6 Thesis Structure .............................................................................................................. 7  Background and Theory .............................................................................................9 2.1 Air Addition .................................................................................................................... 9 2.1.1 In-cylinder Strategies in PIDING Engines ................................................................. 9 2.1.2 Conceptual Combustion ............................................................................................ 12 vii  2.1.3 Pollutant formation ................................................................................................... 13 2.1.3.1 Particulate Matter .............................................................................................. 13 2.1.3.2 Oxides of Nitrogen ............................................................................................ 15 2.1.3.3 Total Unburnt Hydrocarbons ............................................................................ 17 2.1.3.4 Carbon monoxide .............................................................................................. 18 2.2 Diluent Compression .................................................................................................... 18 2.2.1 Compression Efficiencies ......................................................................................... 19 2.2.2 Mechanical Efficiency .............................................................................................. 23 2.2.3 Transmission Efficiency ........................................................................................... 25 2.3 Summary / Literature Gap ............................................................................................ 26  Experimental System .................................................................................................27 3.1 Single Cylinder Research Engine (SCRE) .................................................................... 27 3.1.1 Data Acquisition and Instrumentation ...................................................................... 30 3.1.2 Flow Measurement.................................................................................................... 31 3.1.3 In-Cylinder Pressure Measurement........................................................................... 32 3.1.4 Particulate Emissions Measurement ......................................................................... 34 3.1.5 Gaseous Emissions Measurement ............................................................................. 35 3.2 Diluent (Air) Compression System ............................................................................... 38 3.2.1 Development of Prototype 2-stage configuration ..................................................... 40 3.2.2 Compressor Pressure, Temperature, and Flow measurement ................................... 42 3.2.3 Motor Power, RMS Voltage and Shaft Speed measurement .................................... 44 3.2.4 Data Acquisition System........................................................................................... 45 3.3 Measurement Uncertainty ............................................................................................. 45 viii   On-Engine Air Addition Experiments .....................................................................49 4.1 Methodology ................................................................................................................. 49 4.2 Combustion Effects ....................................................................................................... 52 4.2.1 Apparent Heat Release Rates .................................................................................... 53 4.2.2 Combustion Duration ................................................................................................ 57 4.2.3 Combustion Efficiency and Stability ........................................................................ 60 4.2.4 Adiabatic Flame Temperatures ................................................................................. 61 4.2.5 Maximum Pressure Rise Rate ................................................................................... 64 4.3 Emissions ...................................................................................................................... 65 4.3.1 PM and CO ............................................................................................................... 65 4.3.2 THC........................................................................................................................... 67 4.3.3 NOx ........................................................................................................................... 70 4.4 Expansion Work of Diluent .......................................................................................... 73 4.5 Gross Indicated Efficiency ............................................................................................ 76 4.6 Conclusions ................................................................................................................... 81  Development and Characterization of Diluent Compressor .................................83 5.1 Methodology ................................................................................................................. 83 5.2 Power Consumption and Output Flowrate .................................................................... 84 5.3 Polytropic Index ............................................................................................................ 86 5.4 Friction Estimation........................................................................................................ 87 5.4.1 Overlap Method ........................................................................................................ 87 5.4.2 Motor Efficiency Method ......................................................................................... 90 5.4.3 Analysis..................................................................................................................... 92 ix  5.5 Compressor Performance .............................................................................................. 93 5.5.1 Specific Power .......................................................................................................... 93 5.5.2 Volumetric Efficiency ............................................................................................... 96 5.5.3 Mechanical Efficiency and Overall Efficiency ......................................................... 98 5.6 Conclusions ................................................................................................................. 101  System Level Study of Air Addition.......................................................................103 6.1 Response Surfaces of Experiments ............................................................................. 103 6.2 NOx Control Strategy ................................................................................................. 105 6.3 System Efficiency ....................................................................................................... 108 6.4 Optimum Operating Spaces ........................................................................................ 112 6.5 Conclusions ................................................................................................................. 117  Conclusions and Future Work ...............................................................................119 7.1 General Conclusions ................................................................................................... 119 7.2 Recommendations for future work ............................................................................. 124 Bibliography ...............................................................................................................................126 Appendices ..................................................................................................................................137 Appendix A Calculation of Uncertainty ................................................................................. 137 A.1 Uncertainty Analysis: Instrument Specifications.................................................... 137 A.2 Sample Uncertainty Calculations ............................................................................ 138 Appendix B On-Engine Air Addition Experiments: Additional Information ........................ 142 B.1 Determination of Realistic Engine Operating Mode .............................................. 142 B.2 Effect of EGR and Φ on AHRR .............................................................................. 146 B.3 Calculation of Adiabatic Flame Temperatures ....................................................... 148 x  B.4 Sensitivity Analysis of Expansion Power of Diluent .............................................. 151 B.5 Summary Tables ..................................................................................................... 153 Appendix C System Level Study of Air Addition: Additional Information ........................... 162 C.1 Response Surface Relevant Information ................................................................. 162 C.2 ANOVA Summary Table ....................................................................................... 168 Appendix D P&ID .................................................................................................................. 169 Experimental Facility: Additional Information ...................................................................... 171  xi  List of Tables Table 2.1 Effects of various in-cylinder strategies of PIDING engines on emissions and performance .................................................................................................................................. 11 Table 2.2 Relevant specifications  and efficiencies  of  reciprocating compressors in various applications ................................................................................................................................... 24 Table 3.1 Key specifications of Single cylinder research engine ................................................. 27 Table 3.2 List of general instrumentation used in SCRE experiments ......................................... 31 Table 3.3 Key specifications of 3-stage compressor .................................................................... 39 Table 3.4 Instrumentation used in compressor characterization ................................................... 43 Table 3.5 Repeat points of experiments in Chapter 4 ................................................................... 48 Table 3.6 Uncertainty analysis at the primary repeatability point ................................................ 48 Table 4.1 Engine operating set points for air addition experiments ............................................. 51 Table 4.2 Comparison of Engine operating set points: air addition experiments and previous N2 addition study [23] ........................................................................................................................ 52 Table 5.1 Compressor operating set points. .................................................................................. 84 Table 6.1 Response surface equations and constants.................................................................. 105 Table 6.2 The proposed EGR for NOx control at various baseline EGRs and Φs, and its qualitative effect on PM emissions ............................................................................................................... 108 Table 6.3 Optimization solution and relevant data ..................................................................... 116 Table 6.4 Sensitivity of optimum solution to the target values in merit function ...................... 116 Table 7.1 Uncertainty relevant information: Exhaust Gas Analyzer. ......................................... 137 Table 7.2 Uncertainty relevant information: PM measurements. ............................................... 137 Table 7.3 Uncertainty relevant information: Miscellaneous measurements. .............................. 138 xii  Table 7.4 Sensitivity of expansion power of diluent to the assumed temperature at BDC. ....... 151 Table 7.5 Sensitivity of expansion power of diluent to the assumed stagnation temperature of the diluent jet .................................................................................................................................... 151 Table 7.6 Standard errors of various response surfaces .............................................................. 162  xiii  List of Figures Figure 1.1 US EPA Heavy Duty On-road Engines Emission Standards of PM and NOx [13] ...... 2 Figure 2.1 Schematic of typical compression cycle showing various losses ................................ 20 Figure 3.1 Single cylinder research engine ................................................................................... 28 Figure 3.2 Flow diagram of the experimental system................................................................... 30 Figure 3.3 The 3-stage reciprocating diluent (air) compressor ..................................................... 38 Figure 3.4  P&ID of the compressor in its ‘stock’ configuration with relevant instrumentation . 40 Figure 3.5  P&ID of the compressor in its 2-stage configuration ................................................. 42 Figure 3.6  Nature of uncertainty variations in different parameters (a) constant relative error variation for PM (b) constant error variation in gross indicated efficiency .................................. 47 Figure 4.1  Fuel pulse diagram: (a) HPDI (b) HPDI with air addition ......................................... 53 Figure 4.2 Effect of FDR on AHRR at various EGR and Φ ......................................................... 54 Figure 4.3  Late mixing controlled phase: (a) HPDI (b) HPDI with air addition ......................... 55 Figure 4.4 Effect of EGR on AHRR at two different FDR and Φ= 0.68 ..................................... 57 Figure 4.5 Effect of FDR on (a) overall combustion duration (b) late cycle duration (c) early cycle duration, at various EGR and Φ .................................................................................................... 59 Figure 4.6 Effect of FDR on combustion efficiency at various EGR and Φ ................................ 60 Figure 4.7 Effect of FDR on the COV of GIMEP with FDR at various EGR and Φ ................... 61 Figure 4.8 Effect of FDR on adiabatic flame temperature at various Φ and (a) EGR= 25% (b) EGR= 0% ...................................................................................................................................... 64 Figure 4.9 Effect of FDR and Φ on maximum pressure rise rate at (a) EGR= 25% (b) EGR= 0%....................................................................................................................................................... 64 xiv  Figure 4.10 Effect of FDR on PM emissions at various Φ and (a) EGR= 25% (b) EGR= 0% (c) EGR= 25% with PM emissions in log scale (d) EGR= 0% with PM emissions in log scale ....... 66 Figure 4.11 CO-PM correlation .................................................................................................... 67 Figure 4.12 Effect of FDR on THC at various EGR and Φ .......................................................... 68 Figure 4.13 Effect of FDR on NOx emissions at various Φ and (a) EGR= 25% (b) EGR= 0% .. 70 Figure 4.14 Correlation of NOx emissions index with reciprocal of representative adiabatic flame temperature at various EGR and FDR .......................................................................................... 71 Figure 4.15 PM-NOx trade-off as a function of FDR and EGR at Φ= 0.77 ................................. 72 Figure 4.16 Effect of FDR on expansion power of diluent at various Φ and EGR= 25% ............ 76 Figure 4.17 Effect of FDR on gross indicated efficiency and combustion independent indicated efficiency at various Φ and (a) EGR= 25% (b) EGR= 12.5 % (c) EGR= 0% .............................. 78 Figure 4.18 Effect of FDR on exhaust energy % at various Φ and (a) EGR= 25% (b) EGR= 12.5 % (c) EGR= 0% ............................................................................................................................ 80 Figure 5.1 Effect of compressor inlet and outlet pressure on (a) compressor outlet mass flowrates and (b) compressor motor power consumption, for 2-stage and 3-stage compressor configurations........................................................................................................................................................ 84 Figure 5.2 Variation of n of the last stage of respective configurations with (a) compressor inlet and outlet pressures and (b) compressor outlet mass flowrate ..................................................... 87 Figure 5.3 Schematic representation of friction estimation based on overlap method ................. 88 Figure 5.4 Schematic depicting work and energy transfer in various control volumes of the compressor in 3-stage configuration. ............................................................................................ 90 Figure 5.5 Motor efficiency variation with Load .......................................................................... 91 xv  Figure 5.6 Estimated friction power with overlap method as a function of third stage outlet pressure. ........................................................................................................................................ 93 Figure 5.7 Estimated friction power with motor efficiency method as a function of reference load........................................................................................................................................................ 93 Figure 5.8 Effect of total pressure ratio on (a) specific motor power and specific isentropic power and (b) specific isentropic power and specific indicated power. .................................................. 95 Figure 5.9 Effect of total pressure ratio on the specific power consumption: motor and (isentropic + friction) ...................................................................................................................................... 96 Figure 5.10 Comparison of volumetric efficiencies: Measured, modelled, and applications in literature. ....................................................................................................................................... 98 Figure 5.11 Comparison of overall efficiencies: 2-stage, 3-stage, and applications in literature...................................................................................................................................................... 100 Figure 5.12 Comparison of mechanical efficiencies: 2-stage, 3-stage, and applications in literature...................................................................................................................................................... 100 Figure 6.1 Qualitative effect of the NOx control strategy on the PM emissions at baseline EGR= 12.5% .......................................................................................................................................... 108 Figure 6.2 Comparison of motor specific power of tested with hypothetical compressor configurations ............................................................................................................................. 111 Figure 6.3 Effect of FDR on system efficiency at various Φ and EGR of (a) 3-stage tested and (b) 3-stage hypothetical compressor configurations ......................................................................... 112 Figure 6.4 Effect of FDR on system efficiency at various EGR and Φ of (a) 2-stage tested and (b) 2-stage hypothetical compressor configurations ......................................................................... 112 Figure 6.5 Variation of merit function with FDR and EGR at (a) Φ= 0.77 and (b) Φ= 0.68 ..... 115 xvi  Figure 7.1 Determination of ‘mode flexibility’. ......................................................................... 145 Figure 7.2 Determination of high PM ‘flexible modes’. ............................................................ 145 Figure 7.3 Effect of EGR on AHRR at various Φ and FDR ....................................................... 146 Figure 7.4 Effect of Φ on AHRR at various EGR and FDR ....................................................... 147 Figure 7.5 Sensitivity of AFT to (a) reactant pressure and (b) reactant temperature (c) fuel jet temperature ................................................................................................................................. 150 Figure 7.6 Effect of air addition on modelled motored in-cylinder pressure and temperature ... 152 Figure 7.7 Accuracy of response surface fit of NOx .................................................................. 163 Figure 7.8 Accuracy of response surface fit of PM .................................................................... 164 Figure 7.9 Accuracy of response surface fit of gross indicated efficiency ................................. 165 Figure 7.10 Accuracy of response surface fit of intake surge tank pressure .............................. 166 Figure 7.11 Accuracy of fit of response surface fit of compressor motor power ....................... 167 Figure 7.12 Accuracy of response surface fit of compressor outlet mass flowrate .................... 168 xvii  Nomenclature AHRR Apparent heat release rate ATDC After top dead center CV Control volume DOC Diesel oxidation catalyst DPF Diesel particulate filter EGR Exhaust gas recirculation rate EPA  Environmental protection agency GID Gas ignition delay GISFC Gross indicated specific fuel consumption GRP Gas rail pressure or injection pressure GSOC Start of natural gas combustion IHR Integral of heat released IHRx Crank angles at x deg. ATDC of maximum IHR (x= 5, 10, 50, and 90) LHV Lower heating value PAH Polycyclic aromatic hydrocarbon PID Pilot ignition delay PIDING Pilot ignited late cycle direct injection natural gas engine PSEP Pilot separation PSOC Pilot start of combustion RIT Relative injection timing SCR Selective catalytic reduction xviii  𝐶𝑝  Specific heat (of air) at constant pressure 𝜂𝑏𝑒𝑙𝑡  Belt efficiency (of compressor) 𝜂𝑐  Combustion efficiency (of engine) 𝜂𝑖,𝑔  Gross indicated efficiency (of engine) 𝜂𝑖𝑠𝑜  Isothermal compression efficiency 𝜂𝑖𝑠𝑒𝑛   Isentropic compression efficiency 𝜂𝑂𝐸  Overall efficiency (of compressor) 𝜂𝑚𝑒𝑐ℎ  Mechanical efficiency (of compressor) 𝜂𝑚𝑜𝑡𝑜𝑟  Compressor motor efficiency 𝜂𝑝𝑜𝑙𝑦  Polytropic compression efficiency 𝜂𝑣,𝑚  Measured volumetric efficiency (of compressor stage) 𝜂𝑠𝑦𝑠  System efficiency (of engine + diluent compressor system) 𝛾  Ratio of specific heats (of air) h Specific enthalpy  ?̇?    Mass flowrate (of compressor outlet) ?̇?𝑖𝑛  inlet mass flowrate (of compressor)  n Polytropic index (of compressor stage) 𝑁  Rotational speed (rpm) (of engine; of compressor) p Pressure 𝑝𝑖𝑛  Inlet pressure (of compressor) 𝑝𝑜𝑢𝑡  Third stage outlet pressure or delivery pressure (of compressor) xix  𝑃2𝑃1  Pressure ratio across a stage (of compressor) 𝑃𝑖,𝑔  Gross indicated power (of engine) 𝑃𝑓  Rate of frictional heat dissipation (of compressor) 𝑃𝑖,𝑠, 𝑃𝑙,𝑠, and 𝑃𝑓,𝑠 Specific indicated, loss, and friction power (of compressor) 𝑃𝑚𝑜𝑡𝑜𝑟  Motor power (of compressor) 𝑃𝑚,𝑠  Motor specific power (of compressor motor) 𝑃𝑠ℎ𝑎𝑓𝑡  Shaft power (of compressor) Φ  Global equivalence ratio (of engine) R Specific gas constant of air or correlation coefficient 𝜌𝑖𝑛  inlet density (of compressor) 𝑆𝑚, 𝑆𝑠, and 𝑆𝑟  Measured, synchronous, and rated shaft speed of compressor motor T1 Temperature (of compressor inlet) 𝑉(𝜃)  Volume at a crank angle (of engine) 𝑉𝑐  Clearance volume (of compressor) 𝑉𝑑 or 𝑉𝑠 Swept/ displaced volume (of compressor stage; of engine cylinder) ?̇?   the rate of the measured indicated work (of compressor) ?̇?𝑖𝑑𝑒𝑎𝑙  Rate of work consumed by a reversible compression cycle 𝑊𝑖,𝑔  Gross indicated work (of engine) ?̇?𝑖𝑠𝑒𝑛  Isentropic power (of compressor) xx  Acknowledgements I believe every journey, every interaction, every choice that we make, shapes our life in the most beautiful ways. In my journey of pursuing this MASc degree, there were many special moments, and most importantly, special people, who left an indelible mark in my life. Among them, one that truly stands out, is my supervisor Dr. Patrick Kirchen. Aside from providing support and guidance in my project, Dr. Kirchen has had faith in me in my lowest moments during the project. His personal stories about his own struggle in communicating his research as well as his countless meticulous feedback on my slides, has greatly helped me in developing my presentation skills. It is his mentorship which has helped me in the completion of this thesis.  I would like to acknowledge Dr. Steven Rogak for his many feedback, guidance, and fruitful technical discussions. The weekly SCRE meetings with him and Dr. Kirchen have greatly helped me in enriching my understanding about engine and its subsystems. I would also like to acknowledge the technical guidance and support received from Westport’s engineers, in particular: Drs. G. McTaggart-Cowan, S. Munshi, J. Huang, and N. Wu, as well as Mr. A. Singh and R. Cahill. Dr. G. McTaggart-Cowan had provided insightful feedback on various aspects of the project while R. Cahill, a co-op student then, and now my friend, had aided me with the compressor experiments.  I would like to acknowledge my colleagues Rene, Pooyan, and Michael, who have made working in the engine lab a great experience. I would like to specially thank Rene for his aid in the calibration of PM measurement instruments during my air addition experiments. Amongst other colleagues, I would like to say my thanks to Jeremy and Mahdiar for always lending their ears for fruitful technical discussions. I have been fortunate to have found amazing friends in Vancouver. I would like to thank my close friends from my home state, in particular: Kuldeep, Samprity, and Debanga, who have made living in Vancouver such a great experience. I would also like to thank my roommate Anand, for always being there to listen to my problems, offering helpful guidance, and English grammar tips. I would also like to thank my best friends in India, Shripad and Amol, who are always there to lift my spirits and with whom I have cherished some wonderful moments of my life.  Finally, I would like to convey my heartfelt thanks to my family for their enduring faith, love and support all these years.  xxi  Dedication To my late grandmother (আইতা), my dear Mom (মাধুৰী), my Dad (ॐ), and my sister (तृष्णा), for their unquestionable faith in me.   To the beautiful Hindustani couplet, which has been a source of my inspiration for almost 16 years: “करत करत अभ्यास के, जड़मतत होत सुजान |  रसरी आवत जात ते, तसल पर परत तनसान ||” Author: Tulsidas (16th century) Consistent and deliberate practice can allow even a simple-minded person to achieve scholarly virtues, just like (even) a soft rope rubbing continuously on a stone, makes a mark on it.   1    Introduction 1.1 CI Engines and Air Pollution Diesel engines have dominated the world in heavy-duty applications (for e.g. marine, trucks, buses, off road applications) due to their inherent benefits of fuel efficiency, high torque at lower speeds, and reliability. However, they have also earned a notorious reputation for being ‘dirty’ due to it being a significant source of air pollutants: In 2014, heavy duty diesel vehicles contributed to about 15% of total black carbon emissions and nearly 34% of total NOx emissions in US [1].  Air pollutants can have severe health and environment impact. In a 2014 Word Health Organization report, an estimated 7 million premature deaths annually had been attributed to air pollution [2]. A common proxy indicator of air pollution is particulate matter (PM), which is liquid and solid aerosols suspended in atmosphere. It contains diverse toxic materials, with some of them being carcinogenic (arsenic, nickel, chromium and PAHs). A correlation between chronic and acute exposure to PM pollutants and respiratory and heart diseases has been noted in various studies [3–5]. Furthermore, carbonaceous aerosols also contribute to global warming through absorption and scattering of sunlight because of its higher Global Warming Potential (20-year GWP of Black Carbon is 3200 [6]). Other effects of aerosols include its ability to affect cloudiness [7], changing reflecting surfaces of snow and ice, which may further cause permafrost, thinning of polar ice etc. [8]. Amongst other pollutants, NOx is known to cause eye and respiratory tract irritation with moderate exposures [9]. Environmental impact resulting from NOx include the formation of photochemical smog and ozone through its reaction with reactions with volatile organics (VOCs) in the presence of sunlight [10]. NOx is also an important precursor in the 2  formation of nitrate aerosols which aside from being a significant fraction of PM2.5 [10], also leads to acid rain [9]. Other emissions like methane and CO2 are known to contribute to global warming [11].  1.2 Diesel Pollutant Regulations With increasing industrialization globally, air pollution has become a global problem. Given that a significant fraction of the major pollutants like PM and NOx is from heavy duty diesel engines, tens of countries over the world have enforced emission standards. Emission regulations in North America (US and Canada) typically align with that established by US EPA [9] while most Asian countries have adopted regulations mandated by European Commission with respective phase-in delays. In most jurisdictions, compliance of diesel engines is determined via a series of emission cycle tests, which determines whether the engine met the legislated levels for each pollutant (see [12] for more details).   Figure 1.1 US EPA Heavy Duty On-road Engines Emission Standards of PM and NOx [13] Note: The NOx standards for 2007 (0.27 g/kWhr) were phased in on a percent of sales basis, with 100% implementation in 2010. For the NOx standard of 2015, manufacturers may choose to certify to California Optional Low NOx standards of 0.13, 0.07 or 0.027 g/kWhr.  3  As can be seen from Figure 1.1, emission standards have become increasing stringent over the years [13]. Since 1974, specific emission limits of heavy duty vehicles in US for PM, NOx and CO have dropped by nearly a factor of 60, 54, and 3, respectively. Furthermore, latest Emission standards from European Commission, i.e. Euro 6, has additionally placed a limit on number of particles greater than 100 nm to address growing concerns for severe health effects associated with finer particles (PM2.5). Emission standards also regulate various sources of air pollution, aside from automotive sources, and together they have been effective in improving the ambient air quality with nearly 39%, and 56% reductions achieved in ambient PM10 (24 hour), and NOx (annual) concentrations, respectively during the years 1990 to 2016 [14] in US. At the same time, it has also necessitated automotive manufacturers to continuously innovate new emission reduction strategies.  1.3 Emission Reduction in CI Engines 1.3.1 Existing Systems and Strategies  Manufacturers have used a combination of in-cylinder and exhaust aftertreatment strategies to meet the US EPA 2007 and later standards [12]. A popular in-cylinder strategy is exhaust gas recirculation (EGR), which reduces thermal NOx by reducing combustion temperatures. Although EGR was sufficient to meet the EPA 2007 NOx requirements for most manufacturers, 2010 and later standards required additional NOx exhaust aftertreatment systems like selective catalytic reduction (SCR) [12]. Other exhaust aftertreatment systems that have become a commonplace in heavy duty diesel engines since US EPA 2007 regulations, include the diesel particulate filter (DPF) for the filtering of PM, and the diesel oxidation catalyst (DOC) for the oxidation of CO, hydrocarbons, non-solid diesel particulates as well as non-regulation emissions such as aldehydes or PAHs [15].  4  Aftertreatment systems are incredibly effective (e.g. NOx reduction ~ 70-98% with SCR [16]; PM reduction in excess of 90% with DPF [17]) in reducing emissions; however, they may impose significant fuel penalty (e.g. via regeneration in DPFs [17]), or require the consumption of an additional catalytic fluid (e.g. NOx reduction by SCR requires). Additionally, they add restriction to the exhaust stream, and thereby increase pumping work (additional work required to expel exhaust gases). Although aftertreatment with combustion optimization may result in net improvement in system efficiency [18], significant issues of system complexity, capital cost still call for research and development of alternate strategies to achieve further emissions reduction.  1.3.2 Natural Gas Fuelling: HPDI A popular alternative fuel to diesel is natural gas. Natural gas provides numerous environmental and economic benefits, which arises mainly because of its primary constituent hydro-carbon: CH4 (87-97% CH4, [18]). CH4 by virtue of being the simplest hydrocarbon, possesses the least carbon-to-hydrogen ratio, and thereby results in the least energy specific CO2 emissions. The lack of carbon-carbon molecular bonds results in reduced probability of benzene ring formation during combustion, which further results in reduced formation of carcinogenic polycyclic aromatic hydrocarbons (PAH) and soot [19]. Relative to diesel and gasoline, natural gas also burns with lower adiabatic flame temperatures, and thereby results in lower thermal NOx. From a commercial perspective, natural gas reserves are available globally, with relatively lower price and volatility than other fossil fuels; however, being a gaseous fuel at standard conditions, natural gas brings infrastructural and storage challenges.   A technology to facilitate late cycle, direct injected natural gas in diesel engines is High Pressure Direct Injection (HPDI) [20]. It uses a small pilot injection of diesel fuel (5-10%) to provide increased localized temperatures as well as necessary radicals for combustion of the direct 5  injected primary fuel (NG), due to which this combustion strategy is also referred as pilot ignited direct injection of natural gas (PIDING). HPDI, therefore, allows a diesel engine to combust a relatively cleaner fossil fuel- NG with its inherent benefits while retaining near diesel engine like efficiencies [20] (Consult [21] and [22] for further details on HDPI). Further work, however, is needed for reducing emissions to the very low levels dictated by current standards while minimizing requirements and usage of complex after treatments systems.  1.4 Air Addition in PIDING Engines A potential in-cylinder strategy to improve PIDING engines further is fuel dilution. In an N2 addition study in a PIDING engine, diluting the primary fuel- natural gas with N2, resulted in substantial reductions in emissions (PM, CH4, THC, and CO) while achieving slight improvements in fuel economy [23]. The typical NOx-soot trade-off was not seen. Since, diluting the fuel with air (air addition) would be a more practical concept, an O2 addition concept and it’s relevant apparatus was conceived and developed by Westport Fuel Systems, leading to a patent [24]. Subsequently, a research configuration based on the patent was developed, and used for conducting screening tests with N2 addition and air addition in a PIDING engine, in various previous projects supported by Westport fuel systems. The results from the N2 addition reported similar qualitative trends compared to previous results by McTaggart-Cowan et al. [23]. The air addition results were inconclusive due to significant measurement errors but indicated the possibility of a significant NOx penalty. One major objective of this study is to characterize air addition in a relevant heavy-duty engine operating mode and the results and discussion are presented in Chapter 4.  There are major challenges that could potentially limit the applicability of air addition in an on-board PIDING engine: power consumption and sizing of the diluent compressor. Since the primary fuel-natural gas in a PIDING engine is at very high pressures, the diluent (air) would also 6  require to be at similar pressures (>300 bar) for mixing with the fuel. The most appropriate compressor type, meeting such high diluent delivery pressure requirements with sufficient flowrates and reasonable sizing, would be the reciprocating compressor. Therefore, the power consumption and the flowrate of a 3-stage reciprocating compressor system were characterized in Chapter 5. A prototype 2-stage compressor configuration was also developed during this study to reduce the sizing further and was similarly characterized in the same chapter. To prevent thermal and mechanical failures, the total pressure ratio of this 2-stage compressor was reduced by supplying its inlet with a compressed air source having engine relevant pressures, which in this study were taken as near typical intake manifold pressures. Finally, to weigh the benefits of air addition against the parasitic load from the diluent compression, a system level study was conducted, and its results and discussion presented in Chapter 6.  1.5 Objectives The overall objective of the project was to investigate the concept of air addition in a PIDING engine and determine its system implications by evaluation and considering the compression load of suitable diluent compressors required for this concept. Emphasis was placed on having a mechanistic understanding of observed behaviors. The objectives were achieved predominantly through experimental observations, literature review, and by developing mathematical models and formulations.    Specific project objectives for investigating air addition include:  • Evaluate air addition on a realistic heavy duty operating mode on a PIDING engine over a range of exhaust gas recirculation (EGR) rates and global oxygen equivalence ratios (Φ): effects of air addition on emissions and performance. 7  • Achieve basic understanding of pollutant forming behaviors via observations of combustion processes based on in-cylinder pressure measurements and available literature in diesel and PIDING engines. Specific objectives associated for investigating the system implications of air addition include:  • Characterize the flowrate, compressor motor power, and performance of 2-stage and 3-stage compressor configurations. Analyze their performance relative to that in literature.  • Determine the system efficiency of air addition by considering the compression power of considered compressor configurations in the gross indicated efficiency of the engine. • Provide an optimization approach by considering both the system level efficiency and the target emission reductions of PM and NOx, and thereby recommend optimum engine operation zones.  1.6 Thesis Structure The current chapter (Chapter 1) provides an introduction and motivation to the research. Chapter 2 provides the relevant background required for understanding of results and discussion in Chapter 4 and Chapter 5: for pollutant formation mechanisms and for performance evaluation of diluent compressors. Chapter 3 describes the experimental system, defines various parameters, and describes the uncertainty analysis approach used in subsequent chapters. The chapter also provides a brief overview of the development of a prototype diluent compressor of reduced sizing i.e. 2-stage compressor. Chapter 4 presents and discusses the results of air addition characterization, while Chapter 5 discusses the results of characterization of diluent compression systems. Chapter 6 uses the results and discussion of Chapter 4 and Chapter 5 together, to discuss the system level implications of air addition by considering parasitic diluent compression loads from various compressor configurations. The same chapter also discusses an optimization 8  approach and proposes an optimum operating space within the tested space. The final conclusions of this study are summarized in Chapter 7. 9   Background and Theory The use of Westport fuel systems HPDI technology in PIDING engines has been shown to achieve PM reduction by nearly 70% and NOx reduction by nearly 45% over conventional diesel fuelling systems of 2003 [25]. For further performance improvements as well as to achieve better emission control, various in-cylinder strategies were pursued over the years. Their effects on emissions and performance are discussed in Section 2.1.1. Fuel dilution showed the potential to overcome many challenges faced by these strategies, and therefore is the primary reason for the pursuit of air addition in this study. The background relevant for understanding the effects of air addition on pollutants (discussed in Section 4.3) is provided in Section 2.1.3. A major objective of this study is to evaluate the system implications of air addition, by further considering the compression loads of suitable compressors required to execute this strategy. Since this is expected to depend on the compressor performance, the relevant background for performance assessment of reciprocating compressors is discussed in Section 2.2.   2.1 Air Addition 2.1.1 In-cylinder Strategies in PIDING Engines Table 2.1 lists various in-cylinder strategies, similar to that in diesel engines, which have been characterized over the years in PIDING engines with objectives of reducing emissions and improving performance. Although all the strategies, except late post injection (LPI), resulted in a NOx-penalty while reducing PM emissions, the NOx-penalty in certain cases was significantly reduced by operating in ultra-high EGR. Other common drawbacks encountered while reducing PM include increased CH4 emissions, maximum pressure rise rates, combustion variabilities, and loss in fuel economy. However, diluting fuel with N2, as is discussed below, was found to be free 10  of these limitations [23], and therefore fuel dilution with air, a more practical concept, was pursued in this study. Effect of fuel dilution with N2 at various combustion timings with 30% EGR in mode B75 (speed= 1500 rpm; high load= 75% of peak load) was investigated by McTaggart-Cowan et al. [23]. This concept achieved up to 60% reduction in both CO and undenuded PM (primarily via the soot reduction) emissions at intermediate combustion timings (at 10oATDC, 15oATDC), with slight reductions in NOx. Performance benefits include improved fuel economy (GISFC reductions ~4%), enhanced combustion stability (via reduced cycle-cycle variabilities), and reduced combustion intensity (reduction of peak heat release rates). Although this strategy was not found to be favorable at the earliest combustion timing (i.e. 0oATDC), the uniqueness of this strategy lies in the fact that in all other timings the emission reductions were obtained without typical trade-offs of NOx and fuel economy.    11   Table 2.1 Effects of various in-cylinder strategies of PIDING engines on emissions and performance Note: ↓: reduced/ reduction/ decrease; ↑: increased/ increase; GRP: gas rail pressure or injection pressure; GISFC: gross indicated specific fuel consumption ;  RIT: relative injection timing; IHR50: represents combustion timing as defined in Section 3.1.3; mode B75: mode with load= 75% of nominal and maximum and speed= 1500 rpm; PSEP: pilot separation as shown in Figure 4.1 Ref. # In-cylinder strategy  Desirable effects of the strategy on emissions and performance Limitations [26]   GRP (21, 30 MPa) at various modes (load-speed combination) Significant PM ↓ only at high loads ↑ NOx, max pressure rise rate, and heat release rate. PM reduction was load dependent [27] Global oxygen equivalence ratio (Φ) at mode B75 ↓ Φ resulted in ↓ PM, CO, hydrocarbons, and GISFC; PM and GISFC was most sensitive to Φ, amongst other considered parameters of injection timing, injection pressure, pilot mass, and load ↓ Φ results in ↑ NOx, ↑ peak in-cylinder pressure [28] EGR rate (0 to 50%) at various engine modes and injection timings Over 80% reduction in NOx emissions achievable. NOx ↓ linearly with inlet O2 mass fraction, till NOx reached 20% of non-EGR levels.  ↑ PM, CO, and hydro-carbons. ↑ GISFC and combustion variabilities at high EGRs (> 30%) [29] GRP (17 to 25 MPa), IHR50 (0o, 10o, and 20oATDC), RIT (-0.4, 0.2, and 1.8 ms) over a range of EGR rates (0 to 50%) ↑ GRP resulted in ↓ PM over all EGR, with most significant reductions at highest EGR rates; ↑ NOx insignificant at > 40% EGR. Advanced IHR50 reduced PM, CO, and THC trade-offs from EGR, with ↑ reductions at > 30% EGR; GISFC improved as well.  Shortening of RIT from short positive to negative RIT mitigated the PM trade-off from EGR; GISFC improved as well.  At > 30% EGR, ↑ GRP resulted in ↑ PM and CO. Advanced IHR50 significantly ↑ max pressure, and its rise rates. Shortening of RIT resulted in ↑ CH4 and NOx, although NOx penalty was insignificant at highest EGR rates. [22] [30] RIT (via PSEP variation) at various Φ (0.6; 0.7) and EGR (18%; 25%) for mode B75 ↑ Φ eliminated the CH4 penalty, which typically results from RIT shortening. For negative RIT, PM and CO were independent of both Φ and EGR. Suggested optimum point based on max PM reduction (~90%) and near baseline NOx and CH4, had improved fuel economy by 2% Recommended optimum point had drawbacks of cycle to cycle variability and increased max pressure rise rates [31] [22] Late Post Injection (LPI) optimized for mode B75  Reductions of PM, CO, and CH4 specific emissions by nearly 70%, 70%, and 20%, respectively by optimization of both the timing and the quantity of fuel injected in the second injection pulse. No NOx trade-offs. The robustness of the strategy was demonstrated at two other modes Minor fuel consumption penalty (<1% ↑ in GISFC) at the recommended optimum point   12   2.1.2 Conceptual Combustion Although a conceptual model of PIDING combustion, like that for diesel by Dec [32], doesn’t exist, a working conceptual model can be outlined from the insights provided by CFD [33–36] and  years of engine testing (e.g. [21,22,32,37]). Ongoing optical diagnostics by Rochussen et al. [38] aims to develop a more conceptual model of PIDING combustion by characterizing combustion mechanisms in this working model. The working model is briefly described below.  In conventional PIDING operation, the pilot fuel (diesel) is injected first, followed by the injection of natural gas, which at high loads is typically around the start of diesel combustion or slightly prior to it [22]. As the relatively colder and under-expanded gas jet adjusts to combustion chamber conditions via a barrel shock [36], the momentum of the injected fluid provides the driving force to penetrate into the combustion chamber charge while entraining oxidizer along the way [21]. Fuel concentration gradients ranging from pure fuel at the jet core to leaner mixtures on the outer peripheries of the jet, are formed because of oxidizer entrainment prior to ignition [21]. Combustion occurs when the gas jet interacts with the hot products of pilot combustion, which based on a recent imaging study, is indicated to start along premixed peripheries of the jet’s head followed by the sides of the quasi-steady jet [39].  Following the consumption of the premixed mixture, the flame stabilizes to a lifted turbulent diffusive jet flame, where the reaction zone is expected to have a near stoichiometric fuel-air mixture, similar to diesel diffusive burning [21]. Following the end of NG injection, the momentum transfer to the separated ‘puff’ jet diminishes, while mixing and burning continues till the reactions are unsustainable: due to insufficient fuel, oxidizer or temperature [21]. Combustion can be influenced significantly by various injection strategies, resulting in significant effects in emissions and performance (e.g. too low injection pressures can result in 13  slower and incomplete combustion. The next section provides a brief overview of how pollutant formation is affected by combustion.   2.1.3 Pollutant formation This section provides a brief background on the formation mechanisms of various pollutants of concern in a PIDING engine. Air addition, similar to N2 addition, in a PIDING engines is expected to affect the pollutant formation behavior through a dilution effect and increased turbulence intensity effect (kinetic energy from the added diluent mass) [23], and therefore a brief overview on the effects of fuel dilution on the PM and NOx forming behaviors, observed in various fundamental methane diffusion flame studies are also provided. While the applicability of these results to PIDING engines is unclear due to possible influence of varied fluid dynamics, these results illuminate potential effects of varying physics (from various in-cylinder strategies) on the pollutant formation mechanisms. The key results and pollutant formation mechanisms from an N2 addition study in PIDING engines [21,23], which are more relevant to the current study, will be discussed in context with effects of air addition in Chapter 4, and therefore is not discussed here.  2.1.3.1 Particulate Matter Solid carbon (soot) is the major constituent of particulate matter (PM) at high-loads in both direct-injected diesel engines [40] and PIDING engines [27]. The processes involved in soot formation and the effects of various physics on the soot formation in non-premixed methane flames are discussed below.  As of now, six processes have been identified which constitute soot formation mechanism: pyrolysis, nucleation, surface growth, coalescence, agglomeration, and oxidation [19,41]. Pyrolysis involves the breakdown of fuel in a high temperature oxygen deficient environment to 14  produce soot precursors or the basic building blocks for soot [42].  In case of methane combustion, a major soot precursor acetylene (C2H2) is formed when methyl radicals (CH3) react to form C2H5 or C2H6 molecules and then undergoes successive hydrogen abstraction [41]. Majority of the C2H2 formed undergoes oxidation to form CO, while some react to form benzene (C6H6), a constituent of the base ring structure of polycyclic aromatic hydrocarbons (PAHs), and thereby starts the growth of soot particles [41]. Nucleation or the soot particle inception stage, converts the gaseous precursors  to first particles or nuclei in the order of 1-2 nm [19,41]. Mass is added to the surface of nucleated particle in the surface growth phase. Coalescence phase involves particles sticking with each other to form closed structures, and thereby achieve high rates of growth [19,41]. Agglomeration is identified as the last phase of soot growth, where non-coalescing particles join with each other in open fractal-like structures. Soot particles at any stage of the formation process (from pyrolysis through agglomeration), can be oxidized as long as the environment has oxidizing species (e.g. O, OH , and O2) and sufficiently high temperature [19].  Effect of fuel dilution on the soot formation of fundamental non-premixed methane flames has been widely investigated. Addition of N2 to co-flow laminar methane diffusion flame resulted in a proportional decrease of peak soot volume fraction proportionally due to a dilution effect resulting from reduced carbon per unit mass [43]. In the same study by Gulder [43], O2 addition to methane reported an additional chemical effect (due to added O2 species participating in chemical reactions), which resulted in the reduction of C2H2 (a major soot precursor in methane combustion [41]) concentration, and thereby chemically suppressed soot formation [43]. A similar study by Cao et al. [44] also reported the peak soot volume fraction to decrease with N2 addition to co-flow laminar methane diffusion flame in pressures up to 2.7 atm. A numerical investigation of N2 addition to co-flow methane-air laminar diffusion flames on the soot volume fraction, with 15  a one-step soot model by Tian et al. [45], revealed the dilution effect to be comparable with thermal effect, which is due to change in flame temperature field with diluent addition, and residence time effects.  2.1.3.2 Oxides of Nitrogen As is identified in literature, nitric oxides (NOx: NO and NO2) are formed primarily via 4 different mechanisms or routes: thermal (Zeldovich) route, prompt (Fenimore) route, nitrous oxide (N2O) route, and the fuel-bound route. Its formation in the context of internal combustion engine conditions are well understood and are detailed in [19,46]. Nitrogen oxide (NO), which forms a major fraction of NOx, is primarily formed with the thermal route [19,46]. The rate limiting step of this mechanism requires high activation energy, due to which thermal route is highly sensitive to temperature, and typically requires> 1700 K for activation. Additionally, the slow nature of this mechanism requires sufficiently long residence times to reach equilibrium, and therefore may not achieve equilibrium in the turbulent mixing environment of diesel engines. Regardless, it is the dominant mechanism for most (exception: low temperature combustion strategies) engine relevant combustion. The second route is prompt (Fenimore NO) [19,46], which primarily contributes to formation of NO in the flame front or on the richer side of a diffusion flame. This is because the rate limiting step of this mechanism requires CH radicals or its precursors (like C2H2), which are only formed at these locations. Since activation energy required for prompt NO is much lower, prompt NO can form in combustion environments with temperatures as low as 1000 K (e.g. at very high EGR). N2O route involves the reaction of N2 with oxygen radical in presence of a third body to form NO. This reaction is similar to thermal route, although the additional presence of the third body in this route allows NO formation at even 1000 K. This mechanism, otherwise insignificant to NO formation, can dominate when combustion conditions don’t favor prompt and thermal NO, 16  such as lean flames and low temperatures (e.g. negative RIT with low Φ). The fourth NO source is fuel-bound N, which becomes significant in the combustion of fuel containing significant amounts of chemically bound nitrogen, such as ammonia and coal. The significant amounts of molecular nitrogen (N2) found in natural gas or in natural gas diluted with N2 are unlikely to contribute to NO from fuel-bound N [21]. Changes in adiabatic flame temperature (AFT) of diffusion flames may act as a good indicator of NOx variations of the thermal route. This is because: (1) AFT as characteristic flame temperature in regions were radiation effects are weak, has been justified in previous studies by Gomez and Glassman [47] and by Axelbaum and Law [48] (2) flame temperature is good indicator of the typically dominant, thermal NOx route. Gulder [43] provided further support for (1) by demonstrating a linear relationship between the measured flame temperatures and calculated AFT for both co-flow and counter-flow diffusion flames of a variety of non-methane hydrocarbon fuels. A PIDING engine study across a range of operating conditions (load, speed, combustion timing, and intake oxygen fractions) by Hill and McTaggart-Cowan [49] demonstrated (2) by correlating NOx emissions to intake oxygen fraction, which in turn was shown to have a linear relationship with the reciprocal of AFT. However, deviations in the correlation between exhaust NOx emissions and calculated AFT have been noted in diesel engines, where a major cause was attributed to soot radiation affecting the flame temperature-AFT relationship [50]. The influence of fuel dilution, and soot radiation on the thermal NOx emissions was explored in many previous methane diffusion flame studies. A numerical study on flameless combustion of counterflow methane flames by Mohamed et al. [51] reported a decrease of NOx emissions with dilution of methane with N2 due to reduced fuel supply to flame zone, which would decrease both overall rate of chemistry as well as AFT [51]. The reduction of NO with N2 addition 17  in co-flow laminar methane diffusion flames was also seen in another study by Feese and Turns [52], where it was attributed to reduction in peak flame temperatures. This study also demonstrated that the effectiveness of N2 addition on NOx reduction may vary with the sooting nature of the flame. A sooting flame would result in reduced soot with N2 addition, which in turn would result in reduced radiation losses [52]. This would not allow N2 addition to reduce the flame temperature effectively, and therefore would reduce the effectiveness of NOx reduction. This effect of soot radiation on the NOx emissions has also been noted in both fundamental diffusive flames (e.g. [53,54]) as well as in turbulent diffusive combustion environment of diesel engines [50]. Therefore, a possibility of high soot reductions resulting in a corresponding NOx increase effect exists in PIDING engines.  2.1.3.3 Total Unburnt Hydrocarbons There are two major causes of total unburnt hydrocarbons (THC) in direct injection compression ignition engines: (1) overleaning i.e. fuel mixed to leaner than lower flammability limit, such as during ignition delay (2) undermixing of fuel [46,55]. An overlean mixture is not able to sustain a fast reaction front or to autoignite. Although part of this mixture may participate in combustion from the oxidizer side, some may escape combustion by entering crevices inside combustion chamber. A major source in case of undermixing of fuel is fuel retained in injector sac and nozzle holes, which enters the combustion chamber as the pressure drops towards the end of expansion stroke, and thereby escape combustion [46,55].  Other causes of unburnt hydrocarbons include localized flame extinction and bulk quenching. High turbulent strain rate may cause flames to locally extinct, some of which may never reignite and end up being emitted [19]. Another cause of localized flame extinction is quenching at crevices and walls due to high heat transfer losses [19,46]. Bulk quenching occurs 18  towards the end of the combustion event, when the temperature and pressure suddenly drops due to expansion stroke of piston, and thereby cause the combustion reactions to be unsustainable [46]. This process can be minimized via enhanced late-combustion burn as was achieved with nitrogen dilution in [21,23].  The major constituent of unburnt hydrocarbon emissions in PIDING engines is methane, which form > 95% of total unburnt hydro-carbons (THC) [21,22,27]. 2.1.3.4 Carbon monoxide Traditionally, carbon monoxide (CO) was not a significant concern for direct injection CI engines in regular operation [21,46,55]; however, under extreme operating conditions, such as that with very high EGR (> 40%) in PIDING engines [29] as well as current stringent emission regulations, CO emissions can become a significant concern. CO emissions are a result of incomplete oxidation to CO2, which in overall lean systems (e.g. DI diesel engines, PIDING engines) can be due to insufficient fuel-air mixing, failing to provide sufficient O2 to the reaction zone [21,46]. Reduction of temperatures in the reaction zone, such as that with bulk quenching, can also stop the oxidation of CO [46].   2.2 Diluent Compression  The background relevant for the performance assessment of suitable compressors (i.e. reciprocating compressors) for the air addition concept, is discussed in this section. Performance assessment of the diluent compressor (discussed in Chapter 5) used in this study was carried out by evaluating various compressor relevant efficiencies. Their definitions, theory, and the typical range of reciprocating compressor efficiencies observed in various applications are discussed below.    19  2.2.1 Compression Efficiencies Definitions of compression efficiency vary depending on what constitutes the compression system. A common definition, based on a per stage (of compressor) basis, is given below [56,57]: Compression efficiency =  ?̇?𝑖𝑑𝑒𝑎𝑙?̇? where ?̇?𝑖𝑑𝑒𝑎𝑙 is the rate of work consumed by a reversible compression cycle, having an isobaric suction and discharge process, and ?̇? is the rate of the measured indicated work. The reference compression and expansion processes can together be chosen as any of isothermal, isentropic or a polytropic process with polytropic index n ∈ (1,1.4), in which case their compression efficiencies are referred to as isothermal (𝜂𝑖𝑠𝑜), isentropic (𝜂𝑖𝑠𝑒𝑛), and polytropic (𝜂𝑝𝑜𝑙𝑦), respectively. They are described by Equation (2.1) given in [57] and [58], Equation (2.2) given in [58], and Equation (2.3) given in [57]:   𝜂𝑖𝑠𝑒𝑛 =?̇?(ℎ2,𝑖𝑠𝑒𝑛 − ℎ1)?̇?=  ?̇? ∫ 𝐶𝑝𝑑𝑇𝑇1(𝑃2𝑃1)𝛾−1𝛾𝑇1?̇? (2.1)   𝜂𝑝𝑜𝑙𝑦 =?̇?𝑛𝑛 − 1 𝑅𝑇1 ((𝑝2𝑝1)𝑛−1𝑛− 1)?̇? (2.2)  𝜂𝑖𝑠𝑜 =?̇?𝑅𝑇1 (log (𝑝2𝑝1))?̇? (2.3) where ?̇?, (ℎ2,𝑖𝑠𝑒𝑛 − ℎ1), 𝐶𝑝, and R are the compressor outlet mass flowrate, calculated change in specific isentropic enthalpy based on measured inlet (suction) temperature (𝑇1) and stage pressure ratio (𝑃2𝑃1) for air, the specific heat for air at constant pressure, and the specific gas constant, respectively.  20  The compression power in an ideal compression cycle increases with decreasing wall heat transfer to ambient of the chosen reference process, resulting in least power consumption in compressor with isothermal stage walls, while maximum power consumption with adiabatic stage walls [56,57,59]. This can be understood by calculating the  compression power  of the various reference processes (represented by numerator of Equation (2.1),(2.2), and (2.3)) for the same parameters of ?̇?, 𝑇1, and 𝑃2𝑃1. Figure 2.1, elucidates graphically that the p-V area (i.e. the boundary work done on the fluid) enclosed by an adiabatic compression process is more than that by isothermal. This results in multiple stages inherently leading to more power savings, since it makes the overall compression process tend more towards isothermal compression [59].   Figure 2.1 Schematic of typical compression cycle showing various losses  The suitability of the chosen reference process is dependent on the wall heat transfer in the actual compressor [57,60]. A reference process with wall heat transfer closer to that of the actual compressor makes the definition of compression efficiency more valid by highlighting the 21  irreversibilities (i.e. the difference between actual and ideal work would then majorly be the irreversibilities in the process). This also ensures that compression efficiency remains less than 1, which may not always be the case. Therefore, a compressor with high heat transfer from wall to ambient would best be represented by an isothermal reference process, while those with low values of the same, would be best represented by an isentropic reference process. Typically, practical compression systems have wall heat transfer between isothermal and adiabatic, and therefore makes polytropic reversible process a very suitable candidate for reference process.   Polytropic reference process requires determination of the polytropic exponent (n). An accurate determination of n requires in-cylinder pressure and volume data, which requires intrusive modifications of the compressor stages for measurement of crank resolved in cylinder pressure and volume data. The negative slope of the linear fitted line of the compression process in the log p-log V graph, would then be the compression polytropic exponent. An easier method, typically used for hermetic compressors [59,61,62], only requires the measured stage pressure and temperature ratio, and is described by Equation (2.4) below:   𝑛 =log (𝑝2𝑝1)log {(𝑝2𝑝1)(𝑇2𝑇1)} (2.4) Although the accuracy of this method, and its sensitivity to the thermocouples used is unclear, this method was used for preliminary calculations of n in this study due to its ease.  In reciprocating compressors, 𝜂𝑖𝑠𝑒𝑛 determination is typically preferred over 𝜂𝑝𝑜𝑙𝑦 due to reasons stated below. For refrigeration compressors, the polytropic exponent is found to be highly refrigerant dependent [58], making it an inconsistent reference process for the evaluation of the 22  same compressor with different refrigerants. 𝜂𝑖𝑠𝑒𝑛 provides a consistent and universally accepted standard of compressor stage efficiency analysis, since typically the wall heat transfer to ambient is not significant enough to reduce compression power, such that the power savings are more than power consumed from losses (which would result in 𝜂𝑖𝑠𝑒𝑛 > 1). It is acknowledged in [57,63], that typical industrial reciprocating compressor processes are far from isothermal, though various research work has been undertaken to develop isothermal compressors and save power consumption ([64]).  𝜂𝑖𝑠𝑒𝑛 of reciprocating compressors in various applications has been reported in literature and are collectively tabulated in Table 2.1. Although 𝜂𝑖𝑠𝑒𝑛 varying in similar ranges were noted for both large (petrochemical compressors [57] ) and small compressors ([65,66]), 𝜂𝑖𝑠𝑒𝑛 of twin cylinder compressor, having displacement volume (𝑉𝑠) comparable to the one used in this study, was relatively lower. These values are much lower than unity due to significant irreversibilities from valve pressure losses. An ideal compression process assumes isobaric intake and exhaust processes, which would require the valve diameters to be equal to bore diameter, and therefore is not possible in a realistic compressor. The resulting differential pressure build up causes under-pressure and overpressure in the suction and exhaust stroke, respectively, and are illustrated in Figure 2.1.  𝜂𝑖𝑠𝑒𝑛 also depends on compressor operating conditions, primarily the stage pressure ratio, increasing which in [66] resulted in increased 𝜂𝑖𝑠𝑒𝑛 (by 8%-point) at its lower values (<10) before stabilizing to a constant value. Although the author couldn’t locate any study detailing the effects of inlet pressure on 𝜂𝑖𝑠𝑒𝑛, [57] comments that reduced inlet pressures can result up to a decrease of 20%-point in overall efficiency (𝜂𝑂𝐸 = 𝜂𝑖𝑠𝑒𝑛. 𝜂𝑚𝑒𝑐ℎ). However, it is not clear which of the two constituent parameters contribute to its reduction.  23  2.2.2 Mechanical Efficiency Mechanical efficiency (𝜂𝑚𝑒𝑐ℎ =𝐶𝑜𝑚𝑝𝑟𝑒𝑠𝑠𝑜𝑟 𝐼𝑛𝑑𝑖𝑐𝑎𝑡𝑒𝑑 𝑃𝑜𝑤𝑒𝑟𝐶𝑜𝑚𝑝𝑟𝑒𝑠𝑠𝑜𝑟 𝑆ℎ𝑎𝑓𝑡 𝑃𝑜𝑤𝑒𝑟) of a compressor measures the effectiveness of converting input shaft power to indicated power of the compressor stages. 𝜂𝑚𝑒𝑐ℎ of a reciprocating compressors, similar to reciprocating engines, is a major function of the ratio of friction power dissipated from the machinery parts (e.g. bearings, piston rings [67]) to the ratio of shaft power [56]. Compressors in the opposite ends of size and speed spectrum, showed 𝜂𝑚𝑒𝑐ℎ to be in near similar ranges i.e. 0.85 to 0.95 (Table 2.1). This is interesting, since an influence of operating speed on 𝜂𝑚𝑒𝑐ℎ was expected. 𝜂𝑚𝑒𝑐ℎ also varied with stage pressure ratio, increasing which in [66] resulted in an initial increase of 𝜂𝑚𝑒𝑐ℎ by 3%-point at lower pressure ratio (<10) before near-linearly decreasing with further increase. The author couldn’t locate any study reporting the effects of inlet pressure on 𝜂𝑚𝑒𝑐ℎ.          24    Table 2.2 Relevant specifications  and efficiencies  of  reciprocating compressors in various applications Note: 𝑁: compressor shaft speed; 𝜂𝑖𝑠𝑒𝑛: isentropic efficiency; 𝜂𝑚𝑒𝑐h: mechanical efficiency; 𝜂𝑂𝐸: overall efficiency; 𝑉𝑠: swept volume; 𝑉𝑠,𝑡: swept volume of first stage of compressor used in this study = 255 cm3; 𝑉𝑐𝑉𝑠: clearance fraction i.e. ratio of clearance volume to swept volume of a compressor stage Ref. # 𝑁 (𝑟𝑝𝑚) (𝑃2𝑃1) 𝜂𝑖𝑠𝑒𝑛 % 𝜂𝑚𝑒𝑐h % 𝜂𝑂𝐸 = 𝜂𝑖𝑠𝑒𝑛. 𝜂𝑚𝑒𝑐h % # of stages 𝑉𝑠 (× 𝑉𝑠,𝑡) 𝑉𝑐𝑉𝑠%  Type of Reciprocating compressor [59]  -  -  - 88-95  - 1 - 4-16 Data from Ingersoll Rand Reciprocating compressors [56] 450-650 - - 90-93  - 1, 2 > 10𝑉𝑠,𝑡 5-15 Petrochemical compressors [57] > 900 >4 73-79 95 70-75 1, 2 > 10𝑉𝑠,𝑡 - Petrochemical compressors   [65]  -  - 80-83 92 <  - 1 < 0.1𝑉𝑠,𝑡 - Household refrigeration  [66] 3000 5-15 70-73 92-95 64-69 1 0.03𝑉𝑠,𝑡 0.77, 2.01 Hermetic refrigeration [68]   950 5-10 56-57 -   - 1 2.3𝑉𝑠,𝑡 4.5 Twin cylinder compressor  25  2.2.3 Transmission Efficiency  The transmission efficiency quantifies losses from various subsystems transferring the compressor input power to the compressor shaft. For converting the input power (electric or fuel) to torque required to drive the compressor, a suitable driver is recommended based on the size of the compressor [57], which for this study is a 7.5 hp induction motor. The other subsystem transfers driver shaft power to the compressor shaft, which for this study is a V-belt. The losses in transmission power from each of these subsystems are characterized by their respective efficiencies as discussed below.  Motor efficiency (𝜂𝑚𝑜𝑡𝑜𝑟) characterizes the power lost in the transfer of input power to shaft power [69,70] of a motor. In acceptable operating region (50 to 100% of full load), it is reported to vary in the range 82 to 94%, depending on the size, synchronous speed, and the operating load of the motor [70]. The load dependence of 𝜂𝑚𝑜𝑡𝑜𝑟 is significant. While higher loads typically retain a constant 𝜂𝑚𝑜𝑡𝑜𝑟 (nearly equal to nameplate value at 3/4th of rated power), lower than 50% load may result in an exponential decrease of 𝜂𝑚𝑜𝑡𝑜𝑟 [70]. This suggests that an estimation of motor shaft power from measured motor input power at lower loads may require 𝜂𝑚𝑜𝑡𝑜𝑟 as a function of load to be known, for example, from manufacturer or measurements. [70] lists several methods to determine 𝜂𝑚𝑜𝑡𝑜𝑟 experimentally, one of which is used in this study to validate the nameplate motor efficiency: the slip method.   Belt efficiency (𝜂𝑏𝑒𝑙𝑡) characterizes the transmission losses via a belt. [71] provides an exhaustive list of experimentally measured industrial belt efficiencies varying with pitch diameter, type of belt, range of torque and speed. The belt used in this study was a V-type wrapped belt, which based on [71] should have 𝜂𝑏𝑒𝑙𝑡 in the range 90 to 95%. 26  2.3 Summary / Literature Gap  A prior study had investigated fuel dilution with N2 in a PIDING engine at various combustion timings. Similar effects are also expected with air addition, however, the O2 species present in added air may contribute to an additional chemical effect, which may affect emissions and performance differently. Although effects of fuel dilution (with both O2 and N2) on the pollutants (NOx and soot) have been investigated in fundamental non-premixed methane combustion studies, the applicability of these results to the combustion environment of a PIDING engine is not clear. Since the magnitude of these effects affect the feasibility of the air addition for on-board commercial applications, an independent investigation of the effects of air addition in a PIDING engine is warranted. This study also considers variation of EGR rate and Φ, parameters which significantly affect NOx and PM, to further elucidate their interaction effects with air addition, while also indicating regions of optimum operation. Another major consideration in determining the feasibility of air addition, is the performance of the diluent compressor (reciprocating). The vast majority of reciprocating compressor research are for refrigeration and petrochemical compressors, and as such the author couldn’t locate sufficient literature on the performance parameters of compressors with sizing relevant for air addition application. Furthermore, existing research didn’t explore areas relevant to current study, for example, effects of inlet pressures on compressor efficiencies; evaluation of relevant performance parameter: specific motor power consumption; performance of compressors at greater than 10 stage pressure ratios; empirical relations for ball park estimation of friction. This study, therefore, experimentally explores these areas in a suitable diluent compressor in Chapter 5, with the aid of existing theory and research presented in this chapter. 27    Experimental System  The air addition concept was characterized by using a single cylinder research engine (SCRE) and an Industrial reciprocating compressor. The development and implementation of this facility is discussed below.  3.1 Single Cylinder Research Engine (SCRE) A single cylinder research engine (SCRE) modified from a six-cylinder Cummins ISX 450 series 4 stroke diesel engine was used for characterizing air addition. Previous studies using the SCRE, have demonstrated performance very similar to that of an equivalent multi cylinder research engine [72]. It also allows a wide range of possible test conditions making it a very useful research tool. Specifications for the engine are provided in Table 3.1:     Table 3.1 Key specifications of Single cylinder research engine Parameter Value Compression ratio 17 Clearance volume (m3) 1.55×10-4  Displacement volume (m3) 24.91×10-4 Bore (m) 0.137 Connecting rod length (m) 0.2615 Crank radius (m) 0.0845  Rated speed (RPM) 1800 Crank encoder offset 3o BTDC  28   Figure 3.1 Single cylinder research engine   The SCRE running at the highest loads is able to produce enough power to overcome frictional losses from its non-firing cylinders; however, lower loads require an additional torque to continue operating. The facility therefore used an external 30 kW electric motor (Baldor) controlled by a vector drive to provide additional constant torque (more than required to overcome engine friction). A water-cooled eddy-current dynamometer controlled by a Digilog controller, was mounted in series to the vector drive motor to absorb this additional torque. It also absorbed any fluctuations in torque levels to maintain a constant engine speed with 1% of setting. The engine speed was measured with a flywheel encoder as well as an optical encoder.  For operating the engine with boost air, compressed air (100-120 psi) was supplied by an Ingersoll-Rand rotary screw compressor. The drier used in conjunction with the compressor was not in full operation during the study, and therefore a relative humidity of 20-30% is expected in the supplied air. Variations in pressure were reduced by a buffer tank, while delivery pressure was maintained using a manual pressure regulator and a pneumatic regulator. Using this system, the fluctuations were ±2 kPag, measured upstream of the intake surge tank. The intake surge tank SCRE Intake  Exhaust DAQ EGR valve 29  damps the pressure pulses generated by the periodic breathing of the engine. (See Figure 3.2). The recirculated exhaust gas flow is controlled by the EGR valve position and the back pressure with respect to intake pressure. The latter is adjusted using a motorized ‘back pressure valve’ (shown in Figure 3.2) and a wastegate valve (Appendix D). The fuel system was composed of a natural gas (NG) and a diesel system. The natural gas was compressed with a 3-stage reciprocating compressor to provide natural gas at pressures 3600-3700 psi. Diesel from a gravimetric pail was pumped to a high-pressure pump, and subsequently used to set the NG delivery pressure. The high diesel pressure is used as a pilot pressure to a dome loaded regulator and reduced the supply natural gas pressure to about 0.5- 1 MPa less than the diesel rail pressure. The NG-diesel bias, was controlled with a needle valve, and was intended to prevent natural gas from entering diesel lines. The injector used for this study is a 1st generation HPDI (J36-0831) injector from Westport Fuel Systems.  A research-oriented air addition system was used with the SCRE, and as shown Figure 3.2, consisted of three major components: (1) an industrial three stage compressor, which compressed the ambient air to delivery pressures of ~ 300 bar (2) a manual needle valve, used to regulate the diluent flow to be mixed with natural gas, and (3) a mixing chamber, to mix the NG and compressed diluent at high pressures (220 bars). The diluted mixture was in turn injected to the combustion chamber using an unmodified HPDI injector. A buffer tank was installed downstream of the diluent compressor to damp pressure fluctuations for steady flow control. The 3-stage compressor was also used to evaluate other compression strategies and characterize the power requirements and delivery flowrate. This is discussed in more detail in Section 3.2. For a more detailed reference of the air addition system, consult P&ID in Appendix D. 30   Figure 3.2 Flow diagram of the experimental system  3.1.1 Data Acquisition and Instrumentation The data acquisition system was from National Instruments modules and described in more detail in [21]. A LabVIEW 6.0 interface in the control room computer was used to operate, monitor, control, and log data from the engine. The system allowed recording of fast response in-cylinder pressure measurements (described in Section 3.1.3) for 45 consecutive cycles at ½o CA. All other engine operation and measurement data was recorded via separate ‘slow’ data modules at a sampling frequency of 1 Hz for a duration of 180s. The fast response data was recorded prior or at the beginning of the ‘slow’ data. Fuel flow rate, intake pressure, back pressure, intake CO2, and heat release rate plots were continuously monitored during data recording to ensure engine operation stability. Additional plots to facilitate the monitoring of fuel dilution ratio (described in 31  Section 3.1.2) and compressor outlet pressure were added to the traditional LabVIEW code. Following the recording of a data point, the calculated parameters of GIMEP, IHR50, gas rail pressure (GRP), diesel flowrate, Φ, EGR, and FDR were rechecked to ensure that the point didn’t drift.  A list of general instrumentation used in SCRE are given in Table 3.2. Some of them are described in more detail in Appendix P. Instrumentation and their measurements used for the calculation of parameters relevant to the conclusions of this study are described in  more detail in Sections 3.1.2 - 3.1.5. Table 3.2 List of general instrumentation used in SCRE experiments Measurement Instrument Pressure at various locations for monitoring Setra Model 209 gauge pressure transducer Temperature  Omega K-type thermocouple Diesel flowrate (?̇?𝑑) Closed loop gravimetric scale  (Adam Equipment PGW 4502e precision balance) Intake air mass flowrate (?̇?𝑖𝑛,𝑎𝑖𝑟) Venturi system ?̇?  UBC machined venturi flowmeter (Cv = 0.995) Omega PX2300-2DI differential pressure transducer  𝜌 Setra model 209 gauge pressure transducer K type thermocouple  3.1.2 Flow Measurement  The amount of diluent-air was quantified by the fuel dilution ratio (FDR) below:  𝐹𝐷𝑅 =  (?̇?𝑑𝑖𝑙,𝑎𝑖𝑟?̇?𝑛𝑔) 100 (3.1) where  ?̇?𝑑𝑖𝑙,𝑎𝑖𝑟 is the diluent-air flowrate and ?̇?𝑛𝑔 is the natural gas fuel flowrate. The diluent and NG flowrate were measured using two separate Coriolis flowmeters (Promass 80 A) as can be seen in Figure 3.2. The specifications of the instrument relevant to uncertainty analysis are presented in Appendix A 32  3.1.3 In-Cylinder Pressure Measurement  A flush-mounted water-cooled piezo-electric pressure transducer (Model: Kistler 6067C) was used to measure the crank resolved (0.5 CAo resolution) in-cylinder pressure response of the SCRE. This was pegged at BDC (bottom dead centre) with a piezo-resistive transducer (PCB piezotronics Model: 1501C02EZ) at the intake manifold to obtain the absolute in-cylinder pressure variation with CAo  (𝑝(𝜃)). The in-cylinder pressure measurement (𝑝(𝜃)) forms the basis of thermodynamic analysis with SCRE because many primary engine performance parameters are defined in terms of indicated parameters (GIMEP: gross indicated mean effective pressure, 𝑃𝑖,𝑔: gross indicated power, GISFC: gross indicated specific fuel consumption). The gross indicated work, i.e. the work delivered to the piston during the compression and expansion stroke in a 4-stroke internal combustion engine cycle is given by Equation (3.2)   𝑊𝑖,𝑔 = ∫ 𝑝(𝜃)𝑑𝑉(𝜃)𝜃=180𝜃=−180 (3.2)  where 𝑝(𝜃) and 𝑉(𝜃) are the absolute in-cylinder pressure and engine in-cylinder volume at a crank angle, respectively. The variation of engine cylinder volume with 𝜃 can be calculated based on known geometric parameters of piston, crank shaft, in-cylinder geometry etc. as described in [46]. The gross indicated power (𝑃𝑖,𝑔) was used for normalizing emissions to derive power specific emissions and to calculate gross indicated efficiency (𝜂𝑖,𝑔). 𝑃𝑖,𝑔 was calculated as shown in Equation (3.3):  𝑃𝑖,𝑔 =𝑊𝑖,𝑔 × 𝑁2 × 60 (3.3) 33  where 𝑁 is the engine rpm. 𝜂𝑖,𝑔 was calculated as shown in in Equation (3.4)  𝜂𝑖,𝑔 =  𝑃𝑖,𝑔(?̇?𝑁𝐺 . 𝐿𝐻𝑉𝑁𝐺 +?̇?𝑑. 𝐿𝐻𝑉𝑑) (3.4) where ?̇?𝑁𝐺 is mass flowrate of natural gas consumed, 𝐿𝐻𝑉𝑁𝐺 is lower heating value of natural gas= 48,810 kJ/kg,  ?̇?𝑑 is mass flowrate of pilot diesel consumed and 𝐿𝐻𝑉𝑑 is lower heating value of diesel = 42,772 kJ/kg. For the calculation of 𝑊𝑖,𝑔 and all indicated parameters derived from it, the average value from 45 consecutive cycles is used.  The apparent net heat release rate (AHRR or 𝑑𝑄𝑛𝑑𝜃) represents the combustion energy released less the heat lost to the engine cylinder walls and is calculated with Equation (3.5) below:  𝑑𝑄𝑛𝑑𝜃=𝛾𝛾 − 1𝑝𝑑𝑉𝑑𝜃+1𝛾 − 1𝑉𝑑𝑝𝑑𝜃 (3.5) where 𝛾 is ratio of specific heats (assumed constant= 1.3), p and 𝑉are pressure and volume at a particular 𝜃. It should be noted that temperature variation in the combustion chamber would cause 𝛾 to vary, therefore, a constant assumption would be applicable only in the context of qualitative analysis.  The integral of the heat released (IHR= ∫𝑑𝑄𝑛𝑑𝜃𝑑𝜃) was used to define the combustion timing. The combustion timing is taken as the crank angle corresponding to 50% of the maximum integral heat released and will be hereafter referred to as IHR50. Similarly, IHR5, IHR10 and IHR90, are taken as crank angles corresponding to 5, 10, and 90% of the maximum IHR.  IHR was also used to define the start of pilot combustion (PSOC) and the start of natural gas combustion (GSOC) using a method detailed in [22]. These definitions were used to determine the pilot ignition delay (PID) and gas ignition delay (GID) as follows: 𝑃𝐼𝐷 = (𝑃𝑆𝑂𝐶 − 𝑃𝑆𝑂𝐼) 34  𝐺𝐼𝐷 = (𝐺𝑆𝑂𝐶 − 𝐺𝑆𝑂𝐼) where PSOI is the pilot injection timing and GSOI is the gas injection timing.  3.1.4 Particulate Emissions Measurement The particulate emissions measurement systems consisted of an engine exhaust sampling system and an aerosol measurement device: TSI DustTrak DRX (Model: 8533). The sampling system diluted the exhaust sample to keep the PM concentrations within maximum limits of sampling instruments. The dilution air was sourced from the same compressor used for SCRE intake, and it was heated to nearly the temperature of the sampled exhaust (~70 oC) at this location. Mixing of diluent (air) with the sampled exhaust at similar temperatures, prevents the initial condensation and nucleation of semi-volatile components [27]. With the aid of an ejector dilutor, raw exhaust was sampled to a secondary dilution in a mixing chamber. This secondary dilution cooled the sample to near 55oC which inhibited further PM growth, resulting in frozen PM emissions. The dilution ratio was calculated by comparing the CO2 concentration of the diluted sample (measured using California Analytical NDIR analyzer (Model 100)) to the exhaust CO2 concentration (measured using AVL gaseous emissions bench as described in Section 3.1.5) [27]. Two rotameters allowed for the control of primary and secondary dilution to achieve dilution ratio from 6, at the lowest PM emissions (low EGR and high air addition operating conditions) to nearly 64, at the highest PM emissions (high EGR and zero air addition operating conditions). It was ensured that even at the lowest PM emissions, the measured concentration of PM (minimum = 0.035 mg/m3) was more than the instrument’s resolution by at least an order of magnitude (see Appendix A.1 for instrument specifications of TSI DustTrak DRX. The sample line then goes via a UBC developed thermodenuder system by Patchuk [27] where it is heated to 200oC for removing volatile components of the PM, before being sampled by the TSI DustTrak DRX. 35  The TSI DustTrak DRX works on the principle of light scattering dictated by Raleigh theory. The instrument uses a laser to illuminate the aerosol particles, which by virtue of its much smaller size compared to the laser’s wavelength, scatters light with the intensity proportional to the sixth power of the particle diameter (mass2), as well as the refractive index of the particle. The scattered light is collected by an inbuilt photodetector producing a voltage proportional to the intensity of scatter light. The voltage is then related to PM concentrations based on a calibration with Arizona Test Dust (ISO 12103-1 A1).  However, the calibration dust unlike engine PM has a much broader distribution, and a different refractive index, which would cause the instrument not to be perfectly correlated with PM mass. Also, previous studies by Patychuk and Rogak [73],  had demonstrated significant mass concentration overestimations resulting from semi-volatile fractions because of their differences in refractive index. The thermodenuder system just upstream of the DustTrak, helped resolve issues from volatile compounds; however, the possibility of a mass estimation error resulting from volatile compounds in the sample should always be acknowledged. For all its limitations, the instrument still has a fast response time (1s) compared to most PM sampling instruments and a very low detection limit, therefore, making it a valuable PM sampling instrument. Attempt has been made in this study, to analyze the DustTrak measurements with detailed uncertainty analysis as presented in Appendix A.2.  3.1.5 Gaseous Emissions Measurement The measurement of the gaseous emissions was done with an AVL CEB II Emissions Bench. The sampling location was far downstream of the engine exhaust port, and therefore provided sufficiently long residence times for the homogeneous mixing of the exhaust gases. A small sample flow of at least 14 LPM was drawn through the sampling location, passed through a heated filter before being sent through a temperature-controlled sample line to the AVL CEB II 36  Emissions Bench. This sample flow was then distributed to individual analyzers (all except intake CO2) for measuring different gaseous emissions. For the measurement of intake CO2, the sample is extracted from a location just downstream of the intake surge tank as can been seen in Figure 3.2.  Various analyzers in the AVL CEB II Emissions Bench were used for the measurement of different gases. The total unburnt hydrocarbon emissions (THC) and methane (CH4) were measured using individual Flame Ionizing Detectors (Pierburg FID 4000 hhd). The sample is passed via the first FID, which measures the propane equivalent of unburnt hydrocarbons. A portion of this sample is passed via a thermochemical converter to convert all non-methane hydrocarbons into CO2 and H2O before being measured of CH4 via a second FID. During post-processing with the propane-equivalent measurement of the THC is converted to a methane equivalent measurement with a division by 3. The NOx (NO2 and NO) emissions were measured with a chemiluminescence detector (Pierburg CLD 4000 hhd). Non-dispersive infrared absorption analyzers (ABB Uras 14 EGA) were used to measure the CO, CO2 from exhaust and the CO2 from intake. The O2 because of its magnetizing properties in an external magnetic field, was measured with paramagnetic detector analyzer (Magnos 106 EGA). The analyzers tend to drift, and therefore were calibrated with reference gases at the beginning of each testing day. In addition to the calculation of specific emissions, the emission measurements were used to characterize the combustion efficiency (𝜂𝑐). The 𝜂𝑐 is a measure of how effectively the heat content of a fuel is transferred into usable heat and is defined by Equation (3.6) based on its definition given in Heywood [46].  37   𝜂𝑐 = 1 −∑ 𝑥𝑖𝐿𝐻𝑉𝑖𝑖?̇?𝑁𝐺𝐿𝐻𝑉𝑁𝐺 ?̇?𝑖𝑛+  ?̇?𝑑𝐿𝐻𝑉𝑑 ?̇?𝑖𝑛 (3.6) where ?̇?𝑖𝑛 is the sum of mass flowrates of natural gas, pilot-diesel, fresh intake air, air addition air and recirculated exhaust gas. 𝐿𝐻𝑉𝑖 and 𝑥𝑖  are the lower heating value and the measured mass fraction, respectively of the ith species (𝐿𝐻𝑉𝑐𝑜 = 10.1 MJ/kg, 𝐿𝐻𝑉𝐶𝐻4 = 50 MJ/kg, 𝐿𝐻𝑉𝑛𝑀𝐻𝐶 = 46.35 MJ/kg (non-methane hydrocarbon), and 𝐿𝐻𝑉𝑠𝑜𝑜𝑡 = 32.8 MJ/kg) measured using exhaust gas analyzers.  The EGR rate (EGR%) was calculated from emission gas measurements using relations described in [74]. The EGR rate in general is defined by Equation (3.7) as follows:  EGR%=?̇?𝐸𝐺𝑅?̇?𝐸𝐺𝑅+?̇?𝑖𝑛,𝑎𝑖𝑟× 100 (3.8) where ?̇?𝐸𝐺𝑅 is the mass flowrate of recirculated exhaust flow (?̇?𝐸𝐺𝑅) and ?̇?𝑖𝑛,𝑎𝑖𝑟 is the amount of fresh intake air flowrate. The calculated value of EGR% and the measured value of ?̇?𝑖𝑛,𝑎𝑖𝑟 was used for the calculation of ?̇?𝐸𝐺𝑅.  An oxygen based equivalence ratio (Φ) was proposed in [21] since for EGR applications, the oxygen contained in the recirculated exhaust cannot be accounted with the traditional air based equivalence ratio as defined in Heywood [46]. For this study, Φ has been corrected further to also include the oxygen contained in air addition and is defined with the Equation (3.9):  Φ =  (?̇?𝑜2, 𝑖𝑛  ?̇?𝑁𝐺 + ?̇?𝑑)𝑠𝑡𝑜𝑖𝑐ℎ𝑌𝑜2,𝑖𝑛?̇?𝐸𝐺𝑅 + ?̇?𝑖𝑛,𝑎𝑖𝑟 + ?̇?𝑑𝑖𝑙,𝑎𝑖𝑟?̇?𝑁𝐺 + ?̇?𝑑 (3.9) 38   where (?̇?𝑜2, 𝑖𝑛  ?̇?𝑁𝐺+?̇?𝑑)𝑠𝑡𝑜𝑖𝑐ℎ is the oxygen mass fraction in intake air at stoichiometric conditions, ?̇?𝑑𝑖𝑙,𝑎𝑖𝑟 is the mass flowrate of air addition air, and 𝑌𝑜2,𝑖𝑛is the intake oxygen mass fraction calculated as described in [21].  3.2 Diluent (Air) Compression System    Figure 3.3 The 3-stage reciprocating diluent (air) compressor  A 3-stage reciprocating compressor (Northshore Artic 1000 series) was used to provide compressed air for air-addition experiments. The compressor was driven with a three-phase motor through the usage of an industrial V-belt.  An automatic drainage system facilitated the automatic discharge of the water collected in moisture separator every 20 minutes, and thereby eliminated the need for the operator to enter the engine cell while the system was running. The compressor was splash lubricated and uses a di-ester based synthetic oil. Synthetic oil decomposes at temperatures >220oC, and therefore the operator must operate the compressor with careful monitoring of discharge temperatures. Also, the oil was changed prior to the compressor 3-phase Compressor motor Auto-drain 3-stage compressor 39  characterization measurements (Chapter 5) since power consumed by friction is important for assessing the mechanical performance of the compressor. Pressure reliefs valves existed at each stage outlets to prevent over pressure and vessel bursts. A pressure switch, also ensured that the compressor shuts off when the third stage outlet pressure reaches 330 bar. The relevant manufacturer based specifications of the compressor from [75] is presented in Table 3.3 below:  Table 3.3 Key specifications of 3-stage compressor  Model Artic 1000 Operating Pressure 5000 psi Delivery with 7.5 HP motor 7.5 CFM FAD (9 CFM Charge Rate) Number of stages 3 First stage bore 90 mm Second stage bore 36 mm Third stage bore 14 mm Piston stroke 40 mm Rotation speed 1500 rpm Moisture drainage Auto (stock) for air addition experiments Manual for compressor characterization experiments  As discussed in Section 1.4, the assessment of the diluent (air) compression system performance is a critical component in determining the system level implications of air addition. Therefore, the 3-stage compressor was instrumented (Figure 3.4) to characterize the power and flowrates for 3-stage and 2-stage configurations. The 2-stage was of interest for future system integration studies and is discussed in Section 3.2.1. In both these configurations, the third stage outlet pressure or the delivery pressure was controlled manually via a back-pressure regulator. The relevant measurement instrumentation (Figure 3.4), and the measurement based calculated parameters required for characterizing of the compressor performance are discussed in Sections 3.2.2 - 3.2.4.  40    Figure 3.4  P&ID of the compressor in its ‘stock’1 configuration with relevant instrumentation  3.2.1 Development of Prototype 2-stage configuration A 2-stage configuration prototype was developed from the 3-stage compressor to the asses the performance of diluent compression system with reduced sizing. This required several adaptations to the three-stage reciprocating compressor. The first stage was bypassed by connecting the 1st stage outlet to its inlet in a three-port connection, with the open port drawing additional air from atmosphere via a rotameter (Figure 3.5). This arrangement allowed the first                                                  1 Manufacture supplied configuration i.e. 3-stage configuration with the crankcase vent connected to 1st stage inlet 41  stage to discharge at ambient pressure, and thereby reducing its indicated power consumption (ideally zero). It also allowed for the measurement of leak contribution to blowby from first stage. The 2-stage configuration, therefore, used only the 2nd and 3rd stages of the 3-stage configuration for diluent compression. To achieve the required delivery pressures, a 2-stage diluent compressor would require very high stage pressures ratios (>17). To reduce the pressure ratios, compressed air was provided to 2nd stage inlet to reduce the overall and the stage pressure ratios. The compressed air was taken from the supercharger used to supply intake air to the SCRE. The range of intake pressures were based on relevance to typical engine operation, taken here as typical intake manifold pressures. The compressor was also converted to manual drainage from auto drainage. The auto drainage system as shown in 3-stage configuration of the compressor (Figure 3.4) is designed to leak if the compressor operated at pressure ratios outside the range of regular operation. Since the 2-stage configuration was supplied with inlet pressures that were less than half its design pressures, the stages would vent with manual drainage. A P&ID of the compressor in its 2-stage configuration is presented in Figure 3.5. 42  3.2.2 Compressor Pressure, Temperature, and Flow measurement  Instruments used for the measurement of pressure, temperature, and mass flowrate are listed in Table 3.4. The measurement locations of respective configurations are shown in Figure 3.4 and Figure 3.5.   Figure 3.5  P&ID of the compressor in its 2-stage configuration   43  Table 3.4 Instrumentation used in compressor characterization Measurement Instrumentation Temperature at inlet and outlet of various stages Omega ungrounded K-type thermocouples Pressure at inlet and outlet of various stages Setra Model 209 gauge pressure transducer Atmospheric pressure taken as the 1st stage inlet pressure in the 3-stage configuration Digital Barometer  Compressor outlet mass flowrate Endress+Hauser Promass 80A coriolis mass flowmeter Compressor inlet mass flowrate Beko Technologies FS 211 hot film mass flowmeter  These measurements were used for ensuring safe operation, determination of steady state, and calculation of parameters relevant for analysis. Safe operation was ensured by monitoring the stage outlet pressure and temperatures. Steady state was attained when rise in stage outlet temperatures were less than 1oC per minute. Parameters relevant for the results and discussions presented in Chapter 5 are defined below:  The measured volumetric efficiency (𝜂𝑣,𝑚) was determined as follows:  𝜂𝑣,𝑚 =?̇?𝑖𝑛(𝜌𝑖𝑛𝑉𝑠𝑁)60 (3.10) where ?̇?𝑖𝑛 is the compressor inlet mass flowrate, 𝜌𝑖𝑛 is the inlet density, 𝑉𝑠 is the swept volume of the concerned stage, and N= 1500 rpm is the speed of the compressor.  The isentropic power (?̇?𝑖𝑠𝑒𝑛,𝑖) of a stage was determined as follows:  ?̇?𝑖𝑠𝑒𝑛 = ∑ ?̇?𝑖𝑠𝑒𝑛,𝑖𝑖=𝑘𝑖=1=  ∑ ?̇? ∫ 𝐶𝑝𝑑𝑇𝑇𝑖,1(𝑃𝑖,2𝑃𝑖,1)𝛾−1𝛾𝑇𝑖,1𝑖=𝑘𝑖=1 (3.11)  (𝑃𝑖,2𝑃𝑖,1) is the measured pressure ratio across the ith stage, 𝑇𝑖,1 is the corresponding measured inlet temperature for the ith stage, k is the total number of stages, and ?̇? the outlet mass flowrate of the 44  compressor. ?̇?𝑖𝑠𝑒𝑛 was further used to calculate isentropic specific power (𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛=1 = ?̇?𝑖𝑠𝑒𝑛?̇?), and indicated specific power at a particular 𝜂𝑖𝑠𝑒𝑛= a (𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛=𝑎 = ?̇?𝑖𝑠𝑒𝑛?̇?×1𝑎). 3.2.3 Motor Power, RMS Voltage and Shaft Speed measurement  For power requirement calculation, the three-phase compressor motor power consumption was taken as an indication of compressor shaft power (𝑃𝑠ℎ𝑎𝑓𝑡) consumption. The is because motor power (𝑃𝑚𝑜𝑡𝑜𝑟) relates linearly with compressor shaft power when the motor efficiency (𝜂𝑚𝑜𝑡𝑜𝑟) and the belt efficiency (𝜂𝑏𝑒𝑙𝑡) remains constant:    𝑃𝑠ℎ𝑎𝑓𝑡 =  𝜂𝑚𝑜𝑡𝑜𝑟𝜂𝑏𝑒𝑙𝑡𝑃𝑚𝑜𝑡𝑜𝑟 (3.12) The 𝜂𝑏𝑒𝑙𝑡 of the belt configuration used i.e. V-belt depends on the torque and speed but is not expected to be significant for the range of considered torque and speed [71]. The 𝜂𝑚𝑜𝑡𝑜𝑟 also typically remains constant at higher operating loads (> 50%) [70]. Since, 𝜂𝑚𝑜𝑡𝑜𝑟 determination was required for analysis presented in Section 5.4, this was also measured and is discussed below. The measurement of  𝑃𝑚𝑜𝑡𝑜𝑟 is also discussed below.  The three-phase motor power (𝑃𝑚𝑜𝑡𝑜𝑟 ) was measured with a power meter (Elitepro). Since three phase 4-wire ‘wye’ circuit configuration was used to provide power to the motor, the total power was calculated by adding the measured power from each of the three phases [76]. The measured power of a phase was in turn calculated by measuring their RMS phase voltage (𝑉𝑝ℎ𝑎𝑠𝑒) and current.  The motor efficiency was calculated using the slip method [70]. This method additionally required the motor shaft speed measurement (𝑆𝑚), which was measured with a stroboscope with a resolution of 1 rpm. 𝜂𝑚𝑜𝑡𝑜𝑟 was calculated using the following equation: 45   𝜂𝑚𝑜𝑡𝑜𝑟 =𝑃𝑟𝑃𝑚𝑜𝑡𝑜𝑟𝑆𝑠 − 𝑆𝑚(𝑆𝑠 − 𝑆𝑟)(𝑉𝑟𝑉 )2× 100% (3.13) where 𝑃𝑟, 𝑆𝑠, 𝑆𝑟, 𝑉𝑟 and 𝑉 are rated motor power  (5.6 kW), motor synchronous speed (3600 rpm for the considered motor at the power supply frequency of 60 Hz), rated motor speed (3520 rpm), rated voltage, and line to line 3-phase rms voltage ( √3. 𝑉𝑝ℎ𝑎𝑠𝑒), respectively.  3.2.4 Data Acquisition System  The data acquisition system to record ‘slow’ data (pressure, temperature, and mass flowrate data) and the data recording procedure used were the same as that used for SCRE (Section 0). The compressor motor power and the phase voltage measurement data from the dent instruments Elitepro logger was acquired separately by connecting the logger to the operator's laptop with a USB. A software- ELOG 15 provided by the manufacturer (Dent instruments) was used to connect the computer to the logger and active data logging. The sampling rate was once per 3 s and the data was captured over a duration of 3 minutes. Separate self-developed MATLAB routines, were used for the simultaneous processing of the Elitepro logger data file and the ‘slow’ data recorded by SCRE. These routines are included in the electronic appendix in the ‘Compressor Characterization’ folder.  3.3 Measurement Uncertainty The precision was found to be the more significant uncertainty for the current study, and has been traditionally used by previous researchers using the SCRE [21,27,37,77]. This is because analysis of experimental uncertainty was primarily aimed at identifying statistically significant results, specifically for the interpretation of effects of air addition on emissions and gross indicated efficiency. Ideally, such an uncertainty would be estimated by calculating confidence intervals for a sample mean value obtained with its repeated measurements (usually >3). Repeating each point 46  multiple times is very time intensive, and therefore alternate approaches which required less number of repeat points were pursued and are discussed below.   Two different approaches of uncertainty calculation were used in this study. The first approach calculated confidence intervals from standard errors (𝜎𝑠𝑒 =𝜎𝑠√nr; where 𝜎𝑠 is sample standard deviation and nr is number of observations) of various repeat points to obtain a linear relationship with the measured mean (due to varying operating conditions) values of the repeat points: • The repeat points, tabulated in Table 3.5, were used to determine 95% confidence intervals (Δ) and the mean values of a parameter (e.g. specific PM emissions, Gross indicated efficiency). Student’s t-values (𝑡95) were used to calculate Δ through the following Equation [78]:  Δ =𝑡95𝜎𝑠√nr (3.14) • A single linear relationship between the confidence intervals and the mean value of a parameter was obtained using least squares fit, which was then used to calculate the Δ, taken as the representative uncertainty (𝑢𝑡), at any measured value for that parameter.  This approach was used to determine the representative uncertainty of various parameters discussed in Chapter 4 and Section 6.3. A second approach was used for parameters calculated based on response surfaces, particularly that used in Section 6.4. This approach is also based on the calculation of standard error (𝜎𝑠𝑒) but calculated with a different formulation designed for curve fits [78]:   𝜎𝑠𝑒 =  √∑ (𝑦𝑖 − ?̂?𝑖)2𝑛𝑖=1𝑛 − 2  (3.15) where  𝑦𝑖, ?̂?, and 𝑛 is the measured value of the response variable, the corresponding predicted response variable, and the number of points used for the calculation of the response surface, 47  respectively. The representative uncertainty with this approach is absolute and equal to the 95% confidence interval based on 𝜎𝑠𝑒 (1.96𝜎𝑠𝑒). The first approach allowed for a consistent determination of whether the error or the relative error of a parameter remained constant. As seen in Figure 3.6a, Δ varied proportional to the mean value of specific PM emissions, favoring a constant relative error for 𝑢𝑡. This was also seen in all other specific emission parameters. However, in case of 𝜂𝑖,𝑔, a constant error was found to be more representative (Figure 3.6b).  The representative uncertainty, based on first approach (𝑢𝑡), was also compared to the uncertainty calculated (𝑢𝑐𝑎𝑙) with the propagation of error [78], using the instrument specifications (see Appendix A for 𝑢𝑐𝑎𝑙 discussion). As listed in Table 3.6, this was done for certain parameters at the primary repeat point, which is highlighted in bold in Table 3.6. In most cases, 𝑢𝑐𝑎𝑙 was smaller than 𝑢𝑡. This is likely due to other sources of error (e.g. operator set point errors, repeatability of engine, and its operating points) contributing significantly to 𝑢𝑡, which cannot be fully accounted by just the instrument uncertainties.     Figure 3.6  Nature of uncertainty variations in different parameters (a) constant relative error variation for PM (b) constant error variation in gross indicated efficiency   48  Table 3.5 Repeat points of experiments in Chapter 4 Note: t95: Student’s t-values for a 95% confidence interval [78], nr: number of repeat points. The operating point highlighted in bold is the primary repeat point which was repeated every day prior to measurements conducted on that day. Constant parameters EGR % Φ FDR % nr t95 GIMEP: 16.8 ±0.3 bar Engine speed: 1350 ±10 rpm IHR50: 10 ±0.75 oATDC PSEP: 0.3 ms GRP: 22 ±0.2 MPa Diesel flow: 8.5 ±1.25 mg/inj     12.5 0.58 0 16 2.131 12.5 0.58 25 8 2.365 25 0.58 0 6 2.571 25 0.58 10 3 4.303 25 0.68 0 3 4.303 25 0.77 0 3 4.303 12.5 0.68 0 5 2.776 12.5 0.68 10 4 3.182 12.5 0.68 50 4 3.182 12.5 0.77 0 5 2.776 12.5 0.77 10 3 4.303 12.5 0.77 25 3 4.303 12.5 0.77 50 3 4.303 0 0.58 0 6 2.571 0 0.68 0 4 3.182  Table 3.6 Uncertainty analysis at the primary repeatability point Parameter ?̅?𝑠 COV= 𝜎𝑠?̅?𝑠% 𝑢𝑡 =∆?̅?𝑠% 𝑢𝑐𝑎𝑙% DustTrak PM (g/kWhr) 0.05475  13.51 20 4.87 NOx (g/kWhr) 2.5558  5.15 5.3 4.46 THC (g/kWhr) 0.4314  12.84 15.4 4.48 𝜂𝑖,𝑔% 43.877  0.79 0.8 1.34 49   On-Engine Air Addition Experiments   The on-engine air addition experiments were intended to determine the effects of air addition on the emissions (especially PM and NOx) and performance. Since in-cylinder strategies for emission reduction could potentially bring trade-offs (e.g. typical NOx-soot tradeoff), air addition was investigated at various EGR rates and equivalence ratios (Φ). EGR was considered as a relevant parameter since it has less adverse effects on thermal efficiency at moderate levels compared to other NOx control strategies like IHR50 [29,37]. Since Φ significantly affects emissions and efficiency, it was chosen as the other relevant parameter [27]. Also, Φ relates with the intake manifold pressure, which is chosen as the compressed air source for the inlet of prototype compression system in this study (Chapter 5). Due to time, the study was carried out at a single representative heavy-duty mode (load-speed combination). This mode also allowed a wide EGR- Φ-FDR space to be considered while satisfying the safety constraints. A systematic procedure discussed in detail in B.1 was used to determine this mode (referred to as Mode Z in this study). The testing methodology is described in Section 4.1, and the results and their discussion is presented in Section 4.2 - 4.5.  4.1 Methodology   An engine operating point was defined by 9 parameters (Table 4.1), out of which only three parameters- Φ, EGR rate, and fuel dilution ratio (FDR) were varied between different operating points. The remaining parameters were selected based on previous test results from this single-cylinder engine [21,22,79]. Φ was chosen to be in a range representative of actual heavy-duty operation, with the upper limit restricted to Φ= 0.8 by exhaust temperature. EGR rate was based on suggested EGR fractions in [74], where higher EGR rates resulted in increasing combustion 50  instabilities. Although at a particular EGR rate, the oxygen equivalence ratio (Φ) can account for the varying exhaust O2 concentration in the recirculated exhaust, the intake oxygen mass fraction (𝑌𝑜2,𝑖𝑛) is a more relevant parameter for analyzing emissions  (e.g. NOx  [49]). The Fuel dilution ratios were restricted to 100% because of increasing flowrate instabilities at higher diluent (air) flowrates. For the case of EGR= 0%, the exhaust pressure was set to be the same as at EGR= 12.5% (within 15 kPa), for the same Φ and FDR. This was done as large variations in exhaust pressure at EGR= 0%, could  affect emission characteristics because of varying exhaust gas residuals ([74]). Equalizing the exhaust pressure of a zero-EGR point to that of an equivalent point with EGR, enabled differences resulting from varying EGR to be isolated without the influence of back pressure. Randomization of EGR and Φ was achieved by taking EGR-Φ combinations on different days. A repeatability point (highlighted in bold in the Table 4.1), was measured at the beginning of each testing day to account for day to day variabilities (Section 3.3). Following the repeat point, the data was analyzed to identify for significant deviations (> 3𝜎𝑠) in emissions. If so, facility troubleshooting was conducted prior to next set of measurements. The test matrix is presented in Table 4.1:        51  Table 4.1 Engine operating set points for air addition experiments Note: The repeatability point is in bold. GIMEP: gross indicated mean effective pressure; PSEP: pilot separation or the time interval between end of pilot pulse and beginning of natural gas pulse; GRP: NG gas rail pressure; Φ: oxygen-based equivalence ratio; EGR: exhaust gas recirculation flow rate; FDR: fuel dilution ratio; 𝑌𝑜2,𝑖𝑛: intake oxygen mass fraction.  Constant parameters Variable parameters EGR% Φ 𝑌𝑜2,𝑖𝑛 FDR%  GIMEP: 16.8 ± 0.3 bar  Engine speed: 1350 ±10 rpm  IHR50: 10 ±0.75 oATDC  PSEP: 0.3 ms  GRP: 22 ±0.2 MPa  Diesel flow: 8.5± 1.25 mg/inj  25±1.75 0.58±0.025 0.194 [0; 10; 25; 50; 100] ±2.75 0.68±0.025 0.189 [0; 10; 25; 50; 100] ±2.75 0.77±0.025 0.185 [0; 10; 25; 50; 100] ±2.75 12.5±1.75 0.58±0.025 0.214 [0; 10; 25; 50; 100] ±2.75 0.68±0.025 0.21 [0; 10; 25; 50; 100] ±2.75 0.77±0.025 0.207 [0; 10; 25; 50; 100] ±2.75 0 0.58±0.025 0.23 [0; 10; 25; 50; 100] ±2.75 0.68±0.025 0.23 [0; 10; 25; 50; 100] ±2.75 0.77±0.025 0.23 [0; 10; 25; 50; 100] ±2.75  For the presentation of majority of the results in the following sub-sections, variations in tested Φ and FDR are grouped under a EGR% and depicted in a single plot. A qualitative understanding of trends associated with EGR= 12.5% can easily be achieved from the trends observed with EGR= 25% and EGR= 0%, and therefore plots associated with EGR= 12.5% are not presented in this chapter (B.4 contains the summary data corresponding to EGR= 12.5%). However, its data was used for the development of response surfaces in Chapter 6. To provide context, the current results are discussed with respect to previous results from the N2 addition study [23], in part to elucidate the effect of O2 present in air addition. It should be noted that different operating modes (Table 4.2) and instrumentation are considered for the two studies. Also, the comparisons of PM between both the studies should be taken on a qualitative basis since the instrument as well as the measurement principle were different. The current study measured 52  denuded light scattering mass concentration using DustTrak (see Section 3.1.4 above) while the N2 addition study [23] measured un-denuded mass concentration of PM using a tapered element oscillating microbalance (TEOM). Patychuk et al. [27] demonstrated that although the denuded PM measurements with DustTrak correlated linearly with TEOM, under predictions of PM by up to 30% with DustTrak may result due to the nearly absent semi-volatile fractions in denuded PM.   Table 4.2 Comparison of Engine operating set points: air addition experiments and previous N2 addition study [23]  Note: GIMEP: gross indicated mean effective pressure; PSEP: pilot separation or the time interval between end of pilot pulse and beginning of natural gas pulse; GRP: NG gas rail pressure; EGR: exhaust gas recirculation flow rate; FDR: fuel dilution ratio. Operating parameters Air addition N2 addition Gross indicated Power (kW) 47 35 GIMEP (bar) 16.8 13.5 EGR rate (%) 0; 12.5; 25 30 Intake oxygen mass fraction [0.185; 0.189; 0.194; 0.207; 0.21; 0.214; 0.23] 0.19 IHR50 (oATDC) 10 [0; 5; 10; 15] GRP (MPa) 22 21 PSEP (ms) 0.3 1 Diesel (% of total fuel energy) 4 5 FDR (%) [0; 10; 25; 50; 100] [0; 43; 116]  4.2 Combustion Effects Air addition results in major effects on the combustion, particularly on the apparent heat release rates (AHRR). This is because of two key differences that result in dilution of fuel with air: reduction of energy content per unit mass of injected fuel and longer injection durations. Longer injection durations are required with air addition to ensure that comparable amount of fuel with respect to non-diluted cases were delivered (Figure 4.1). To maintain IHR50 with the longer 53  injection duration, the pilot injection timing (i.e. PSOI) was advanced with a constant PSEP (pilot separation).   Figure 4.1  Fuel pulse diagram: (a) HPDI (b) HPDI with air addition  The longer injection duration results in increased total kinetic energy transferred to cylinder charge. The better mixing from this added kinetic energy has been argued to be the cause of reduction of PM and CO emissions observed with N2 addition in PIDING engine in [23]. This section will discuss the combustion effects with air addition at different EGR and Φ.  4.2.1 Apparent Heat Release Rates The apparent heat release rate (AHRR) is broadly categorized into three regions, namely the pilot combustion region, the partially premixed combustion phase, and the mixing-controlled combustion phase. The first peak in the AHRR (Figure 4.2) corresponds with the pilot combustion. The partially premixed phase is limited by the kinetics of the reactants, and in the context of this study represents the region in the AHRR from the first minima to the absolute maxima. The partially premixed combustion phase is followed by mixing controlled combustion, where the (a) (b) 54  combustion is limited primarily by the rate of fuel injection, and its ability to mix to a combustible mixture.    Figure 4.2 Effect of FDR on AHRR at various EGR and Φ  The pilot combustion region showed increasing AHRR with air addition, particularly for the case of 100% (Figure 4.2). This is unexpected since both the diesel rail pressure and the PPW were kept constant and could be due to increased injected pilot mass (up to 1 mg/injection, which is nearly 12% increase in pilot energy content from a nominal pilot mass of 8.5 mg/inj) caused by lower in-cylinder pressures at the advanced PSOI with higher FDR. It is also possible that the nominal pilot mass was lower (by about 1.5 mg/injection) than measured because of an unaccounted leak from natural gas dome loaded regulator, discovered after the air addition study.   55  The reduced fuel concentration reduces the intensity of the partially premixed phase of combustion (Figure 4.2). This is due to less natural gas injected and mixed prior to combustion, which could be both because of reduced methane concentration in diluted fuel, as well as reduced ignition delays (~ 16% reduction in gaseous ignition delay (GID) at Φ= 0.7, 25% EGR as tabulated in B.4) comparable to that (~ 17% reduction in GID) seen in the previous N2 addition study [21,23]. Furthermore, the reduced natural gas combusted in the partially premixed phase would also result in reduced fraction of energy released in the partially premixed phase relative to that of mixing-controlled phase.  In agreement with previous N2 addition studies [21,23], air addition also had significant effects on the combustion intensity of the mixing controlled combustion (Figure 4.2). To understand this further, the mixing-controlled combustion phase of a diluted case is further divided into two regions: early mixing-controlled phase and late mixing controlled phase. The former is considered as the combustion occurring in the time interval from the peak AHRR (of the diluted case) to the end of gas injection in the undiluted case, and the latter is the combustion that follows the early part. This is highlighted in the schematic below:  Figure 4.3  Late mixing controlled phase: (a) HPDI (b) HPDI with air addition   (a) (b) Late mixing controlled phase  56  Since mixing controlled combustion is primarily dependent on the instantaneous fuel mass present in the cylinder charge and the injected turbulent kinetic energy [80], the former is reduced with air addition, which in turn reduces the intensity of AHRR relative to non-diluted cases in this phase. The latter should not be significantly affected in the early part because the kinetic energy of the diluted fuel jet would be comparable to that of an undiluted case at the same nominal injection pressures. However, the injected turbulent kinetic energy in the later part of mixing controlled combustion of the diluted case should relatively increase because of the extra diluent mass and corresponding longer injection durations. This can be clearly seen in the functional form of injected kinetic energy proposed by Chmela and Orthaber [80]:  𝑑𝐸𝑘𝑖𝑛𝑑𝜃= 𝛼. 𝜌𝑓 . (𝑛𝜇𝐴)2. 𝑄𝑓3 (4.1) where 𝜃 is crank angle, 𝛼 is a tuning constant, 𝜌𝑓 is fluid density, 𝑛 is engine speed, 𝜇 is flow coefficient, A is spray hole area, and 𝑄𝑓 is volumetric injection rate. Equation (4.1) clearly shows that longer injection would lead to more net injected kinetic energy for constant 𝜌𝑓 and 𝑄𝑓. With air addition, the 𝜌𝑓 of diluted fuel would increase due to its increased molecular weight. Assuming choked flow, it can be seen that 𝑄𝑓 will also increase due to higher velocity (speed of sound relatively higher in diluted fuel). Both these effects, therefore, should also contribute to an increase of instantaneous injected kinetic energy with air addition. The later part with air addition, compared to that of 0% FDR, also has ongoing fuel injection which would contribute to heat release. The net result of these effects is seen in Figure 4.2, where the early part of the mixing-controlled combustion is of reduced intensity, while the later part has higher AHRRs in the diluted cases.  57  The effects of EGR or Φ on AHRR were relatively less significant than that from FDR (B.2). As was also seen in previous HPDI studies [22,37], decreasing Φ was found to make the peak of AHRR relatively flatter. This effect was observed across all combinations of FDR and EGR (B.2). As shown in Figure 4.4, an increasing EGR resulted in a consistent trend of increased premixed combustion intensity as well as a delay in the occurrence of this maxima, for all considered Φ and FDR. A similar observation was made in [74], and could be due to increased gaseous ignition delay (GID) caused by increased recirculated exhaust (B.4), followed by a more intense premixed combustion. Also, this trend was not significantly affected with variation of FDR (Figure 4.4).    Figure 4.4 Effect of EGR on AHRR at two different FDR and Φ= 0.68  4.2.2 Combustion Duration Similar to other studies [21,23], the combustion duration was considered in two phases: early cycle and late cycle combustion (Figure 4.5), with the former being the duration between IHR10 and IHR50 and the latter taken as the duration IHR50 to IHR90 (Section 3.1.3). In regular PIDING operation (without fuel dilution), the early and late cycle combustion durations are dominated by partially premixed and mixing controlled combustion phases, respectively [21]. Increased intensity of first premixed combustion maxima 58  However, this may change as the duration of mixing controlled phase increases relative to partially premixed combustion phase, as in the case with air addition, and thereby may occupy a large part of the early cycle combustion. This is elucidated in Figure 4.2, where the peak AHRR occurs before IHR50 (10o ATDC).  The effect of FDR on the overall combustion duration and its constituent phases at various Φ  and EGR is shown in Figure 4.5. Overall combustion duration (IHR10 to IHR90) increased with increasing Φ (Figure 4.5a) owing to longer gas pulse widths required at higher Φ. This trend is sensitive to FDR and EGR (Figure 4.5a). The trend of increasing combustion duration with increasing air addition at 25% EGR, didn’t agree with previous N2 addition results [21,23] having similar EGR levels (30% EGR), where the duration remained constant. The early cycle duration increased with air addition in all tested cases of EGR and Φ (Figure 4.5c). This was also observed in previous N2 addition study [21,23], and is due to the reduced duration of partially premixed combustion phase with air addition. Interestingly, the early cycle duration was also found to be independent of EGR and Φ (Figure 4.5c). Late cycle duration shows different behavior with air addition at different EGR levels, and therefore is the underlying source of overall duration trends. With no EGR, the late cycle duration, similar to N2 addition [21,23], decreased due to increased late cycle turbulence, and therefore a faster mixing controlled combustion. This mixing effect from the diluent is also expected at higher EGR (25%) but didn’t result in decreased late cycle combustion duration (Figure 4.5b), and therefore suggests the possibility of an additional counteracting effect at high EGR. However, the fact that a decrease in combustion duration was observed in the N2 addition study [21,23] conducted at similar EGR (30%) and combustion timing (IHR50), suggests that another effect (e.g. from the presence of O2 in added air) may become more significant at higher EGRs.  59                     Figure 4.5 Effect of FDR on (a) overall combustion duration (b) late cycle duration (c) early cycle duration, at various EGR and Φ  (a) (b) (c) 60  4.2.3 Combustion Efficiency and Stability  Figure 4.6 Effect of FDR on combustion efficiency at various EGR and Φ  The effect of FDR on combustion efficiency (𝜂𝑐; defined in Section 3.1.5) at various EGR and Φ is shown in Figure 4.6. 𝜂𝑐 increased significantly with air addition at high EGR, with up to 4%-point seen at the highest EGR (25%) and Φ (0.77). The major sources of inefficiencies (based on this method of determining 𝜂𝑐) were identified to be unoxidized CO and unoxidized THC (effects of air addition on CO and THC will be discussed in Section 4.3), which constituted greater than 70% and 20% of the inefficiency, respectively. 𝜂𝑐 is improved at low EGRs due to improved CO and THC oxidation at high combustion temperatures, and thereby air addition doesn’t affect 𝜂𝑐 at low EGR (0% EGR). Air addition had a greater effect at higher EGR. At 25% EGR and lower Φ (0.58 and 0.68), 𝜂𝑐 approached 99% with 100% FDR, while at Φ= 0.77, 𝜂𝑐= 98%, indicating potential for 𝜂𝑐 improvement. 61    Figure 4.7 Effect of FDR on the COV of GIMEP with FDR at various EGR and Φ  The combustion stability is commonly characterized by the coefficient of variance (COV) of GIMEP [21] (COVGIMEP), with a lower value indicating higher combustion stability (or reduced variability) and vice versa. The COVGIMEP is calculated as described below:  COVGIMEP=𝜎?̅? (4.2) where 𝜎 is the standard deviation and ?̅? is the mean of GIMEP, calculated from 45 consecutive engine cycles.  The effect of FDR on COVGIMEP at various Φ and EGR is shown in Figure 4.7. COVGIMEP was found to be independent of EGR and Φ. Increasing FDR resulted in up to 0.6%-point decrease in the COVGIMEP, indicating that air addition results in improvements in the combustion stability. This effect of fuel dilution on combustion stability was seen in N2 addition study [21], where up to 0.3%-point decrease in COVGIMEP was achieved. 4.2.4 Adiabatic Flame Temperatures Representative adiabatic flame temperature (AFT) of the mixing controlled natural gas flame (diffusive flame) was calculated using a methodology similar to that by Hill and Mc-Taggart-Cowan in [49], where AFT was calculated to correlate with the NOx emissions of a 62  PIDING engine. This methodology detailed in B.3, determines an equilibrium-based, constant pressure adiabatic flame temperature (calculated with Cantera 2.3 [81]) for stoichiometric methane oxidation (natural gas combustion assumed similar to stoichiometric diffusive burning of diesel [46] in engines) with a portion of intake mixture (air and recirculated gas) from the IVC (intake valve closing). For the equilibrium analysis, the reactant pressure was selected as 86 bar, which is the average of in-cylinder pressures (86 ± 20 bar) obtained at GSOC (Section 3.1.3) for all data points. The reactant temperature was selected as 875 K and is the average of temperatures (875 ± 40 K) determined at GSOC based on an isentropic compression starting from BDC at a temperature taken equal to intake manifold temperature. A sensitivity analysis (B.3) indicated that a change in reactant pressure from 50 to 150 bar, caused a change of only 10 K in the AFT. Increasing the reactant temperature with respect to a baseline reactant temperature (875 K) increased the AFT with a difference nearly equal to that of the considered temperatures; however, this difference reduced with increasing reactant temperature due to equilibrium effects at higher AFTs (B.3). Although not considered for the AFT calculation presented in this study, the temperature of the fuel jet (including the diluent) inside the combustion chamber would be lower than that of the compressed intake gases. A sensitivity analysis revealed that for a fuel jet at 300 K and intake gases at 875 K, the AFT can reduce up to 50 K (Appendix B.3), where the diluent in the fuel jet can contribute to up to 10 K reductions at the highest FDR (100%). It is important to note that a major assumption involved, which may have significant effects on the representative AFT, is that the composition of the intake mixture from IVC is unchanged after undergoing the partially premixed combustion.   The effect of FDR, EGR, and Φ on adiabatic flame temperature with air addition is shown in Figure 4.8. As expected, the most significant effect on AFT is that of EGR, resulting in a 63  decrease of up to 400 K from 0% EGR to 25% EGR. The effects of FDR and Φ on AFT also become more significant with EGR. This is because recirculated exhaust is necessary to influence the mole fractions of the primary gases in the intake mixture, without which the representative AFT of mixing controlled methane flame (with or without air added) will be constant and equal to that of methane-air premixed stoichiometric combustion, as is seen in Figure 4.8b. With recirculation of exhaust (Figure 4.8a), increasing Φ resulted in reduced adiabatic flame temperature due to reduced oxygen mole fraction in the intake mixture. Similarly, increasing the FDR resulted in increased AFTs; however, its effectiveness in increasing the AFT reduced with increasing Φ, with Φ= 0.77 at 25% EGR resulting in no significant changes in AFT (Figure 4.8a). In the same figure, the AFT with air addition is also contrasted with that of N2 addition, which is calculated for the same mass of diluent added. AFT increases with air addition while decreases with N2 addition, providing further support for the differences in NOx behavior noted for these diluents, which is discussed in Section 4.3.3.     Adiabatic flame temperature, as discussed in Section 2.1.3.2, is a good characteristic flame temperature; however, available evidence  justifying this is applicable to flame zones where effects of radiation is less [43,47,48], for example, soot radiation in diesel engines may affect this [50]. Another aspect not captured in this analysis is the effect of air addition on the localized combustion structure at the flame front.   64      Figure 4.8 Effect of FDR on adiabatic flame temperature at various Φ and (a) EGR= 25% (b) EGR= 0%  4.2.5 Maximum Pressure Rise Rate The effect of FDR on maximum pressure rise rate (max (𝑑𝑃𝑑𝜃)) at various EGR and Φ is shown in Figure 4.9. The effect of FDR was dependent on EGR, with a near linear increase of up to 1.25 bar/deg observed with increasing FDR (0 to 100%) at the highest EGR (25%), while at lower EGR (0%, 12.5%), the effects of FDR were significant only when FDR was greater than 50%. max (𝑑𝑃𝑑𝜃) increased with increasing EGR but was not significantly affected by Φ.     Figure 4.9 Effect of FDR and Φ on maximum pressure rise rate at (a) EGR= 25% (b) EGR= 0%  (a) (b) (a) (b) 65  4.3 Emissions 4.3.1 PM and CO The effect of FDR on the PM emissions at various Φ and EGR is shown in Figure 4.10. Two different scalings were used to illustrate this effect: regular scaling to highlight the magnitude of overall PM reductions (Figure 4.10a and b); and log scaling used to show trends over several orders of magnitude (Figure 4.10c and d). Air addition had a significant effect on PM emissions with more than an order of magnitude reduction at 100% FDR (or 90% to 97% reductions) for all considered Φ and EGR (Figure 4.10a and b). At higher EGR (25%, 12.5%), the possibility of further reductions by increased FDR (beyond 100%) is indicated by the non-zero slopes at 100% FDR in Figure 4.10c. However, PM reduction at the lowest EGR (0%) was found to be insignificant beyond 50% FDR (Figure 4.10d).  The mechanisms of PM reduction with air addition are considered in context with previous fuel dilution studies [23,43]. PM reductions observed with N2 addition in a similar PIDING engine study were attributed to increased late cycle mixing which resulted in either inhibition of PM formation (e.g. reduced soot precursors in diluted fuel jet) or enhancement of PM oxidation (e.g. in diluted fuel jet, by enhanced post-combustion mixing) [23]. Enhanced late cycle mixing from the added turbulent kinetic energy of added diluent mass should be nearly same for both diluents: air or N2 due to similar transport properties [43]. However, O2 present in air addition might have an additional chemical effect via its participation in the kinetics of soot formation and oxidation, as was seen in a fundamental non-premixed methane combustion study with O2 addition [43]. Comprehensive assessment of this additional chemical effect would require a separate N2 addition study in the same engine, with same operating conditions, injectors, same PM measurement 66  principle, and possibly heating of N2 to account for the thermal effect from higher adiabatic flame temperatures with O2 addition. This was not pursued in this project.          Figure 4.10 Effect of FDR on PM emissions at various Φ and (a) EGR= 25% (b) EGR= 0% (c) EGR= 25% with PM emissions in log scale (d) EGR= 0% with PM emissions in log scale Note: The ‘-v-’ marker, overlapping with ‘-o-’ marker, represents mean values. The detection limit of PM in the current study is 0.04 mg/m3. Adjustment of the exhaust sampling dilution ratio was done to keep the PM concentration above the detection limit at lower concentration.    Similar to PM emissions, significant reductions in CO (45% to 95% reductions) emissions were obtained by varying FDR at all considered Φ and EGR (B.4). This was also seen with N2 addition [23], where it was attributed to significant improvements in the late cycle mixing. It is (c) (d) (a) (b) Measurements below the detection limt  67  also possible that, compared to N2 addition, the localized O2 in air addition further contributes to CO oxidation by enhancing the CO to CO2 kinetics by increasing oxygen availability.  A strong correlation (R2= 0.89) of CO with PM emissions was observed (Figure 4.11). This correlation was also observed with a range of other in-cylinder strategies in previous PIDING engine studies [27,34,37,82], although deviation from this behavior was also reported, such as that by Patychuk [27]. The fact that this correlation remained strong even with air addition indicates the absence of major deviations (with air addition) from the typical formation mechanisms of CO and soot in PIDING engine combustion. Although a clear understanding of the underlying mechanism for this correlation hasn’t been achieved, some possible causes include: a common precursor to CO and soot, overlap of soot and CO formation zones, common oxidation mechanisms of CO and PM (e.g. enhanced late cycle oxidation rates).   Figure 4.11 CO-PM correlation  4.3.2 THC The effect of FDR on THC (68 to 82% of THC in this study is CH4) emissions at various EGR and Φ is shown in Figure 4.12. Air addition was found to reduce THC by a factor of 4 at the y = 0.013x - 0.0302R² = 0.892900.10.20.30.40.50 10 20 30 40PM mass  (g/kWhr)CO (g/kWhr)68  highest Φ and EGR. The effect of FDR on THC emissions was dependent on the EGR, with higher EGR (12.5%, 25%) resulting in more pronounced reductions (60 to 85%), where the corresponding CH4 reductions were also of the same order (i.e. 60 to 85%) at the highest EGR. Furthermore, at 25% EGR, the THC emissions obtained with 100% FDR were observed to be independent of Φ. Since Φ= 0.77 had the highest baseline THC emissions, this indicates that the effectiveness of air addition to reduce THC was higher at high Φ. FDR was found to result in negligible reductions of THC at 0% EGR as the THC emissions were already low. In previous studies [22,37], THC emissions were found to decrease with increasing Φ, even at EGRs of 25%, where it was attributed to be due to increased combustion temperatures. The fact that a contradictory trend is seen at the highest EGR in the current study, even though both the measured exhaust temperatures and peak mixture average temperature increased with Φ (Appendix B.5), suggests that a competing mechanism (e.g. localized flame extinctions) of THC formation may be more dominating at higher EGR.  Figure 4.12 Effect of FDR on THC at various EGR and Φ The mechanisms for THC reduction with air addition are considered in context with previous fuel dilution studies [23,43]. Although THC reductions were demonstrated with N2 69  addition [23], it was argued that the primary source of the THC reductions were not related to improved combustion stability (via improved mixing from N2 addition) from N2 addition, but because of a fuel-volume effect from a fixed volume of unburned diluted fuel (i.e. from  undermixing, discussed in Section 2.1.3.3). This effect considers the dilution of a fixed-volume of unburned fuel, possibly trapped in injector sac and nozzle passages [23], and can be seen from the consistent linear THC reductions obtained with N2 addition (at various combustion timings). Since no such consistent linear reduction in THC emissions were seen with air addition (Figure 4.12), the fuel-volume effect is not a significant cause of the obtained reductions with the current engine/injector. Furthermore, with THC reductions with N2 addition not attributed to improved mixing [23], it is unlikely that improved mixing would be a significant cause of THC reduction with air addition [23]. This, therefore, indicates the major cause for the THC reductions obtained at 25% EGR to be the presence of an additional effect with air addition (e.g. from O2 species). This may be a result of air addition influencing the late cycle combustion duration, where the possibility of an additional effect from O2 species is also noted at higher EGRs (Section 4.2.2). Although not conclusive, this effect contributes to reduced THC most likely by affecting any of the three THC sources identified for compression ignition engines in literature (Section 2.1.3.3): reduced overleaning, reduced localized flame extinctions, and reduced bulk quenching of fuel-air mixture at the end of the expansion stroke.  70  4.3.3 NOx   Figure 4.13 Effect of FDR on NOx emissions at various Φ and (a) EGR= 25% (b) EGR= 0%  The effect of FDR on NOx emissions at various Φ and EGR are shown in Figure 4.13. A NOx penalty is observed with air addition as NOx increased by a factor of up to 2.3 times at 25% EGR and Φ = 0.58. The smallest increase (~1.5) was obtained for the highest Φ and was similar for all EGRs, indicating that a high Φ (i.e., higher loads) is more attractive for air addition. The observed NOx increase was not noted for N2 addition [23], where NOx remained nearly constant, or decreased slightly at the highest dilution. There, the decrease in NOx was lower than expected based on the reduced flame temperatures from fuel dilution [23], and the changes in mixing rates resulting in increased local residence times were assumed to be the cause.  The mechanism of the NOx increase seen with air addition is considered by eliminating unlikely mechanisms from those identified in literature (Section 2.1.3.2). NOx increases via the fuel-bound N route is unlikely since chemically bound nitrogen is unaffected with air addition. N2O route is favored in combustion conditions unfavorable to prompt and thermal NOx, which as discussed below, is not the case with air addition. The high temperatures obtained in the combustion environment of diesel and PIDING engines often make thermal route the dominant pathway, where an exponential dependence of specific NOx emissions (and emission index) with (a) (b) 71  AFT has been seen over a wide range of intake O2 concentrations in both diesel [46] and PIDING engines [49]. This relationship was also seen with air addition at high EGRs (≥ 12.5%) (Figure 4.14), indicating that increases in the flame temperature affecting the thermal route was most likely the major contributor to increased NOx for air addition at these EGR rates. The high FDR points appear to produce more NOx than that expected from computed AFT; however, it should be recalled that the AFT calculation doesn’t consider the effect of soot radiation, so the true flame temperature for the higher FDR cases should be higher (promoting more NOx) than the computed AFT. Air addition may also enhance prompt NO formation at the jet core (e.g. due to increased CH radical production), however, this effect is unlikely to be enhanced at lower EGR rates, where the higher flame temperatures would further support thermal NOx. Similarly, it is unlikely that a reduction in EGR rates, would significantly enhance other potential ways of thermal NOx (O and N2 species; residence time). Therefore, it is very likely that increases in flame temperature was also the major contributor to increased NOx at low EGRs.   Figure 4.14 Correlation of NOx emissions index with reciprocal of representative adiabatic flame temperature at various EGR and FDR   EGR= 25% EGR= 12.5% EGR= 0% 72  The effect of FDR, and EGR on the PM-NOx tradeoff is shown in Figure 4.15. The measured PM and NOx emissions at an operating point is defined by three control parameters: FDR, EGR, and Φ, which is shown in some cases as three overlapping symbols. In many instances, 1 or 2 control parameters are omitted (e.g. Φ= 0.58, Φ= 0.68, EGR= 12.5%, and FDR= 10 and 25%) to improve clarity.  For such cases, an understanding of the operating space at a particular PM-NOx emissions value can be achieved by observing the effects of each of the control parameters indicated by arrows in the figure. For all fuel dilution ratios, increasing EGR leads to increasing PM emissions while decreasing NOx. A similar effect was also seen with increasing Φ, although the PM increase was more significant. Increasing EGR at higher FDR, resulted in substantial NOx reductions with lower increase in PM emissions. The lowest PM, while achieving near minimum NOx emissions, was found to be at the highest tested EGR (25%), Φ (0.77), and FDR (100%), and therefore is considered as the optimum solution (encircled) from the PM-NOx tradeoff.      Figure 4.15 PM-NOx trade-off as a function of FDR and EGR at Φ= 0.77   Optimum 73  4.4 Expansion Work of Diluent The expansion of injected high-pressure diluent (air) is expected to contribute to the gross indicated work and should increase with FDR. A model based on the first law of thermodynamics was developed for the estimation of this diluent expansion work (𝑊𝑑𝑖𝑙). The underlying assumption of the model is that 𝑊𝑑𝑖𝑙 for a fired cycle in SCRE would be equal to 𝑊𝑑𝑖𝑙 for an equivalent motored cycle. The equivalent motored cycle was a 2 stroke (compression and expansion) cycle with the same thermodynamic conditions as the corresponding fired cycle at that start of compression stroke. It was modelled in 0-dimension with adiabatic walls. The diluent was injected with the same injection rate as that during operation, starting at the gaseous start of injection (GSOI), and over the duration of GPW. The conservation of energy for this system is described as follows:  𝑑𝑄𝑑𝜃 + ∑𝑑𝑚𝑖𝑑𝜃(ℎ𝑖 +𝑉𝑖22+ 𝑧𝑖𝑔)=𝑑𝑊𝑑𝜃+ ∑𝑑𝑚𝑜𝑑𝜃(ℎ𝑜 +𝑉𝑜22+ 𝑧𝑜𝑔) +  𝑑(𝑚𝑢)𝑑𝜃 (4.3) where 𝑑𝑄𝑑𝜃 ~0 is wall heat transfer rate for adiabatic walls, 𝑑𝑚𝑖𝑑𝜃 is mass flowrate of the diluent added, ℎ𝑖 +𝑉𝑖22= ℎ𝑑𝑖𝑙 is the total enthalpy of the diluent, 𝑑𝑚𝑜𝑑𝜃~0 is the mass flowrate of the outgoing fluid, 𝑧𝑖𝑔~0 is the potential energy which can be neglected, m is the mass of the gas in the cylinder, u is the specific internal energy of the gas in the cylinder. Equation (4.3) further reduces to the following equation:  𝑑𝑚𝑑𝜃 ℎ𝑑𝑖𝑙 − p𝑑𝑉𝑑𝜃=𝑑𝑇𝑐𝑦𝑙𝑑𝜃𝐶𝑣𝑚 +𝑑𝑚𝑑𝜃𝑢 (4.4) 74  where 𝑉 is the volume of the cylinder, 𝑇𝑐𝑦𝑙 is the average cylinder gas temperature, and 𝐶𝑣 is the specific heat capacity at constant volume.  Differentiating the ideal gas equation, gives the following:   𝑑𝑝𝑑𝜃𝑉 +𝑑𝑉𝑑𝜃𝑝 = (𝑅𝑑𝑚𝑑𝜃𝑇𝑐𝑦𝑙 + 𝑚𝑅𝑑𝑇𝑐𝑦𝑙𝑑𝜃) (4.5) where 𝑝 is the in-cylinder pressure, and R is the specific gas constant of air ~ 287 (J kg−1 K−1). Assuming an adiabatic nozzle, the total enthalpy across an injector nozzle would remain conserved which results in ℎ𝑑𝑖𝑙 being a function of upstream stagnation temperature of the diluent (𝑇𝑑𝑖𝑙) i.e.  ℎ𝑑𝑖𝑙 =  𝑢𝑑𝑖𝑙 + 𝑅𝑇𝑑𝑖𝑙 Solving, Equation (4.4) and (4.5) for 𝑑𝑝𝑑𝜃 results in the following:   𝑑𝑃𝑑𝜃= (1 − 𝛾)𝑑𝑚𝑉𝑑𝜃(∆𝑢 − 𝐶𝑣𝑇𝑐𝑦𝑙 − 𝑅𝑇𝑑𝑖𝑙) −(𝛾 − 1)𝑃𝑉𝑑𝑉𝑑𝜃(1 +𝐶𝑣𝑅) (4.6) where 𝛾 = 𝐶𝑣𝑅+ 1 is the ratio of specific heats, and ∆𝑢 = ∫ 𝐶𝑣𝑑𝑇𝑇𝑐𝑦𝑙𝑇𝑑𝑖𝑙 is the specific internal energy.  𝑑𝑃𝑑𝜃 was integrated to obtain the pressure trace (𝑝𝑑𝑖𝑙(𝜃)) for the equivalent motored cycle, and therefore 𝑊𝑑𝑖𝑙 can be calculated as follows:  𝑊𝑑𝑖𝑙 = ∑ 𝑝𝑑𝑖𝑙(𝜃)𝑑𝑉(𝜃)𝜃= 180𝑜𝐴𝑇𝐷𝐶𝜃= −180𝑜𝐴𝑇𝐷𝐶 (4.7) The expansion power of diluent can therefore can be calculated as follows:  ?̇?𝑑𝑖𝑙 =𝑊𝑑𝑖𝑙𝑡𝑐𝑦𝑐 (4.8) where 𝑡𝑐𝑦𝑐 is the time taken to complete the corresponding fired cycle.  75  The key assumptions can be summarized as follows:  • Ideal gas and adiabatic walls • Temperature remains uniform inside the cylinder • Adiabatic injector nozzle, total enthalpy across injector nozzle remains conserved • Reservoir conditions at the mixing chamber upstream of injector • Changes in potential energy is negligible The initial conditions at BDC of the equivalent motored cycle were calculated based on measured mass (function of intake air flow, recirculated exhaust flow, engine speed, and residual gas fraction) and a suitable temperature equal to the mean of the intake manifold temperature and the coolant temperature. The temperature was used instead of pressure because of the possibility of error in the absolute value of measured pressure at BDC (low signal to noise ratio). Although temperature was not measured at BDC, the suitability of the chosen temperature was ensured via a sensitivity analysis which showed that varying temperature in the range of 56oC (measured intake manifold temperature) and 80oC (coolant temperature) resulted in no significant effects on ?̇?𝑑𝑖𝑙 (Appendix B.4).  Effects of parameters like 𝑇𝑑𝑖𝑙, GPW, and GSOI on ?̇?𝑑𝑖𝑙 were determined by sensitivity analysis. 𝑇𝑑𝑖𝑙 should be equal to the temperature at the mixing chamber (Figure 3.2). This temperature would be relatively colder with respect to other engine relevant temperatures like coolant (80 oC), intake manifold (50 oC) etc., due to repeated expansions of the diluent from buffer tank to mixing chamber, and therefore 𝑇𝑑𝑖𝑙 in the range 10oC to 30oC was considered. A sensitivity analysis revealed ?̇?𝑑𝑖𝑙 to be insensitive to 𝑇𝑑𝑖𝑙 in the range 10oC to 30oC (Appendix B.4). Since 𝑇𝑑𝑖𝑙 was not measured in this study, 𝑇𝑑𝑖𝑙= 20 oC was chosen for all calculations. Effects of GPW 76  (measured ±1 ms) and GSOI on ?̇?𝑑𝑖𝑙 (measured ±10 ms) were also found to be insignificant (less than 0.03 kW change in ?̇?𝑑𝑖𝑙). The effect of FDR on ?̇?𝑑𝑖𝑙 at various Φ and EGR= 25% is shown in Figure 4.16. ?̇?𝑑𝑖𝑙 is observed to be primarily dependent on FDR at various engine operating conditions. For the considered operating mode, the maximum FDR resulted in ?̇?𝑑𝑖𝑙= 0.36 kW, which contributes to less than 1% increase in the gross indicated power (47 kW).   Figure 4.16 Effect of FDR on expansion power of diluent at various Φ and EGR= 25%  4.5 Gross Indicated Efficiency The effect of FDR on the gross indicated efficiency (𝜂𝑖,𝑔) was evaluated at various EGR and Φ as shown in Figure 4.17. Air addition resulted in a 2-2.5% increase in gross indicated efficiency (𝜂𝑖,𝑔), with the larger improvements being noted at high loads and Φ. These increase in 𝜂𝑖,𝑔 were attributed to improved late cycle mixing in the previous N2 addition study [23]. Beyond this, there are several other mechanisms which may increase 𝜂𝑖,𝑔, for example, the expansion power of diluent (?̇?𝑑𝑖𝑙), improved combustion efficiency (𝜂𝑐), changes in heat transfer, and changes in conversion of heat to work (e.g. HRR phasing). As discussed in Section 4.4, ?̇?𝑑𝑖𝑙 even at the highest FDR increased the gross indicated power by less than 1%, which in terms of 𝜂𝑖,𝑔 77  would contribute to less than 0.5%-point improvement in 𝜂𝑖,𝑔 (at 100% FDR). For clarity and ease of understanding, the effect of ?̇?𝑑𝑖𝑙 on 𝜂𝑖,𝑔 was not separately considered in the mechanistic descriptions below.  To quantify the contribution of 𝜂𝑐 to the increase in 𝜂𝑖,𝑔, 𝜂𝑖,𝑔 was considered as:    𝜂𝑖,𝑔 = 𝜂∗𝜂𝑐     where 𝜂∗ =  𝑊𝑖,𝑔𝑄𝑐𝑜𝑚𝑏 is referred to as the combustion independent indicated efficiency, and is the fraction of actual energy from combustion (𝑄𝑐𝑜𝑚𝑏) that is converted to indicated work (𝑊𝑖,𝑔). 𝜂𝑐= 𝑄𝑐𝑜𝑚𝑏𝑄𝐿𝐻𝑉  (𝑄𝐿𝐻𝑉: energy content of the consumed fuel based on lower heating value) is the combustion efficiency calculated as described in Section 3.1.5. The sources of inefficiencies in 𝜂∗ include heat transfer changes, exhaust enthalpy changes, and conversion of heat to work, but not incomplete fuel conversion.       78   The effect of FDR on 𝜂𝑖,𝑔 and 𝜂∗ at various Φ and EGR are shown in Figure 4.17a. At higher EGR, 𝜂∗ did not increase significantly with increasing FDR, indicating that a very significant portion of the 𝜂𝑖,𝑔 improvements observed at higher EGR were due to improved 𝜂𝑐. For operation without EGR (Figure 4.17b), 𝜂𝑐 was not affected significantly (Figure 4.6), and therefore the improvement of 𝜂𝑖,𝑔 with increasing FDR at low EGR must be due to some other mechanism. This mechanism is elucidated by understanding the effects of FDR on the exhaust gas enthalpy.   A representative rate of exhaust enthalpy (?̇?𝑒𝑥ℎ) was calculated to evaluate the possibility of reductions in exhaust energy being an underlying mechanism of improvements in 𝜂𝑖,𝑔 with air   Figure 4.17 Effect of FDR on gross indicated efficiency and combustion independent indicated efficiency at various Φ and (a) EGR= 25% (b) EGR= 12.5 % (c) EGR= 0% (a) (b) (c) 79  addition at low EGR. This was calculated based on cylinder exhaust temperature measurement (𝑇𝑒𝑥ℎ) with the reference temperature (𝑇𝑟𝑒𝑓) taken as 278 K as follows:   ?̇?𝑒𝑥ℎ = ?̇?𝑒𝑥ℎ. ∫ 𝑐𝑝(𝑇)𝑑𝑇𝑇𝑒𝑥ℎ𝑇𝑟𝑒𝑓 where ?̇?𝑒𝑥ℎ is the measured exhaust mass flowrate and 𝑐𝑝 is the specific heat capacity of exhaust (assumed as air) at constant pressure. Although a more accurate estimation of the exhaust enthalpy would require knowledge of instantaneous exhaust mass flowrate, this calculated parameter should provide a reliable indication of exhaust enthalpy affecting the gross indicated work. This is because the EVO (exhaust valve opening) occurs within the expansion stroke (150o ATDC), and the instantaneous exhaust enthalpy at this point would majorly reflect in the ?̇?𝑒𝑥ℎ. The exhaust enthalpy is then normalized with respect to fuel energy to calculate the exhaust energy % (of fuel energy) and the results are discussed below.   The effect of FDR on the exhaust energy % at various EGR and Φ is shown in Figure 4.18. At the highest EGR, increasing FDR lead to insignificant changes in exhaust energy % verifying that exhaust enthalpy changes didn’t contribute significantly at high EGR. This behavior, however, changes significantly at reduced EGR rates, with up to 2%-point reduction in exhaust energy % at 0% EGR, indicating that reductions in exhaust enthalpy due to air addition at low EGR rates is the most likely mechanism of 𝜂𝑖,𝑔 improvements with FDR at low EGR. However, the underlying mechanism of 𝜂𝑖,𝑔 improvement with decreasing Φ at 0% EGR is unclear and would require consideration of heat transfer changes and changes in conversion of heat to work for a conclusive answer.       80          c   Figure 4.18 Effect of FDR on exhaust energy % at various Φ and (a) EGR= 25% (b) EGR= 12.5 % (c) EGR= 0% (c) (a) (b) 81  4.6 Conclusions This chapter presents the effects of air addition on the emissions (especially PM and NOx) and performance of a PIDING engine at various EGR and Φ. The conclusions are given below:  • The majority of the combustion effects of air addition were similar to that seen in a previous N2 addition study: reduced intensity of partially premixed phase and more mixing dominated combustion (i.e. fraction of energy released in mixing-controlled phase relative to partially premixed phase). The early cycle combustion duration and combustion variability (COV of GIMEP) reduced, both of which were found to be also independent of EGR and Φ. • Several effects on combustion were noted that were not observed for N2 addition, particularly at higher EGR rates: increased adiabatic flame temperatures (AFT), increased overall combustion duration (resulting from unchanged late cycle combustion duration), increased maximum pressure rise rates, and increased combustion efficiency. • Denuded PM (majorly soot) reduced exponentially with air addition (90% to 97% reduction at 100% FDR). Similarly, CO was also reduced, albeit with a lower magnitude (45% to 95%). Except at the lowest EGR (0%), where PM reductions were significant till only 50% FDR, significant PM reductions were observed up to the maximum considered FDR (100%). Similar to N2 addition, improved late cycle mixing resulting from additional turbulent kinetic energy of the diluent is attributed as a cause of these reductions. An additional chemical effect resulting from O2 species in air addition may also contribute to these reductions.   • The THC (68 to 82% is methane) and corresponding unburnt CH4 emission reductions with air addition were significant only at EGR> 12.5% (60 to 84% reductions in THC and CH4 obtained at 25% EGR), where operation with 0% FDR would result in significant CH4 and THC emissions. Also, these reductions were most significant at the highest Φ (0.77). An additional 82  effect of air addition (e.g. from O2 species), relative to N2 addition, which becomes significant at higher EGR rates is attributed for these reductions. • NOx emissions increased with air addition (up to a factor of 2.3), unlike with N2 addition where NOx decreased slightly. The lowest factor of increase (~ 1.5) was obtained at the highest Φ (0.77) and was nearly equal at all EGRs. An increase in flame temperature affecting the thermal NO is attributed as the major contributor to increased NOx obtained with air addition at high EGR rates (≥ 12.5%). The same route is hypothesized to be the major contributor to NOx increase even at low EGR rates.  • Indicated efficiency (𝜂𝑖,𝑔) improvements of the order of 2.5%-point with air addition were observed across a range of tested EGR and Φ. The expansion power of diluent contributed to less than 0.5%-point increase in 𝜂𝑖,𝑔. The indicated efficiency (𝜂𝑖,𝑔) improvements at high EGR resulted from combustion efficiency improvements, while the most likely mechanism for the improvements observed with increasing FDR at the lowest EGR is due to reductions in exhaust enthalpy.  83   Development and Characterization of Diluent Compressor A critical aspect of the system level assessment of air addition is the development of suitable diluent (air) compression strategies (discussed in Section 3.2.1). To evaluate the compression power requirements, the power and flowrate were characterized for two compressions systems in Section 5.2: the 3-stage compressor configuration, and a 2-stage configuration. The data obtained is used to develop response surfaces for system level analysis in Chapter 6. In Section 5.5, the performance of both 2-stage and 3-stage configurations were evaluated to compare the compression strategies, and to identify causes of significant inefficiencies.  5.1 Methodology The motor power consumed, and the outlet flow of the compressor configurations were measured at various operating points. The operating point for the 2-stage configuration (prototype) was defined by the compressor outlet and inlet pressures (delivery pressure) (Section 3.2.1; Figure 3.5), while that of 3-stage configuration (Figure 3.4) was defined only by the outlet pressure.  The pressure for the inlet of 2-stage configuration was chosen to be representative of typical intake manifold pressures. The delivery pressures were considered up to 320 bar since air addition requires diluent to delivered with pressure aligned with the fuel pressure (300 to 320 bar). For characterization of 2-stage configuration, a range of inlet pressure were considered at each delivery pressure (Table 5.1). A repeatability point, shown in bold in Table 5.1, was taken at the start and end of an inlet pressure sweep at an outlet pressure. This was done to ensure compressor operation within reasonable variability, before proceeding to the next outlet pressure. The 3-stage characterization was followed by repeated testing of 4 points, highlighted in bold for 3-stage 84  configuration in Table 5.1, on 3 different days. These repeat points indicated no significant changes in day to day system variability. The test matrix is presented in Table 5.1: Table 5.1 Compressor operating set points. Note: The repeatability points of respective configuration are highlighted in bold. Configuration Outlet pressure (bar) Inlet pressure (bar) 2-stage 75    [1.5, 2, 2.5,3, 3.5] 150 200 250 300 320 3-stage [75, 100, 150, 200, 250, 280, 300, 320] Atmospheric  5.2 Power Consumption and Output Flowrate    Figure 5.1 Effect of compressor inlet and outlet pressure on (a) compressor outlet mass flowrates and (b) compressor motor power consumption, for 2-stage and 3-stage compressor configurations.  Note: The circles represent measurements while the lines represent the response surfaces fitted to those measurements (see Chapter 6 for the development of response surfaces)  The effect of compressor inlet (only for 2-stage) and outlet pressures on compressor outlet mass flowrates are shown in Figure 5.1a. Much lower mass flowrates were observed in the 2-stage configuration relative to the 3-stage configuration. The 2-stage inlet pressures provided for characterization were nearly half than that obtained in 2nd stage inlet in regular 3-stage operation (a) (b) 80% FDR 50% FDR Air addition pressures 85  pressures (Note: The inlet of the 2-stage configuration was the inlet of the 2nd stage of 3-stage configuration as described in Section 3.2.1), and therefore had significantly lower inlet density in 2-stage, resulting in reduced charge mass per swept volume i.e. reduced mass flowrate. Increasing the outlet pressure resulted in reduced outlet flowrates because increasing the stage outlet pressure, at the same stage inlet pressures would result in increased mass of stage residuals, and thereby result in less displaced mass per stroke.  The FDR achievable with the 2-stage configuration is shown in Figure 5.1a. These were calculated for the natural gas flowrates required for the engine mode tested in Chapter 4. For the flowrates obtained in the 2-stage configuration, at relevant air addition pressures (> 280 bar), the 2-stage configuration would be able to provide FDR only up to 80%. The 3-stage configuration could easily achieve up to 150% FDR.   The effect of compressor inlet and outlet pressure on compressor motor power is shown in Figure 5.1b. Increasing the inlet pressure in 2-stage configuration resulted in increased power consumption, due to the increased mass flowrate. At all outlet pressures, significantly lower motor power is consumed in 2-stage configuration. This should not be interpreted as ‘superior performance of 2-stage’, as there was also corresponding reduction in mass flowrates. The effect of increasing system pressure ratio leading to increased power consumption can be seen at each inlet pressure of 2-stage configuration. These effects can be qualitatively understood in the context of polytropic compression process, and therefore polytropic indices at various operating conditions were determined, as discussed in the next section. 86  5.3 Polytropic Index  For each stage, polytropic indices (𝑛) were determined to identify a reference process for analyzing the compressor performance (Section 5.5). This was done using Equation (2.4) from Section 2.2.1 and results are discussed below.   Figure 5.2a shows significant variation in the 𝑛 (Equation (2.4)) of 2-stage configuration obtained by varying inlet pressures (𝑝𝑖𝑛). This contradicted with previous work by S. Motta et al. [61] in reciprocating hermetic compressors, where no monotonous trends in 𝑛, calculated with similar methodology, were seen with varying mass flowrate, inlet pressure, and pressure ratio. A variation of 𝑛 from near isothermal (𝑛 ~1) to near adiabatic (𝑛 > 1.3) for the same compressor stage at varying operating conditions is highly unlikely. Interestingly, 𝑛 was also observed to be a function of mass flowrate (Figure 5.2b), with the 𝑛 at the highest mass flowrate (at 3 stage configuration) indicating near adiabatic heat transfer as is typically expected in reciprocating compressors. Although the actual reason for this is not conclusive, some possible causes include higher inaccuracies in representative stage outlet temperature at low flowrates due to slower heat transfer to the thermocouple junction, exhaust port heat transfer effects etc. The accuracy of thermocouple measured temperature being dependent on operating conditions was also observed in engine exhaust temperature by Caton [83], where it was observed there that the thermocouple measured equilibrium exhaust temperatures didn’t necessarily correlate with the mass weighted average exhaust temperature and the peak gas averaged temperatures [83]. Due to the unreliability of the compressor stage outlet temperature measurements, parameters derived based on it such as 𝑛 were not used to make conclusions in this study. Therefore, as a reference process, 𝑛= 1.4 of an adiabatic reversible or isentropic compression was used.  87    Figure 5.2 Variation of n of the last stage of respective configurations with (a) compressor inlet and outlet pressures and (b) compressor outlet mass flowrate Note: 𝑝𝑜𝑢𝑡 , 𝑝𝑖𝑛 , and n are compressor outlet pressure, 2nd stage inlet pressure, and polytropic index, respectively.   5.4 Friction Estimation Direct measurement of compressor friction power was not conducted in this study because of the lack of relevant experimental resources like dynamometer, piezo-electric transducers inside compressor stages, and crank encoders. However, due to the relevance of friction for calculating critical performance parameters like mechanical efficiency (Section 5.5.3), two independent methods: the overlap method and the motor efficiency method, were developed to estimate friction. The methodology of these respective methods are described in Sections 5.4.1 and 5.4.2, while the combined analysis is presented in Section 5.4.3. 5.4.1 Overlap Method The overlap method is based on the idea that the estimated friction is the common overlapping region of calculated possible friction ranges of 2-stage and 3-stage configurations (Figure 5.3). The possible friction ranges were in turn calculated based on nameplate motor efficiency, a literature-based belt efficiency, range of compressor isentropic efficiencies (𝜂𝑖𝑠𝑒𝑛) (a) (b) 88  observed in literature (discussed in Section 2.2), measured motor power, and measured isentropic power (using Equation (3.11)).   Figure 5.3 Schematic representation of friction estimation based on overlap method  A thermodynamic approach forms the basis of the overlap method, and therefore for clarity and context, the various energy and work transfers are shown in Figure 5.4. The steady state energy equation for control volume CV1 (enclosing pistons, and compressor shaft), as depicted in Figure 5.4, was solved to relate the friction power (𝑃𝑓) to the sum of indicated power of the compressor stages (∑ ?̇?𝑖𝑘𝑖=1 ; k= total number of stages) and the compressor shaft power (𝑃𝑠ℎ𝑎𝑓𝑡), as follows in Equation (5.1):   𝑃𝑓 =   𝑃𝑠ℎ𝑎𝑓𝑡 −  ∑ ?̇?𝑖𝑘𝑖=1   (5.1) The shaft to motor power relations (Section 3.2.3), and the typical definition of 𝜂𝑖𝑠𝑒𝑛 (Section 2.2.1), were included in Equation (5.1) to result in the following:   𝑃𝑓 =   𝜂𝑚𝑜𝑡𝑜𝑟𝜂𝑏𝑒𝑙𝑡𝛼𝑃𝑚𝑜𝑡𝑜𝑟 −  ∑ ?̇?𝑖𝑠𝑒𝑛,𝑖𝑘𝑖=1𝜂𝑖𝑠𝑒𝑛  (5.2) where 𝜂𝑚𝑜𝑡𝑜𝑟 is taken as 0.88 (nameplate value), 𝜂𝑏𝑒𝑙𝑡 is taken as 0.9 typical for a V-belt of the configuration used in this study [71], 𝛼 ∈ [0.9, 1.1] is a sensitivity parameter to account for 89  inaccuracies in 𝜂𝑚𝑜𝑡𝑜𝑟 and 𝜂𝑏𝑒𝑙𝑡, ?̇?𝑖𝑠𝑒𝑛,𝑖 is isentropic work of ith stage (Section 3.2.2), and 𝜂𝑖𝑠𝑒𝑛 ∈[0.56, 0.83] are typical range of isentropic efficiencies found in literature (Table 2.2). Therefore, 𝛼= 1.1 and 𝜂𝑖𝑠𝑒𝑛= 0.83, were used to determine the upper limit of  𝑃𝑓, while the lower limit was determined by 𝛼= 0.9 and 𝜂𝑖𝑠𝑒𝑛= 0.56. This method was done for both compressor configurations. It is expected that friction in a reciprocating compressor is predominantly a function of piston speed, similar to engines [46,55,84]. Since the speed was constant for both compressor configurations, the friction power of both compressor configurations was assumed approximately equal. The overlapping region of the ranges of friction power (based on chosen values of 𝛼 and 𝜂𝑖𝑠𝑒𝑛) of both compressor configurations, therefore, would represent an estimate of the compressor friction using this method. The major assumption with this method is that the friction for both the configurations is nearly same and is insensitive to average pressure changes seen between the configurations.     90   Figure 5.4 Schematic depicting work and energy transfer in various control volumes of the compressor in 3-stage configuration. Note:  ?̇?𝑖: indicated power of the ith stage; ?̇?𝑤,𝑖: wall heat transfer of the ith stage; ?̇?𝑐,𝑖: heat lost from the ith coolers; ?̇?𝑖, ?̇?𝑒 and ?̇?𝑏 are mass flowrate of inlet, exit, and crankcase blowby, respectively; ℎ𝑡,𝑖, ℎ𝑡,𝑒, and ℎ𝑡,𝑏 are total specific enthalpy of inlet, exhaust, and crankcase blowby, respectively  5.4.2 Motor Efficiency Method The motor efficiency method is based on the idea that the shaft power consumed by a compressor at unloaded operation (i.e. all stages venting to atmosphere) is equal to the friction power. Since the compressor shaft power can be related to the measured motor power using known motor and belt efficiencies (Section 3.2.3), the friction power (𝑃𝑓) can be estimated using:   𝑃𝑓 = 𝜂𝑏𝑒𝑙𝑡. 𝜂𝑚𝑜𝑡𝑜𝑟 . 𝑃𝑚𝑜𝑡𝑜𝑟 (5.3) The critical factor determining the accuracy of this method is the accuracy of the motor efficiency at unloaded operation. The power consumed by the compressor motor in an unloaded operation is much lower than 50% of its peak load, a  regime where motor efficiencies are known to be much lower than their nameplate efficiency [70]. Therefore, 𝜂𝑚𝑜𝑡𝑜𝑟 was experimentally 91  determined using the general slip method (Equation (3.13)) discussed in Section 3.2.3 and the results are shown in Figure 5.5.  Figure 5.5 shows that 𝜂𝑚𝑜𝑡𝑜𝑟 is within ± 2% points of nameplate motor efficiency (𝜂𝑚𝑜𝑡𝑜𝑟,𝑟) at loads > 75%.    Figure 5.5 Motor efficiency variation with Load  The friction power was estimated using a modification of the general slip method equation (Equation (3.13)) to reduce dependencies on rated parameters (speed, voltage, and power), which can add significant errors due to manufacturer tolerances [70]. The modification was achieved by referencing the parameters at a particular primary load (𝜂𝑚𝑜𝑡𝑜𝑟,1, 𝑆𝑚,1, 𝑃𝑚𝑜𝑡𝑜𝑟,1, 𝑉1), with respect to the measured parameters at another reference load (𝜂𝑚𝑜𝑡𝑜𝑟,2, 𝑆𝑚,2, 𝑃𝑚𝑜𝑡𝑜𝑟,2, 𝑉2), instead of with respect to rated parameters (𝜂𝑚𝑜𝑡𝑜𝑟,𝑟, 𝑆𝑟, 𝑃𝑟, 𝑉𝑟) as is used in Equation (3.13) of the general slip method. Equation (3.13) is rewritten as Equation (5.4):   𝜂𝑚𝑜𝑡𝑜𝑟,1𝜂𝑚𝑜𝑡𝑜𝑟,2=𝑃𝑚𝑜𝑡𝑜𝑟,2𝑃𝑚𝑜𝑡𝑜𝑟,1𝑆𝑠 − 𝑆𝑚,1(𝑆𝑠 − 𝑆𝑚,2)(𝑉2𝑉1)2× 100% (5.4) The primary loads (Load = 𝜂𝑚𝑜𝑡𝑜𝑟𝑃𝑚𝑜𝑡𝑜𝑟𝑃𝑟) were chosen to correspond to unloaded compressor operation, while the reference loads were loads at which the motor efficiency (𝜂𝑚𝑜𝑡𝑜𝑟,2), determined using the general slip method, was within ± 2%-points of nameplate motor efficiency  92  (𝜂𝑚𝑜𝑡𝑜𝑟,𝑟). As seen in Figure 5.5, these are loads > 75%. Using assumed values of 𝜂𝑏𝑒𝑙𝑡 (0.9), measured values of 𝑃𝑚𝑜𝑡𝑜𝑟,1, and estimated values of  𝜂𝑚𝑜𝑡𝑜𝑟,1, the compressor friction power was estimated based on Equation (5.3). Since friction based on empirical relations for IC engine friction is expected to increase with in-cylinder pressure [46,55,84], the friction power estimated using this method is expected to be a representation of the minimum compressor friction as the compressor was unloaded.  5.4.3 Analysis The estimated friction power of both methods was compared to determine the reliability of these methods and to arrive at a reliable value of friction for subsequent calculations. Estimated friction power using the overlap method as a function of compressor outlet pressure is shown in Figure 5.6. For the chosen ranges of 𝜂𝑖𝑠𝑒𝑛 and 𝛼, the overlapping region of the friction power (𝑃𝑓) of 2-stage and 3-stage configurations lies in the range 1.1 to 1.5 kW. However, the lower limit values of 𝛼 and 𝜂𝑖𝑠𝑒𝑛 are unrealistic since the corresponding 𝑃𝑓 obtained for 3-stage was negative. This is also supported by the results from 2-stage configuration, where only the lower limit values of 𝛼 and 𝜂𝑖𝑠𝑒𝑛 showed significant variability at each outlet pressure, indicating a strong dependence on operating conditions. By increasing the lower limit of 𝑃𝑓 with 𝜂𝑖𝑠𝑒𝑛 = 0.65, the range of the estimated friction reduced to 1.3 to 1.5 kW (not shown in figure), and the results also became more realistic (i.e. no negative 𝑃𝑓 for 3-stage and less variability for 2-stage). The estimated 𝑃𝑓 with the motor method as a function of reference load is shown in Figure 5.7. It is in the range of 1.63 to 1.7 kW. This agrees approximately with the upper limit of estimated friction obtained with the overlap method, suggesting the reliability of these developed methods for 93  estimating friction within ball park ranges. For qualitative analysis presented in Section 5.5.1, the estimated friction power was taken as 1.5 kW.   Figure 5.6 Estimated friction power with overlap method as a function of third stage outlet pressure.  Note: Upper limit: (𝛼 = 0.9, 𝜂𝑖𝑠𝑒𝑛 =0.56); Lower limit: (𝛼 = 1.1, 𝜂𝑖𝑠𝑒𝑛 =0.83)    Figure 5.7 Estimated friction power with motor efficiency method as a function of reference load.  5.5 Compressor Performance This section evaluates the compressor performance of both configurations based on volumetric efficiency (𝜂𝑣,𝑚), mechanical efficiency (𝜂𝑚𝑒𝑐ℎ), overall efficiency (𝜂𝑂𝐸), and specific power (?̂?𝑚,𝑠). The performance based on 𝜂𝑣,𝑚, 𝜂𝑚𝑒𝑐ℎ, and 𝜂𝑂𝐸 were contrasted with that seen in literature. Although similar contrast could not be provided for ?̂?𝑚,𝑠, detailed mechanistic descriptions were provided for understanding underlying behaviours.      5.5.1 Specific Power  The motor specific power is a useful parameter for evaluating the performance of a compressor from the perspective of its applicability for air addition and is defined below: 𝑃𝑚,𝑠 =𝑃𝑚𝑜𝑡𝑜𝑟?̇? 94  where 𝑃𝑚𝑜𝑡𝑜𝑟 is measured motor power and ?̇? is compressor outlet mass flowrate. In comparing 𝑃𝑚,𝑠 of different compressor configurations, a lower value would be an indicator of superior performance. The effect of total pressure ratio (𝑝𝑜𝑢𝑡𝑝𝑖𝑛) on 𝑃𝑚,𝑠 and specific isentropic power (𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛=1, Section 3.2.2) of both compressor configurations is shown in Figure 5.8a. For the 2-stage configuration, the 𝑃𝑚,𝑠 increased exponentially with 𝑝𝑜𝑢𝑡𝑝𝑖𝑛, and was also significantly higher than that of 3-stage. Interestingly, a comparison of their respective specific isentropic power (𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛=1; Section 3.2.2), which is the theoretical minimum indicated specific power, revealed nearly equal values for both configurations, with the 2-stage requiring marginally higher 𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛=1 as was expected due to increased compression power with reduced stages. Furthermore, a worst case estimate of specific indicated power (i.e. 𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛=𝑎  as defined in Section 3.2.2), based on the lowest 𝜂𝑖𝑠𝑒𝑛 (a= 0.56) seen in literature (Table 2.2), resulted in comparable 𝑃𝑚,𝑠 for both the configurations used, which were much lower than the 𝑃𝑚,𝑠 of 2-stage (Figure 5.8b). This indicates that the cause is unlikely due to poor isentropic efficiency.     The underlying cause for high 𝑃𝑚,𝑠 of 2-stage was investigated further by considering it as the sum of its two individual components as follows:   𝑃𝑚,𝑠 = 𝑃𝑖,𝑠 + 𝑃𝑙,𝑠 (5.5) where 𝑃𝑖,𝑠 is the specific indicated power consumption of the compressor (Section 3.2.2), and 𝑃𝑙,𝑠 is the specific loss power consumption. 𝑃𝑙,𝑠 includes losses like motor losses, belt transmission losses, friction losses i.e. losses exclusive of that included in indicated power. Since 𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛=0.56 was nearly comparable for both configurations, the most likely cause is an exponentially rising 𝑃𝑙,𝑠 95  in 2-stage. This was demonstrated by analyzing the sum of 𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛=1 and the major component of 𝑃𝑙,𝑠 i.e. specific friction power (𝑃𝑓,𝑠), where 𝑃𝑓,𝑠 =𝑃𝑓?̇? with 𝑃𝑓 taken equal to 1.5 kW based on its estimated value in Section 5.4. This parameter (𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛=1+ 𝑃𝑓,𝑠), shown in Figure 5.9, closely followed the 𝑃𝑚,𝑠 behaviors observed for the respective configurations. This demonstrates that the exponential behavior at high 𝑝𝑜𝑢𝑡𝑝𝑖𝑛 observed in 2-stage configuration derives from 𝑃𝑓,𝑠. Since 𝑃𝑓 is nearly equal for both configurations, the exponential rise of 𝑃𝑚,𝑠 of 2-stage configuration at high 𝑝𝑜𝑢𝑡𝑝𝑖𝑛 is due to corresponding lower mass flowrates, which result in the exponential increase of 𝑃𝑓,𝑠. Therefore, the inferior performance of 2-stage relative to 3-stage was inherently because of reduced mass flowrates in 2-stage, which were a result of operating the 2-stage configuration with reduced inlet pressures.       Figure 5.8 Effect of total pressure ratio on (a) specific motor power and specific isentropic power and (b) specific isentropic power and specific indicated power.  Note: assumed 𝜂𝑖𝑠𝑒𝑛 = 0.56; 𝒑𝒐𝒖𝒕𝒑𝒊𝒏: total pressure ratio (a) (b) 96    Figure 5.9 Effect of total pressure ratio on the specific power consumption: motor and (isentropic + friction)  5.5.2 Volumetric Efficiency  The volumetric efficiency is often modelled by including a loss term to the ideal volumetric efficiency based on an ideal compression process [56,57,59]. One such commonly acceptable approach to empirically evaluate volumetric efficiency (𝜂𝑣,𝑒)  is discussed below [59,85]:   𝜂𝑣,𝑒 = 1 −𝑉𝑐𝑉𝑠((𝑃2𝑃1)1𝑛− 1) − 𝐿 (5.6) where 𝑉𝑐𝑉𝑠, (𝑃2𝑃1), n, and L are the clearance fraction, stage pressure ratio, polytropic exponent, and loss fraction respectively. Various estimates of L are recommended in literature. [59], [86], and [57] suggests L to be taken as 0.05, (𝑃2𝑃1) (1100), and (5 + 2(𝑃2𝑃1) ) (1100), respectively. For the calculation of 𝜂𝑣,𝑒 in this study, L was taken as (5 + 2(𝑃2𝑃1) ) (1100). n couldn’t be determined due to the challenges discussed in Section 5.3, and therefore n = 1.3 was chosen, such that variation of 𝜂𝑣,𝑒 with (𝑃2𝑃1) had a similar slope to that of the measured volumetric efficiency (𝜂𝑣,𝑚; defined in Section 3.2.2) with 𝑃2𝑃1.  97  The effect of stage pressure ratio (𝑃2𝑃1) on the 𝜂𝑣,𝑚 of both configurations are shown in Figure 5.10. Increasing (𝑃2𝑃1) decreases the volumetric efficiency as is expected from Equation (5.6). No consistent difference on the measured volumetric efficiency of 2-stage relative to 3-stage configuration at a 𝑃2𝑃1 was observed, and therefore indicates similar breathing in both configurations. This also reinforces the previous assertion made in Section 5.2 that the significant reductions of mass flowrate in 2-stage relative to 3-stage is due to reduced inlet fluid density.  Comparisons of 𝜂𝑣,𝑚, 𝜂𝑣,𝑒, and volumetric efficiencies of reciprocating compressors seen in various applications in the literature are shown in Figure 5.10. A range of volumetric efficiencies is observed from literature, with that from a compressor of a comparable displacement closer to the tested 𝜂𝑣,𝑚, while others significantly higher. This indicates that there is unlikely any issues in terms of flow capacity of the tested compressor. However, the fact that the observed 𝜂𝑣,𝑚 is towards the lower end of spectrum with respect to that seen in literature as well as the 𝜂𝑣,𝑒, suggests potential for improvement through redesign (e.g. clearance volume reduction).   98   Figure 5.10 Comparison of volumetric efficiencies: Measured, modelled, and applications in literature. Note: Vs is swept volume of the first stage of the compressor; Superscripts a, b, c refer to References [62], [65], and [68], respectively  Volumetric efficiency improvements can be made by decreasing the clearance fraction (𝑉𝑐𝑉𝑠),  although it may result in decreased isentropic efficiency [59], due to smaller clearances requiring smaller valves leading to more valve losses. However, for high pressure ratio (>10) applications, such as the current 2-stage configuration, [59,86] emphasizes the reduction of clearance fraction over increasing valve sizing  as a general strategy of compressor design.  5.5.3 Mechanical Efficiency and Overall Efficiency The overall efficiency (𝜂𝑂𝐸) definition used in this study is based on [57] (Equation (5.7)). This definition considers the compression system to be only the compressor, exclusive of motor (driver) and belt; however, their respective efficiencies (𝜂𝑚𝑜𝑡𝑜𝑟= 0.88; 𝜂𝑏𝑒𝑙𝑡= 0.9) were included in the evaluation of 𝑃𝑠ℎ𝑎𝑓𝑡 required for the calculation of the efficiencies discussed in this sub-section. The expressions of 𝜂𝑂𝐸 and mechanical efficiency (𝜂𝑚𝑒𝑐ℎ) are discussed below:  𝜂𝑂𝐸 = (Compressor Efficiency) × (Mechanical Efficiency) (5.7) Refrigerationa Refrigerationb Twin-cylinder  Vs~ 2.3Vs of  UBC compressorc 99  The isentropic efficiency (𝜂𝑖𝑠𝑒𝑛) was taken as the compression efficiency, and Equation (5.7) was reduced to eliminate its dependence on the indicated work (?̇?), which couldn’t be measured here:  𝜂𝑂𝐸  =  ?̇?𝑖𝑠𝑒𝑛?̇? ?̇?𝑃𝑠ℎ𝑎𝑓𝑡=  ?̇?𝑖𝑠𝑒𝑛𝑃𝑠ℎ𝑎𝑓𝑡=  ?̇?𝑖𝑠𝑒𝑛𝜂𝑚𝑜𝑡𝑜𝑟𝜂𝑏𝑒𝑙𝑡𝑃𝑚𝑜𝑡𝑜𝑟  (5.8)  𝜂𝑚𝑒𝑐ℎ  =   ?̇?𝑃𝑠ℎ𝑎𝑓𝑡=  𝑃𝑠ℎ𝑎𝑓𝑡 − 𝑃𝑓𝑃𝑠ℎ𝑎𝑓𝑡 =  1 −𝑃𝑓𝜂𝑚𝑜𝑡𝑜𝑟𝜂𝑏𝑒𝑙𝑡𝑃𝑚𝑜𝑡𝑜𝑟 (5.9) where 𝑃𝑓= 1.5 kW.  Comparison of the overall and mechanical efficiencies of 2-stage and 3-stage, and various applications in literature are shown in Figure 5.11 and Figure 5.12, respectively. 𝜂𝑂𝐸 of both 2-stage and 3-stage configurations were found to be lower than those presented in  the literature (petrochemical [57] and refrigeration [66]). Figure 5.12 shows that the behavior of 𝜂𝑚𝑒𝑐ℎ of both configurations is very similar to that of the respective 𝜂𝑂𝐸, indicating that 𝜂𝑂𝐸 was governed by 𝜂𝑚𝑒𝑐ℎ. Also, 𝜂𝑚𝑒𝑐ℎ of both 2-stage and 3-stage was much lower than that observed in literature ([57,66] in Table 2.2), indicating that the predominant performance issue with the compressor used in this study is low 𝜂𝑚𝑒𝑐ℎ, which is most likely due to high friction value of the compressor used. The fact that the 𝜂𝑚𝑒𝑐ℎ of the tested compressor is bounded by petrochemical (low speed ~ 400 rpm; large size) and refrigeration (high speed~ 3600 rpm; small size) compressors, both in terms of size and shaft speed, suggests the possibility of potential improvement in 𝜂𝑚𝑒𝑐ℎ (and 𝜂𝑂𝐸) through optimized redesign of diluent compressor. Although the analysis in this sub-section is dependent on the assumed value of 𝜂𝑏𝑒𝑙𝑡 (0.9), the accuracy of nameplate rated 𝜂𝑚𝑜𝑡𝑜𝑟 (0.88), and the estimated value of 𝑃𝑓 (1.5 kW, Section 5.4), a sensitivity analysis validated the robustness of the conclusions. A variation of either 𝜂𝑚𝑜𝑡𝑜𝑟 or 𝜂𝑏𝑒𝑙𝑡 by ±0.05, resulted in a maximum change of 2%- point and 4-% point of the of the 𝜂𝑂𝐸 of 2-100  stage and 3-stage configuration, respectively. Sensitivity of 𝜂𝑚𝑒𝑐ℎ to similar variations of 𝜂𝑚𝑜𝑡𝑜𝑟 or 𝜂𝑏𝑒𝑙𝑡 resulted in a maximum change of 3.5%- point and 3-% point in the 𝜂𝑚𝑒𝑐ℎ of 2-stage and 3-stage configuration, respectively. Furthermore, sensitivity of 𝜂𝑚𝑒𝑐ℎ to ±10% change in 𝑃𝑓, resulted in a maximum change of 5%- point and 4-% point of 𝜂𝑚𝑒𝑐ℎ of 2-stage and 3-stage configuration, respectively.   Figure 5.11 Comparison of overall efficiencies: 2-stage, 3-stage, and applications in literature.  Note: Stage pressure ratio is calculated by assuming all stages have equal pressure ratios at a total pressure ratio. Superscript a refers to [66] and Campbell, J. refers to [57]  Figure 5.12 Comparison of mechanical efficiencies: 2-stage, 3-stage, and applications in literature.  Note: Stage pressure ratio is calculated by assuming all stages have equal pressure ratios for a total pressure ratio. Superscript a refers to [66] and Campbell, J. refers to [57]  101  5.6 Conclusions  In this chapter, the compression power and the output flowrate of a 2-stage and a 3-stage compressor configuration were characterized to evaluate the performance relative to each other and with that seen in literature. The conclusions are presented below:  • The effects of inlet and outlet pressure on the power and flowrate could be qualitatively understood in the context of polytropic compression process. • Significantly lower mass flowrates were observed in 2-stage compressor configuration relative to 3-stage because of its operation with much lower inlet pressures compared to design values. The reduced mass flowrates also led to reduced power consumption in 2-stage as expected from a polytropic compression power perspective. • The 2-stage configuration, at air addition relevant outlet pressures (> 280 bar), would be able to provide only up to 80% FDR with the fuel flowrates obtained at the tested engine mode (Chapter 4), while the 3-stage configuration could achieve up to 150% FDR. • The friction of the tested compressor, a pre-requisite for mechanical efficiency calculation, was estimated to be 1.5 kW. Two separate methods, independent of indicated work and shaft work measurements, were developed for this. Both methods resulted in comparable values of friction. •  The motor specific power consumption was nearly an order of magnitude higher in 2-stage configuration relative to 3-stage at the highest-pressure ratios. A mechanistic consideration of the constituents of motor specific power revealed that this effect originated from the loss specific power component (friction being the major constituent), which increased exponentially at the low flowrates obtained in 2-stage configuration. This analysis also 102  indicates that improving the flow capacity may allow possible improvements in 2-stage performance.  • Volumetric efficiencies (𝜂𝑣,𝑚) of both configurations at the same stage pressure ratio remained nearly same and were comparable to that seen in literature. This indicates similar breathing in both configurations as well as absence of major flow capacity issues in the tested compressor.  • Overall efficiency (𝜂𝑂𝐸) of both the 2-stage and 3-stage configurations was significantly lower than that seen in literature. This is due to the much lower mechanical efficiencies (3-stage, 𝜂𝑚𝑒𝑐ℎ ∈ [50, 60]%; 2-stage, 𝜂𝑚𝑒𝑐ℎ ∈ [30, 50]%) of the tested compressor, which is a result of high friction of the compressor. Although this analysis is dependent on the assumed value of 𝜂𝑏𝑒𝑙𝑡 (0.9), the accuracy of nameplate rated 𝜂𝑚𝑜𝑡𝑜𝑟(0.88), and estimated value of 𝑃𝑓, a sensitivity analysis validated the robustness of these conclusions. In terms of both size and shaft speed, the tested compressor is bounded by compressors seen in literature, and therefore its 𝜂𝑚𝑒𝑐ℎ indicates potential for improvement.  103   System Level Study of Air Addition This chapter uses experimental results from Chapter 4 and Chapter 5 to conduct a system level assessment of air addition. Section 6.1 discusses the relevant response surfaces based on experimental results that were used for the various analysis presented in this chapter. A NOx control strategy is developed to counter the observed NOx trade-off with air addition in Section 6.2. It is then used to provide a qualitative assessment of PM emissions with NOx. A system efficiency parameter, which considered the engine efficiency improvements from air addition together with the compression power required for various compression strategies, is proposed to discuss the system level implications of air addition in Section 6.3. An optimization methodology, which targeted both reduction in emissions and improvement in system efficiency, was described and used to recommend operating spaces in Section 6.4. 6.1 Response Surfaces of Experiments Response surfaces, which are mathematical functions relating several independent variables to one response variable, were used in previous PIDING engine studies for sensitivity analysis [37], optimization [37], calculation of tolerance limits of input parameters [27] etc. In this study, they were used to develop a NOx control strategy, study system level implications of air addition, and estimate emissions and efficiency for optimization.    The response surfaces used in this study are multi-variable polynomial fits (up to 2nd order) to independent variables varied in the experiments: EGR, Φ, and FDR for engine experiments, and inlet pressure (𝑝𝑖𝑛) and outlet pressure (𝑝𝑜𝑢𝑡) for the compressor experiments. The response surfaces were calculated using MATLAB’s interactive response surface toolbox [87], which 104  follows the general response surface methodology described in [88]. A brief discussion is given below: • The variables were converted to normalized variables by scaling with their ranges, such that they had a mean of zero and a maximum and minimum values of 1, -1, respectively. The scaled independent and response variables are represented by matrix X and y, respectively.  • The output of the model is described by ?̂? = 𝑋𝐵, where 𝐵 is the coefficients matrix calculated by 𝐵 = (𝑋′𝑋)−1𝑋′𝑦 and ?̂? is the output values of the fit. • Normalized variables were unscaled to arrange the response surface equation in terms of actual independent (𝑥𝑖) and dependent variables (𝑦) (Table 6.1) Various response surface fits of polynomials, from linear to 2nd order, were tried for each response. Lower order response surfaces were preferred due to their simplicity; unless higher order response surfaces provided significantly better closeness of fit or correlation. The final response surfaces are tabulated in Table 6.1, while their corresponding residuals, standard error, and R2 values are listed in Appendix C.1. Amongst the tabulated response surfaces, all except PM, can be used for quantitative analysis because of their reasonable fit accuracies, with less than 5% deviations (Appendix C.1). The response surface of PM, even with R2= 0.97, had up to 50% deviations with respect to measured values (Appendix C.1), and therefore was used only for qualitative analysis. Similar fit inaccuracies were also seen with the response surfaces of THC and CO, and they were therefore not used for any analysis in this study. Although possibility of resolving these inaccuracy issues exists through the usage of advanced emission models (e.g. neural networks, Gaussian process regression), they were not attempted in this study.   105  Table 6.1 Response surface equations and constants.  Note: 𝑃𝑚𝑜𝑡𝑜𝑟  and ?̇?𝑜𝑢𝑡 are motor specific power and outlet mass flowrate, respectively of 2-stage compressor configuration. EGR, Φ, and FDR are expressed in percent, fraction, percent, respectively. Units of PM, NOx, 𝜂𝑖,𝑔, 𝑃𝑚𝑜𝑡𝑜𝑟 , ?̇?𝑜𝑢𝑡 , 𝑝𝑜𝑢𝑡, 𝑝𝑖𝑛  are g/kWhr, g/kWhr, unit less in%, kW, kg/hr, bar, and bar, respectively. The tabulated values provided as 𝑏𝑜 − 𝑏𝑛 were rounded to two significant figures, although calculations were made with unrounded values.  Response surface equation  Constant values [𝑏1, 𝑏2, … 𝑏𝑛] R2 𝑙𝑛(𝑃𝑀) = 𝑏1 + 𝑏2𝐸𝐺𝑅 + 𝑏3Φ + 𝑏4𝐹𝐷𝑅  [-7.3, 0.078, 5.7, -0.034] 0.97 NOx = 𝑏1 + 𝑏2𝐸𝐺𝑅 + 𝑏3Φ + 𝑏4𝐹𝐷𝑅 +𝑏5𝐸𝐺𝑅. Φ + 𝑏6𝐸𝐺𝑅. 𝐹𝐷𝑅 +  𝑏7Φ. 𝐹𝐷𝑅 +𝑏8𝐸𝐺𝑅2 +  𝑏9Φ2 + 𝑏10𝐹𝐷𝑅2 [8.8, -0.36, -5.2, 0.043, 0.17, -4.2, -0.031, 0.0036, -1.1, -3.6] 0.99 𝜂𝑖,𝑔 = 𝑏1 + 𝑏2𝐸𝐺𝑅 + 𝑏3Φ + 𝑏4𝐹𝐷𝑅  [0.55, 9.4, -0.18, 2.2] 0.95 𝑝𝑖𝑛,𝑒𝑛𝑔 = 𝑏1 + 𝑏2𝐸𝐺𝑅 + 𝑏3Φ + 𝑏4𝐹𝐷𝑅  [3.9, 0.019, -2.2, -0.0016] 0.98 𝑃𝑚𝑜𝑡𝑜𝑟 = 𝑏1 + 𝑏2𝑝𝑖𝑛 + 𝑏3𝑝𝑜𝑢𝑡𝑝𝑖𝑛+𝑏4. (𝑝𝑖𝑛.𝑝𝑜𝑢𝑡𝑝𝑖𝑛)  [2.3, 0.27, -0.0038, 0.0023] 0.97 ?̇?𝑜𝑢𝑡 = 𝑏1 + 𝑏2𝑝𝑖𝑛 + 𝑏3. (𝑝𝑖𝑛.𝑝𝑜𝑢𝑡𝑝𝑖𝑛)  [-0.42, 2.5, -0.0083] 0.99  6.2 NOx Control Strategy A NOx control strategy is developed in this section to account for the NOx penalty associated with FDR. This strategy uses EGR to control NOx and proposes the required amounts of EGR specific to various baseline EGR-Φ combinations in the considered engine mode, to compensate for NOx increases with 0 to 100% FDR. The NOx response surface (Table 6.1) was used to calculate the EGR at a given FDR and Φ, such that the NOx, relative to that at 0% FDR, remained constant (ΔNOx~ 0). The calculation methodology is discussed below: • For a constant Φ, the NOx response surface of the experiments in this study is a function of two variables and can be differentiated as follows: o 𝑑(𝑁𝑂𝑥) =𝛿𝑁𝑂𝑥𝛿𝐸𝐺𝑅∆𝐸𝐺𝑅 +𝛿𝑁𝑂𝑥𝛿𝐹𝐷𝑅∆𝐹𝐷𝑅 106  • Path of constant NOx was calculated by setting 𝑑(𝑁𝑂𝑥)= 0 and by iteratively incrementing the FDR by 5%-points (𝐹𝐷𝑅𝑖+1) while calculating the corresponding incremented EGR (𝐸𝐺𝑅𝑖+1) as follows: o 𝛿𝑁𝑂𝑥𝛿𝐸𝐺𝑅|𝑖(𝐸𝐺𝑅𝑖+1 − 𝐸𝐺𝑅𝑖) +𝛿𝑁𝑂𝑥𝛿𝐹𝐷𝑅|𝑖(𝐹𝐷𝑅𝑖+1 − 𝐹𝐷𝑅𝑖)= 0 o 𝐸𝐺𝑅𝑖+1 =𝛿𝑁𝑂𝑥𝛿𝐹𝐷𝑅|𝑖𝛿𝑁𝑂𝑥𝛿𝐸𝐺𝑅|𝑖(𝐹𝐷𝑅𝑖 − 𝐹𝐷𝑅𝑖+1) + 𝐸𝐺𝑅𝑖 • The path of constant NOx is the obtained FDR-EGR combination at a constant Φ The proposed amounts of EGR specific to various EGR-Φ combinations, to account for the increased NOx with FDR is tabulated in Table 6.2. The increase in EGR required to keep NOx constant with 100% air addition is represented by ∆𝐸𝐺𝑅|∆𝑁𝑂𝑥=0. In all cases, less than 10%-point EGR increase could compensate for the NOx penalty with up to 100% FDR. The lowest EGR increase was required at the highest Φ, and vice versa. Although relatively higher ∆𝐸𝐺𝑅|∆𝑁𝑂𝑥=0 was required for lower Φ at higher baseline EGR, the ∆𝐸𝐺𝑅|∆𝑁𝑂𝑥=0 at the highest Φ was nearly independent of baseline EGR and nearly equal to 6.5%. This suggests the possibility of the NOx control strategy to be equally effective at higher baseline EGRs if the operation is at high Φ. This strategy also showed no trade-offs in gross indicated efficiency.  The qualitative effect of the NOx control strategy on the PM emissions is tabulated in Table 6.2, while the same effect at EGR= 12.5% is shown in Figure 6.1. This effect was calculated by determining the PM emissions based on its response surface (Table 6.2) corresponding to the FDR-EGR combinations from the path of constant NOx determined above. Although the PM reduction factor at 100% FDR (𝑃𝑀𝐹𝐷𝑅=0𝑃𝑀𝐹𝐷𝑅=100) based on response surface (Table 6.2) remained a constant 28.5 at a given EGR and Φ, it should be noted that this is due to the exponential nature of the fitted PM 107  response and is not representative of that seen with measured results. As expected, the PM reduction factor at 100% FDR with constant NOx (𝑃𝑀𝐹𝐷𝑅=0𝑃𝑀𝐹𝐷𝑅=100|∆𝑁𝑂𝑥=0), decreased with respect to corresponding 𝑃𝑀𝐹𝐷𝑅=0𝑃𝑀𝐹𝐷𝑅=100. This decrease corresponded to ∆𝐸𝐺𝑅|∆𝑁𝑂𝑥=0 (i.e. higher ∆𝐸𝐺𝑅|∆𝑁𝑂𝑥=0 led to lower 𝑃𝑀𝐹𝐷𝑅=0𝑃𝑀𝐹𝐷𝑅=100|∆𝑁𝑂𝑥=0) and is again due to the exponential nature of the chosen response surface for PM. Figure 6.1 demonstrates that a relatively higher PM curve is expected with constant NOx (∆𝑁𝑂𝑥~ 0) at a particular Φ. These comparisons were made with response surface values not actual measured values to highlight the effect of controlled NOx, which may seem insignificant relative to deviations in actual measured PM with respect to PM from response surface fit.  It is important to note that the assessments based on PM response surface (i.e. 𝑃𝑀𝐹𝐷𝑅=0𝑃𝑀𝐹𝐷𝑅=100 and 𝑃𝑀𝐹𝐷𝑅=0𝑃𝑀𝐹𝐷𝑅=100|∆𝑁𝑂𝑥=0), unlike those based on NOx response surface (i.e. ∆𝐸𝐺𝑅|∆𝑁𝑂𝑥=0), are qualitative in nature due to the inaccuracies in the PM response surface (Section 6.1). Also, response surfaces were based on experiments with constant operating parameters of GIMEP, engine speed, IHR50, PSEP, GRP, and diesel flow, and therefore any derived analysis would inherently be on the same engine operating space.         108  Table 6.2 The proposed EGR for NOx control at various baseline EGRs and Φs, and its qualitative effect on PM emissions  Note: 𝑃𝑀𝐹𝐷𝑅=0𝑃𝑀𝐹𝐷𝑅=100|∆𝑁𝑂𝑥=0 and 𝑃𝑀𝐹𝐷𝑅=0𝑃𝑀𝐹𝐷𝑅=100 are response surface based PM reduction factor at 100% FDR, with and without NOx control, respectively  EGR% (baseline) Φ  ∆𝐸𝐺𝑅|∆𝑁𝑂𝑥=0 𝑃𝑀𝐹𝐷𝑅=0𝑃𝑀𝐹𝐷𝑅=100 𝑃𝑀𝐹𝐷𝑅=0𝑃𝑀𝐹𝐷𝑅=100|∆𝑁𝑂𝑥=0 10 0.58 8.5 28.5 14.8 0.68 7.4 28.5 16.0 0.77 6.3 28.5 17.4 15 0.58 9.1 28.5 14.0 0.68 7.8 28.5 15.5 0.77 6.5 28.5 17.2 20 0.58 10.3 28.5 12.8 0.68 8.6 28.5 14.6 0.77 6.7 28.5 16.9    Figure 6.1 Qualitative effect of the NOx control strategy on the PM emissions at baseline EGR= 12.5% Note: PM emissions plotted, with and without NOx control, are based on response surfaces.  6.3 System Efficiency The system level assessment (energy cost basis) of air addition system (diluent compressors, buffer tanks), is primarily dependent on the diluent compressor and the engine operating mode. This was done by defining system efficiency (𝜂𝑠𝑦𝑠), which considered the compression power (𝑃𝑐) required for diluent compression to the engine gross indicated efficiency (𝜂𝑖,𝑔) obtained with air addition. It is described with Equation (6.1) below: Increasing EGR with  respect to baseline 109   𝜂𝑠𝑦𝑠 =𝑃𝑖,𝑔 − 𝑃𝑐?̇?𝑁𝐺 . 𝐿𝐻𝑉𝑁𝐺 +?̇?𝑑. 𝐿𝐻𝑉𝑑 (6.1) Only one engine operating mode was considered, while 4 different compression systems were considered for the calculation of system efficiency. This includes the tested 3-stage and the tested 2-stage (Chapter 5), and a hypothetical 2-stage and hypothetical 3-stage compressor based on expected performance improvements of their respective tested configurations. Although calculations were based on known and measured quantities (e.g. specific motor power, 𝜂𝑖,𝑔), it is important to note that derived parameters like the shaft compressor power and engine brake efficiency may be more relevant in an on-board application context.  The compression power (𝑃𝑐) was evaluated at an FDR using the corresponding diluent flowrate (?̇?𝑑𝑖𝑙,𝑎𝑖𝑟) and the specific motor power consumption (𝑃𝑚,𝑠) at air addition relevant operating conditions:  𝑃𝑐 = ?̇?𝑑𝑖𝑙,𝑎𝑖𝑟. 𝑃𝑚,𝑠 𝑃𝑚,𝑠 was determined at a compressor outlet pressure of 300 bar. Although, 𝑃𝑚,𝑠 for 3-stage tested was a function of only total pressure ratio (Figure 5.8; Figure 6.2), 𝑃𝑚,𝑠 of 2-stage configuration was also a function of inlet pressure (Figure 5.8, Figure 6.2), and therefore 𝑃𝑚,𝑠 of 2-stage configuration was evaluated at air addition relevant inlet pressures. Also, these evaluations were based on response surfaces of  𝑃𝑚𝑜𝑡𝑜𝑟 and ?̇?𝑜𝑢𝑡 described in Table 6.1, due to the lack of measured 𝑃𝑚,𝑠 of 2-stage at each relevant inlet pressures. 𝑃𝑚,𝑠 of the hypothetical 2-stage and 3-stage compressors were evaluated by assuming a typical overall efficiency (𝜂𝑂𝐸  ) expected based on applications in literature:  • Rearranging Equation (5.8): 𝜂𝑂𝐸  =  ?̇?𝑖𝑠𝑒𝑛𝑃𝑠ℎ𝑎𝑓𝑡 = ?̇?(𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛=1)𝑃𝑠ℎ𝑎𝑓𝑡 = ?̇?(𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛=1)𝜂𝑚𝑜𝑡𝑜𝑟𝜂𝑏𝑒𝑙𝑡𝑃𝑚𝑜𝑡𝑜𝑟 = (𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛=1)𝜂𝑚𝑜𝑡𝑜𝑟𝜂𝑏𝑒𝑙𝑡𝑃𝑚,𝑠  110  • Rearranging above equation: 𝑃𝑚,𝑠 =(𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛=1)𝜂𝑚𝑜𝑡𝑜𝑟𝜂𝑏𝑒𝑙𝑡𝜂𝑜𝑒 , assuming 𝜂𝑚𝑜𝑡𝑜𝑟= 0.88  and 𝜂𝑏𝑒𝑙𝑡= 0.9   • 𝜂𝑂𝐸= 0.7, an intermediate overall efficiency value observed in diverse reciprocating compressor applications [57,66], was chosen for further analysis.  • 𝜂𝑂𝐸= 0.7, subsequently referred to as 𝜂𝑂𝐸−𝑑𝑒𝑠, was used to evaluate the expected specific motor power (𝑃𝑚,𝑠−𝑑𝑒𝑠) of both hypothetical 2-stage and 3-stage configurations based on the obtained isentropic specific power (𝑃𝑖,𝑠|𝜂𝑖𝑠𝑒𝑛 = 1) from tested 2-stage and 3-stage configurations, respectively Comparison of motor specific power of tested and hypothetical compressor configurations are shown in Figure 6.2. With the chosen value of 𝜂𝑂𝐸−𝑑𝑒𝑠= 0.7, 𝑃𝑚,𝑠−𝑑𝑒𝑠 of both 2-stage and 3-stage hypothetical configurations were lower than the measured motor specific power (𝑃𝑚,𝑠) of 3-stage. This suggests the possibility of using a 2-stage configuration for air addition concept through optimization and redesign. It is important to note that 𝜂𝑂𝐸 of the hypothetical 2-stage configuration, like the 2-stage configuration may be provided with compressed air at the inlet, and as such would also be dependent on inlet pressures (Section 5.5.3). However, it is assumed constant for the analysis presented in this study.   111    Figure 6.2 Comparison of motor specific power of tested with hypothetical compressor configurations Note: 2 Stage Pm,s-des and 3 Stage Pm,s-des are expected motor specific power of hypothetical 2-stage and 3 stage compressor configurations, respectively.    𝜂𝑠𝑦𝑠 based on Equation (6.1), was calculated for all four considered compressor configurations. The effect of FDR on the system efficiency at various Φ and EGR of all considered configurations are shown in Figure 6.3 and Figure 6.4. In all, except 2-stage tested configuration, a nearly constant or increasing 𝜂𝑠𝑦𝑠 with FDR was seen for all compressor configurations, indicating that the fuel economy benefits with air addition would likely be able to overcome their respective diluent compression power. For these configurations, relatively higher increase in 𝜂𝑠𝑦𝑠 was seen at higher Φs because of higher increase in 𝜂𝑖,𝑔 with FDR observed at higher Φs (Section 4.5). However, 2-stage showed a contradictory trend because of much higher 𝑃𝑚,𝑠 (thereby higher 𝑃𝑐) resulting from reduced compressor inlet pressures, which is caused by relatively lower intake manifold pressure obtained at higher Φs (Section 5.5.1). An ANOVA analysis, tabulated in C.1, revealed that the variations of 𝜂𝑠𝑦𝑠 with EGR as well as its interaction terms with Φ and FDR to be statistically insignificant (p-value> 0.8) in all except the 2-stage tested configuration. This different behavior of 2-stage tested is due to similar reasons, i.e. only the 𝑃𝑚,𝑠 of 2-stage tested was dependent on intake manifold pressures, which in turn is affected by EGR.  112    Figure 6.3 Effect of FDR on system efficiency at various Φ and EGR of (a) 3-stage tested and (b) 3-stage hypothetical compressor configurations     Figure 6.4 Effect of FDR on system efficiency at various EGR and Φ of (a) 2-stage tested and (b) 2-stage hypothetical compressor configurations  6.4 Optimum Operating Spaces A merit function (𝑓𝑚) of the form typically used for engine operation optimization studies [89–91], was used to identify operating spaces for optimum engine operation with air addition. The merit function (𝑓𝑚) assigns high value to the operating spaces with low values of key emissions (PM and NOx) and system specific fuel consumption (SSFC). The 𝜂𝑠𝑦𝑠 (defined in Section 6.3) is inversely proportional to SSFC and is considered in the merit function to include both the effect of air addition on the fuel economy as well as the cost of diluent compression. The (a) (b) (a) (b) 113  operating space (i.e. EGR*, Φ∗, and FDR*) at the highest value of the merit function (𝑓𝑚∗) is taken as the optimum solution. 𝑓𝑚 is defined below:  𝑓𝑚 =1000(𝑃𝑀𝑃𝑀𝑡)2 + (𝑁𝑂𝑥𝑁𝑂𝑥𝑡)2 + (𝜂𝑠𝑦𝑠,𝑡𝜂𝑠𝑦𝑠)  (6.2) where 𝑃𝑀𝑡, 𝑁𝑂𝑥𝑡, and 𝜂𝑠𝑦𝑠,𝑡 are target values of PM emissions, NOx emissions, and system efficiency, respectively. The targets of PM (𝑃𝑀𝑡= 0.01 g/kWhr) and NOx (𝑁𝑂𝑥𝑡= 0.4 g/kWhr) emissions were based on Euro 6 regulations [92] while the 𝜂𝑠𝑦𝑠,𝑡 was kept equal to the mean of measured 𝜂𝑠𝑦𝑠 at FDR= 0% (i.e. 43%). 𝑃𝑀, 𝑁𝑂𝑥, and 𝜂𝑠𝑦𝑠 were calculated from their respective response surfaces. Euro 6 also has regulation targets for CO and THC; however, they were not included in the 𝑓𝑚 to simply the analysis and due to lack of response surfaces with sufficient accuracy as mentioned in Section 6.1. Only 𝜂𝑠𝑦𝑠 of the 3-stage tested configuration was considered for results and discussion in this section since the optimum solution for all considered domains was found to be independent of the compressor configurations. The higher power in the input emissions causes the optimization routine to prioritize search for operation spaces with reduced emissions over reduced fuel consumption, until operation space with near target values of emissions are achieved. However, the chosen form of the merit function (and the associated optimum operating space) is dependent on the objectives of the manufacturers, and a more realistic prioritization may emphasize fuel economy over emissions. It is important to note that the errors in the parameters included in 𝑓𝑚 via propagation would result in an uncertainty in the merit function value at the optimum solution (𝑈𝑓𝑚∗), and therefore the optimum solution (EGR*, Φ∗, and FDR* at 𝑓𝑚∗) for a particular domain, is presented together with the optimum solution range, which is the operating space corresponding to 𝑓𝑚∗ ± 𝑈𝑓𝑚∗(see A.2 for 𝑈𝑓𝑚∗ calculation).  114  The optimum solution and relevant data obtained using the above methodology is tabulated in Table 6.3. The optimum solution within the full considered domain (Domain A) was found at the highest values of the three control variables i.e. at EGR= 25%, Φ= 0.77, and FDR= 100% (Figure 6.5a). Also, a calculation of the gradient at this optimum solution, which also represents the direction of the steepest ascent, revealed that a better optimum for the considered engine operating mode (global optimum solution) would likely be obtained via the increase of all three control variables. A visualization of the gradients at the optimum solution can be obtained from Figure 6.5a, while the value of the gradient vector is tabulated in Table 6.3. With reduction in Φ in the constrained domains (i.e. Φ fixed), Domain B and C, resulted in optimum solution with lower values of FDR (<100% FDR). This is because at reduced Φ, PM is relatively low, while NOx is relatively high, and therefore a lower FDR is sufficient to achieve a near target PM value with the added benefit of lower NOx penalty. For all domains, the optimum solutions were consistently observed at the highest EGR (25%). The EGR* (EGR value at optimum solution) also showed the range of the optimum solution to be only 1%-point, further supporting the robustness of this result. None of the three-optimum solution were found to have a clear advantage in all the parameters (e.g. Domain C had the highest 𝜂𝑠𝑦𝑠, Domain A had the lowest NOx, while Domain B had the lowest CO, THC, and PM). Therefore, a more meticulous determination of the best optimum solution is suggested by including weights for all the parameters used in the merit function, as this would allow to account for their relative commercial values. Also, a determination of optimum solution should be followed by its testing, which was not done for that of Domain B and C, due to timeline and facility constraints of the project. The robustness of the above conclusions was tested via a sensitivity analysis of the target parameters of the merit function and the results are tabulated in Table 6.4. It was observed that 115  variation of any of the considered target parameters as shown in Table 6.4, didn’t result in any significant changes in the optimum solution of Domain A. A factor of 2 changes to the emissions targets resulted in changes in optimum solution of Domains B and C, particularly FDR*; however, no such changes were seen when 𝜂𝑠𝑦𝑠 was changed by ±2%-point. The optimum solution was also found to be independent of all the considered compressor configurations, excluding 2-stage tested.    Figure 6.5 Variation of merit function with FDR and EGR at (a) Φ= 0.77 and (b) Φ= 0.68          (a) (b) 116  Table 6.3 Optimization solution and relevant data Note: 𝑓𝑚∗, Gradient, and 𝑈𝑓𝑚∗ were calculated at the optimum solution. In the tabulated emission’s values, ‘RS’ represents response surface calculated values at optimum solution while ‘meas.’ represents measured values at the nearest point to optimum solution (i.e. for Domain B and C, the measured values were at FDR= 100% and 50%, respectively). Euro 6 based target values for PM, NOx, CO, THC are 0.01, 0.4, 1.5, and 0.13 g/kWhr, respectively. Domains considered for  optimization Domain A:  EGR= [0 25] % FDR= [0 100] % Φ= [0.58 0.77] Domain B: EGR= [0 25] % FDR= [0 100] % Φ= [0.68] Domain C: EGR= [0 25] % FDR= [0 100] % Φ= [0.58] Optimum solution range EGR% [24 25] [24 25] [24 25] Φ [0.74 0.77] 0.68 0.58 FDR% [90 100] [70 100] [45 80] Optimum solution  [EGR* Φ∗ FDR*] [25 0.77 100] [25 0.68 89] [25 0.58 59] 𝑓𝑚∗  100.8 60 42.58 Gradient  [𝜕𝑓𝑚𝜕𝐸𝐺𝑅 𝜕𝑓𝑚𝜕Φ 𝜕𝑓𝑚𝜕𝐹𝐷𝑅] [9.95 554 1.1] [6.41 293 0.003] [3.47 91.9 0] 𝑈𝑓𝑚∗   14.2 6.22 3.55 PM: RS| meas. 0.013| 0.025 0.012| 0.0073 0.0182| 0.034 NOx: RS| meas. 1.05| 1 1.52| 1.5 1.75| 1.7 CO: RS| meas. n/a| 8.3 n/a| 3.18 n/a| 4.8 THC: RS| meas. n/a| 0.3 n/a| 0.26 n/a| 0.51 𝜂𝑠𝑦𝑠: RS| meas.  40.81| 40.76 42.28| 42.67 44.24| 44.36  Table 6.4 Sensitivity of optimum solution to the target values in merit function Parameter varied  Domain A   [EGR* Φ∗ FDR*] Domain B [EGR* Φ∗ FDR*]  Domain C [EGR* Φ∗ FDR*]  𝑃𝑀𝑡/2  [25 0.77 100] [25 0.68 100] [25 0.58 80.5] 𝑃𝑀𝑡 × 2  [25 0.77 100] [25 0.68 65.5] [25 0.58 38] 𝑁𝑂𝑥𝑡/2  [25 0.77 100] [25 0.68 65.5] [25 0.58 38.5] 𝑁𝑂𝑥𝑡 × 2   [25 0.76 100] [25 0.68 100] [25 0.58 80.5] 𝜂𝑠𝑦𝑠,𝑡 = 45%  [25 0.77 100] [25 0.68 89] [25 0.58 59] 𝜂𝑠𝑦𝑠,𝑡 = 41%  [25 0.77 100] [25 0.68 89] [25 0.58 59]   117  6.5 Conclusions This chapter considered the engine and the diluent compression system together to provide a system level assessment of air addition. A NOx control strategy to counter the increased NOx with air addition and an optimization methodology to determine optimum operation space were proposed. The conclusions are presented below: • All the considered response surfaces developed in this chapter, except that for PM, can be used for quantitative analysis. This is because, unlike other considered response surfaces, the response surface for PM couldn’t be determined with sufficient fit accuracies.  • A NOx control strategy, which proposes specific amounts of EGR required at various baseline EGR-Φ combinations, was developed to counter the increased NOx with FDR. It was observed that in all cases, less than 10%-point EGR increase could compensate for the NOx penalty at 100% FDR. The amount of EGR required at the highest Φ (0.77) was independent of baseline EGR and was about 6.5%.  • The 𝜂𝑠𝑦𝑠 analysis revealed that increase in 𝜂𝑖,𝑔 with air addition, irrespective of EGR rates, was sufficient to compensate for the compression load consumed by a three-stage reciprocating compressor when operating at high Φ at the tested heavy-duty engine mode. This also indicates the possibility of the air addition concept to be more feasible (i.e. free of a net parasitic load) at high load engine operations, where Φs are constrained to higher values. • System efficiency (𝜂𝑠𝑦𝑠), a proposed parameter for the system level assessment of air addition system in an energy cost basis, revealed that for all considered compressor configurations except 2-stage tested configuration, the fuel economy benefits of air addition would most likely be able to overcome their respective diluent compression load depending on the operating 118  condition. For these configurations, relatively higher increase in 𝜂𝑠𝑦𝑠 with FDR was seen at higher Φs, indicating high Φs to be most favorable for air addition concept. • A merit function of the form typically used for engine operation optimization studies, was used to identify operating spaces for optimum engine operation with air addition. The optimum solution within the full considered domain was found to be at the highest values of the three control variables: EGR, Φ, and FDR. A calculation of the gradient at this optimum solution, revealed that a better optimum for the considered engine operating mode (global optimum solution) would likely be obtained via the increase of all three control variables. For constrained domains (i.e. Φ fixed) with lower Φ, optimal solution was obtained with lower values of FDR (< 100% FDR). For all domains, the optimum solutions were consistently observed at the highest EGR (25%). These solutions, particularly that of the constrained domains, were sensitive to changes in emission targets, while were not sensitive to various considered compressor configurations.  119   Conclusions and Future Work This work investigates the concept of air addition at a representative operating mode (GIMEP= 16.75 bar; Speed= 1350 rpm) of a heavy duty PIDING engine. Air addition, quantified by FDR (ratio of diluent air to natural gas), was characterized by experimentally analyzing its effects on the combustion, emissions, and performance at various EGR rates and global equivalence ratios (Φ). The system level implications of the air addition concept were determined by considering the compression power required to provide the required diluent flowrates from typical compressors. For this, an industrial 3-stage reciprocating compressor was characterized, and used to develop a 2-stage compression system. This chapter provides a summary for the major conclusions of this work, as well as the guidelines for its continuation.  7.1 General Conclusions  Air addition was demonstrated to be an effective emission (PM, CO, and THC) reduction strategy for a representative heavy-duty mode at various EGR rates and global oxygen equivalence ratios (Φ). Denuded PM (primarily soot) was reduced exponentially with air addition, resulting in reductions comparable to that obtained from modern diesel particulate filter (90% to 97% at 100% FDR). Similarly, CO was also reduced, albeit with a lower magnitude (45% to 95%). The THC (68 to 82% methane) and unburnt methane emissions were significantly reduced only at high EGR (> 12.5%). The reductions in THC and unburnt methane (CH4) emissions were of similar magnitude: 60 to 85% reductions at the highest EGR (25%) with 100% FDR. This is particularly beneficial, as CH4 reductions are not generally possible using exhaust after treatment. Air addition resulted in indicated efficiency improvements (𝜂𝑖,𝑔) of the order of 2.5%-point, due to different mechanisms dependent on the EGR rate. By defining a parameter 120  combustion independent efficiency (𝜂∗), it was determined that efficiency improvements at high EGR resulted from combustion efficiency improvements, while the most likely mechanism for the improvements observed with increasing FDR at the lowest EGR is due to reductions in exhaust enthalpy. Expansion work of diluent contributed to nearly 0.5%-point increase at 100% FDR. 𝜂𝑖,𝑔 improvements, especially at high Φs, where the 𝜂𝑖,𝑔 increase was maximum, indicated the potential to counter for parasitic loads of diluent compressor and is discussed below.   Some limitations to air addition were observed, including increased NOx (up to a factor of 2.5 increase), which was not seen in a previous N2 addition study [21,23]. Based on a response surface analysis, at the highest Φ (0.77), an increase of 6.5% in EGR can mitigate the NOx increase. Increase in maximum pressure rise rate with air addition was noted at higher FDR (100%), which at higher EGR (25%) was significant even at lower FDRs (≥ 25%).  To elucidate the mechanistic effects of air addition on the pollutants, a previous N2 addition study [21,23] and thermodynamic analyses were considered. Similar to N2 addition, the reduced PM and CO emissions are also attributed to improved late cycle mixing from the additional turbulent kinetic energy of the diluent. This is because of similar turbulence effects of both the diluents due to similar transport properties of O2 and N2. Although not conclusively demonstrated, air addition may have an additional chemical effect (relative to N2 addition) on PM and CO emissions, due to additional O2 influencing the chemical kinetics. This may reduce soot precursors such as C2H2 [43], or enhance the CO oxidation due to increased localized oxygen availability. In contrast to the previous N2 addition study, the THC reductions with air addition are not due to a fuel-volume effect (a major THC source), where fuel trapped in injector sac are released into the combustion chamber at lower pressures during the exhaust stroke, and which resulted in linear and consistent reductions in THC with fuel dilution. Such consistent linear reductions in THC 121  emissions were not observed with air addition. Similar to N2 addition, the THC reductions are unlikely due to improved mixing, and therefore the reductions are attributed to an additional effect of air addition (e.g. from O2 species), relative to N2 addition, which becomes significant at higher EGR rates. This effect could be due to other THC sources identified for compression ignition engines in literature (e.g overleaning, reduced localized flame extinctions, reduced bulk quenching of fuel-air mixture at the end of the expansion stroke). An exponential dependence of specific NOx emissions with adiabatic flame temperature (AFT) at higher EGRs indicates that increased flame temperatures affecting the thermal NO route was the major contributor to increased NOx for air addition at these EGR rates. Although this relationship was not seen at low EGR rate, the same NO formation route is hypothesized to be major contributor to NOx increase even at low EGR rates. This is because the effect of air addition on these other pathways (prompt NO at jet core; residence time and species contributions to thermal NOx), are unlikely to be enhanced with reduced EGR rates, where higher flame temperatures would be even more conducive to thermal NOx.  The air addition concept could result in no net parasitic load to the engine depending on engine operating conditions, the number of compressor stages, and performance of the chosen compressor. The dependency on engine operating conditions was assessed by defining a system efficiency (𝜂𝑠𝑦𝑠), which considered the compression power (𝑃𝑐) required for various diluent compression strategies along with the engine gross indicated efficiency (𝜂𝑖,𝑔) obtained with air addition. This indicated that increase in 𝜂𝑖,𝑔 with air addition, irrespective of EGR rates, was sufficient to compensate for the compression load of a three-stage reciprocating compressor, when operating at high Φ for the considered engine mode. This also indicates that the air addition concept to be more feasible (i.e. free of a net parasitic load) at high load engine operations, where 122  Φs is generally higher. The compressor significantly affects the range of engine operation in which the concept would be feasible. It was determined that the considered three-stage reciprocating compressor had significantly lower overall efficiency (due to low mechanical efficiency) compared to typical values seen in literature. A hypothetical 3-stage compressor configuration with performance estimated based on a reasonable overall efficiency from literature, demonstrated that improved compressor performance can extend the feasibility of the air addition concept to even lower Φ engine operation. Sizing of the diluent compressor is an important consideration for an actual on-board system. Therefore, a prototype a 2-stage compressor configuration was developed and considered for 𝜂𝑠𝑦𝑠 analysis. The 2-stage configuration concept, even with reduced total pressure ratios relative to 3-stage, resulted in significant parasitic loads (order of 4%-point decrease in 𝜂𝑠𝑦𝑠) at all considered engine operating conditions. However, this was determined to be more of a performance issue of the considered compressor rather than the concept of 2-stage compressor operation itself since the specific isentropic power consumed for both 3-stage and 2-stage configuration were comparable. Furthermore, a hypothetical 2-stage configuration with performance estimated similar to that of hypothetical 3-stage indicated the feasibility of air addition across a range of Φs.  Various effects of air addition at different Φ and EGR rates, as well as optimum engine operating space determined by two separate methods, suggest that air addition as an emission reducing in-cylinder strategy would be most effective with engine operating at high values of EGR, Φ, and FDR. The lowest NOx increase was observed at the highest Φ, where the factor of increase was nearly constant (~1.5). The lowest PM-NOx trade-off was observed at the highest EGR, Φ, and FDR. Furthermore, at higher EGR (≥ 12.5%), PM emissions trends showed further potential for reductions beyond highest tested FDR (100%). The highest increase in combustion efficiency 123  (𝜂𝑐) was obtained at the highest EGR and Φ, where potential for further 𝜂𝑐 improvements via increased FDR was also observed. This also indicates continued 𝜂𝑖,𝑔 improvements beyond 25% EGR, suggesting the possibility of air addition to be used in conjunction with ultra-high EGR (> 40%), where significantly low levels of both NOx and soot can be achieved. Combustion instability (COVGIMEP), a major concern with ultra-high EGR operation, was almost linearly reduced with FDR, providing further support for ultra-high EGR operation. Finally, the optimum engine operation space, determined based on a merit function considering reductions of NOx and PM emissions and system efficiency, indicated that air addition should operate best at high values of EGR, Φ, and FDR.   The conclusions presented should be interpreted in context with the limitations of this study. The work considered one operating mode and as such, the effectiveness of air addition on pollutant reduction and 𝜂𝑖,𝑔 improvements reported in this study may vary at other relevant operating modes. For example, the turbulent kinetic energy from added diluent may become less significant at higher speeds, and thereby may significantly affect the magnitude of emission reductions. Also, only two operating parameters were varied, and the effects of air addition may be influenced by other parameters like relative injection timing, injection pressure etc. The conclusions on the pollutant formation mechanisms were drawn based on available literature on diesel and PIDING engines and a basic understanding of combustion process from in-cylinder pressure measurements. Without the availability of optical access in the current engine, these couldn’t be verified based on visual observations of the combustion process. The system efficiency analysis was based on known and measured quantities (e.g. specific motor power, 𝜂𝑖,𝑔), however, other derived parameters like the shaft power of compressor, brake efficiency of the engine may be more relevant for an on-board application context. Only PM and NOx reductions were 124  considered as merits in the merit function to identify optimum engine operating space, although reductions of THC and CO are also considered important in the latest emission regulations. Although the recommended optimum was tolerant to changes (up to a factor of 2) in the emission and system efficiency targets, changes to these targets and use of weights can significantly affect the optimum solution.   7.2 Recommendations for future work The results presented in this study provide guidance to continued assessment and development of the in-cylinder strategy of air addition for commercial applicability. The author provides the following recommendations for future work: • Further mechanistic understanding of the effects of air addition may be achieved by the following: o Demonstrating the chemical effect of O2 species (present in added air) on pollutants and combustion by experimental comparing air addition with nitrogen addition at the same experimental parameters. This can also be achieved via CFD simulations. o Assessment of effects of air addition on the heat transfer, especially soot radiation at different EGR via optical diagnostics. This may help in identifying the mechanism of indicated efficiency improvements at low EGR rate.  • Determine an optimum engine operating space at the tested mode by further increasing EGR, FDR, and Φ. Ultra-high EGR (> 40%) may be especially beneficial to significantly reduce both NOx and soot, while using air addition to ameliorate combustion instability and high THC emissions typical associated with ultra-high EGR  •  Other injection parameters, particularly injection pressure, should also be considered while determining the optimum as they may influence the effectiveness of air addition 125  • Depending on the objective, more meticulous approaches for the determination of the optimum solution should be formulated, for example, by usage of weights; including THC and CO reduction as merits. Determined optimum operation space should be tested on engine.  • The robustness of air addition as a PM, CO, and THC reduction strategy should also be demonstrated at other modes. This may be attempted by determining optimum operating space and testing in the nearby operating space. At the same time, this would also provide more insight on the feasibility of the air addition concept at these modes.  • Since sizing of the diluent compressor is an important consideration, characterizing the power and flowrates of a 2-stage configuration of improved performance is recommended. This may be achieved by adapting an available (in market) 3-stage reciprocating compressor or by custom designing and manufacturing one with required performance. For the former, a 3-stage compressor with specific motor power consumption nearly equal to that of the hypothetical 3-stage compressor (~0.33 kW/kg/hr) is recommended as the selection criteria. For the latter, it is recommended to focus on certain key areas of design: flowrate capacity and friction.  Higher flowrate capacity can be obtained by decreasing the clearance volume to increase volumetric efficiency or by just increasing the swept volume of stages. To reduce friction, a validated system model, based on friction measurements of a 2-stage compressor with detailed friction modelling is recommended. 126  Bibliography [1] U.S. Environmental Protection Agency, 2017, “Profile of Version 1 of the 2014 National Emissions Inventory,” (April) [Online]. Available: https://www.epa.gov/sites/production/files/2017-04/documents/2014neiv1_profile_final_april182017.pdf. [2] 2014, “WHO | 7 Million Premature Deaths Annually Linked to Air Pollution,” WHO [Online]. Available: http://www.who.int/mediacentre/news/releases/2014/air-pollution/en/. [Accessed: 07-Jul-2018]. [3] Brauer, M., Hoek, G., Smit, H. A., de Jongste, J. C., Gerritsen, J., Postma, D. S., Kerkhof, M., and Brunekreef, B., 2007, “Air Pollution and Development of Asthma, Allergy and Infections in a Birth Cohort,” Eur. Respir. J., 29(5), pp. 879–888. [4] Brunekreef, B., and Holgate, S. T., 2002, “Air Pollution and Health.,” Lancet (London, England), 360(9341), pp. 1233–42. [5] Wang, C., Tu, Y., Yu, Z., and Lu, R., 2015, “PM2.5 and Cardiovascular Diseases in the Elderly: An Overview,” Int. J. Environ. Res. Public Health, 12(7), pp. 8187–8197. [6] Bond, T. C., Doherty, S. J., Fahey, D. W., Forster, P. M., Berntsen, T., Deangelo, B. J., Flanner, M. G., Ghan, S., Kärcher, B., Koch, D., Kinne, S., Kondo, Y., Quinn, P. K., Sarofim, M. C., Schultz, M. G., Schulz, M., Venkataraman, C., Zhang, H., Zhang, S., Bellouin, N., Guttikunda, S. K., Hopke, P. K., Jacobson, M. Z., Kaiser, J. W., Klimont, Z., Lohmann, U., Schwarz, J. P., Shindell, D., Storelvmo, T., Warren, S. G., and Zender, C. S., 2013, “Bounding the Role of Black Carbon in the Climate System: A Scientific Assessment,” J. Geophys. Res. Atmos., 118(11), pp. 5380–5552. [7] Jacobson, M. Z., 2010, “Short-Term Effects of Controlling Fossil-Fuel Soot, Biofuel Soot 127  and Gases, and Methane on Climate, Arctic Ice, and Air Pollution Health,” J. Geophys. Res. Atmos., 115(14). [8] Hansen, J., and Nazarenko, L., 2004, “Soot Climate Forcing via Snow and Ice Albedos,” Proc. Natl. Acad. Sci., 101(2), pp. 423–428. [9] Seinfeld, J. H., and Pandis, S. N., 2006, Atmospheric Chemistry and Physics: From Air Pollution to Climate Change, New York. [10] World Health Organization, “Ambient (Outdoor) Air Quality and Health” [Online]. Available: http://www.who.int/news-room/fact-sheets/detail/ambient-(outdoor)-air-quality-and-health. [Accessed: 17-Jul-2018]. [11] Lashof, D. A., and Ahuja, D. R., 1990, “Relative Contributions of Greenhouse Gas Emissions to Global Warming,” Nature, 344(6266), pp. 529–531. [12] “How Emissions Are Regulated” [Online]. Available: https://www.dieselnet.com/standards/intro.html. [Accessed: 08-Jul-2018]. [13] “Emission Standards: USA: Heavy-Duty Onroad Engines” [Online]. Available: https://www.dieselnet.com/standards/us/hd.php. [Accessed: 17-Jul-2018]. [14] US EPA, O., “Air Quality - National Summary” [Online]. Available: https://www.epa.gov/air-trends/air-quality-national-summary. [Accessed: 07-Jul-2018]. [15] Russell, A., and Epling, W. S., 2011, “Diesel Oxidation Catalysts,” Catal. Rev., 53(4), pp. 337–423. [16] Piumetti, M., Bensaid, S., Fino, D., and Russo, N., 2015, “Catalysis in Diesel Engine NOx Aftertreatment: A Review,” Catal. Struct. React., 1(4), pp. 155–173. [17] Majewski, W. A., “Diesel Particulate Filters” [Online]. Available: https://www.dieselnet.com/tech/dpf.php. 128  [18] Gekas, I., Gabrielsson, P., Johansen, K., Bjørn, I., Kjær, J. H., Reczek, W., and Cartellieri, W., 2002, “Performance of a Urea SCR System Combined with a PM and Fuel Optimized Heavy-Duty Diesel Engine Able to Achieve the Euro V Emission Limits,” Sae Tech. Pap., pp. 2002-01-2885. [19] Warnatz, J., Mass, U., and Dibble, R. W., 2009, Combustion, Springer (India) Private Limited, New Delhi. [20] Harrington, J., Munshi, S., Nedelcu, C., Ouellette, P., Thompson, J., and Whitfield, S., 2002, “Direct Injection of Natural Gas in a Heavy-Duty Diesel Engine Reprinted From : Diesel Engine Experiments,” SAE Tech., pp. 2001-01-1630. [21] McTaggart-cowan, G., 2006, “Pollutant Formation in a Gaseous-Fuelled, Direct Injection Engine, PhD Thesis,” University of British Columbia, Vancouver, Canada. [22] Faghani, E., 2015, “Effect of Injection Strategies on Particulate Matter Emissions from HPDI Natural-Gas Engine, PhD Thesis,” University of British Columbia. [23] McTaggart-Cowan, G. P., Rogak, S. N., Hill, P. G., Munshi, S. R., and Bushe, W. K., 2008, “The Effects of Fuel Dilution in a Natural-Gas Direct-Injection Engine,” Proc. Inst. Mech. Eng. Part D J. Automob. Eng., 222(3), pp. 441–453. [24] Huang, J., Hill, P. G., Munshi, S., and McTaggart-Cowan, G. P., 2014, “Air-Enriched Gaseous Fuel Direct Injection for an Internal Combustion Engine.” [25] Harrington, J., Munshi, S., Nedelcu, C., Ouellette, P., Thompson, J., and Whitfield, S., 2002, “Direct Injection of Natural Gas in a Heavy-Duty Diesel Engine,” SAE Tech. Pap. 2002-01-1630. [26] McTaggart-Cowan, G. P., Bushe, W. K., Rogak, S. N., Hill, P. G., and Munshi, S. R., 2005, “PM and NOx Reduction by Injection Parameter Alterations in a Direct Injected, 129  Pilot Ignited, Heavy Duty Natural Gas Engine with EGR at Various Operating Conditions,” SAE Tech. Pap. 2005-01-1733. [27] Patychuk, B. D., 2013, “Particulate Matter Emission Characterization From a Natural Gas High-Pressure Direct-Injection Engine, MASc Thesis,” University of British Columbia, Vancouver, Canada. [28] McTaggart-Cowan, G., Bushe, W. K., Hill, P. G., and Munshi, S. R., 2004, “NOx Reduction from a Heavy-Duty Diesel Engine with Direct Injection of Natural Gas and Cooled Exhaust Gas Recirculation,” Int. J. Engine Res., 5(2), pp. 175–191. [29] McTaggart-Cowan, Bushe, Rogak, McTaggart-Cowan, G., Bushe, W., and Rogak, S., 2003, “Injection Parameter Effects on a Direct Injected, Pilot Ignited, Heavy Duty Natural Gas Engine with EGR,” SAE Tech. Pap. 2003-01-3089. [30] Faghani, E., Kheirkhah, P., Mabson, C. W. J., McTaggart-Cowan, G., Kirchen, P., and Rogak, S., 2017, “Effect of Injection Strategies on Emissions from a Pilot-Ignited Direct-Injection Natural-Gas Engine- Part II: Slightly Premixed Combustion,” SAE Tech. Pap. 2017-01-0763. [31] Faghani, E., Kheirkhah, P., Mabson, C. W. J., McTaggart-Cowan, G., Kirchen, P., and Rogak, S., 2017, “Effect of Injection Strategies on Emissions from a Pilot-Ignited Direct-Injection Natural-Gas Engine- Part I: Late Post Injection,” SAE Int. 2017-01-0774. [32] Dec, J., 1997, “A Conceptual Model of DI Diesel Combustion Based on Laser-Sheet Imaging,” Sae, (412), p. 970873. [33] Huang, J., 2006, “Natural Gas Combustion Under Engine-Relevant Conditions, PhD Thesis,” University of British Columbia, Vancouver, Canada. [34] Kheirkhah, P., 2015, “CFD Modeling of Injection Strategies in a High-Pressure Direct- 130  Injection (HPDI) Natural Gas Engine, MASc Thesis,” (April). [35] Ouellette, P., Mtui, P. L., and Hill, P. G., 1998, “Numerical Simulations of Directly Injected Natural Gas and Pilot Diesel Fuel in a Two-Stroke Compression Ignition Engine,” SAE Tech. Pap. 981400. [36] Li, G., Ouellette, P., Dumitrescu, S., and Hill, P. G., 1999, “Optimization Study of Pilot-Ignited Natural Gas Direct-Injection in Diesel Engines,” SAE Tech. al Pap. 1999-01-3556. [37] Mabson, C. W. J., 2015, “Emissions Characterization of Paired Gaseous Jets in a Pilot-Ignited Natural -Gas Engine, MASc Thesis,” University of British Columbia, Vancouver, Canada. [38] Rochussen, J., Khosravi, M., and Kirchen, P., 2018, “Effects of Relative Injection Timing on the Combustion Processes of Pilot Ignited Direct Injected Natural Gas,” Proceedings of Combustion Institute, Spring Technical Meeting. [39] Hatzipanagiotou, A., Marko, F., Koenig, G., Krueger, C., Wenzel, P., and Koch, T., 2018, “Numerical and Optical Analysis of Heterogeneous Gas Combustion with Diesel Pilot Ignition in a Commercial Vehicle Engine,” Int. J. Engine Res., 19(1), pp. 109–119. [40] Kittelson, D. B., 1998, “Engines and Nanoparticles: A Review,” J. Aerosol Sci., 29(5–6), pp. 575–588. [41] Roesler, J. F., Martinot, S., McEnally, C. S., Pfefferle, L. D., Delfau, J. L., and Vovelle, C., 2003, “Investigating the Role of Methane on the Growth of Aromatic Hydrocarbons and Soot in Fundamental Combustion Processes,” Combust. Flame, 134(3), pp. 249–260. [42] Tree, D. R., and Svensson, K. I., 2007, “Soot Processes in Compression Ignition Engines,” Prog. Energy Combust. Sci., 33(3), pp. 272–309. [43] Omer L Gulder, 1995, “Effects of Oxygen on Soot Formation in Methane Diffusion 131  Flames,” Combust. Sci. Technol., 101, pp. 302–310. [44] Cao, S., Ma, B., Giassi, D., Bennett, B. A. V., Long, M. B., and Smooke, M. D., 2018, “Effects of Pressure and Fuel Dilution on Coflow Laminar Methane–air Diffusion Flames: A Computational and Experimental Study,” Combust. Theory Model., 22(2), pp. 316–337. [45] Tian, B., Gao, Y., Zhang, C., and Hochgreb, S., 2018, “Soot Measurement in Diluted Methane Diffusion Flames by Multi-Pass Extinction and Laser-Induced Incandescence,” Combust. Flame, 192, pp. 224–237. [46] Heywood, J. B., 1988, Internal Combustion Engine Fundamentals, McGraw-Hill Education, New York. [47] Gomez, A., and Glassman, I., 1988, “Quantitative Comparison of Fuel Soot Formation Rates in Laminar Diffusion Flames,” Symp. Combust., 21(1), pp. 1087–1095. [48] Axelbaum, R. L., and Law, C. K., 1991, “Soot Formation and Inert Addition in Diffusion Flames,” Symp. Combust., 23(1), pp. 1517–1523. [49] Hill, P. G., and McTaggart‐Cowan, G. P., 2005, “Nitrogen Oxide Production in a Diesel Engine Fueled by Natural Gas,” SAE Tech. al Pap. 2005-01-1727. [50] Musculus, M. P. B., 2004, “On the Correlation between NOx Emissions and the Diesel Premixed Burn,” (724). [51] Mohamed, H., Bentîcha, H., and Mohamed, S., 2009, “Numerical Modeling of the Effects of Fuel Dilution and Strain Rate on Reaction Zone Structure and Nox Formation in Flameless Combustion,” Combust. Sci. Technol., 181(8), pp. 1078–1091. [52] Feese, J. J., and Turns, S. R., 1998, “Nitric Oxide Emissions from Laminar Diffusion Flames: Effects of Air-Side versus Fuel-Side Diluent Addition,” Combust. Flame, 113(1–2), pp. 66–78. 132  [53] Beltrame, A., Porshnev, P., Merchan-Merchan, W., Saveliev, A., Fridman, A., Kennedy, L. A., Petrova, O., Zhdanok, S., Amouri, F., and Charon, O., 2001, “Soot and NO Formation in Methane-Oxygen Enriched Diffusion Flames,” Combust. Flame, 124(1–2), pp. 295–310. [54] Turns, S. R., 2012, An Introduction to Combustion : Concepts and Applications, McGraw-Hill, New Delhi. [55] Stone, R., 1993, Introduction to Internal Combustion Engines, Society of Automotive Engineers International, Warrendale. [56] Ludwig, E. E., 2001, Applied Process Design for Chemical and Petrochemical Plants: Vol 3, Gulf Professional Publishing. [57] Campbell, J. M., 1992, “Gas Conditioning and Processing: The Equipment Modules,” Campbell Petroleum Series, pp. 193–238. [58] Rasmussen, B. D., and Jakobsen, A., 2000, “Review of Compressor Models and Performance Characterizing Variables.,” Int. Compress. Eng. Conf., pp. 515–522. [59] 1981, “Gas Properties and Compressor Data: Form 3519-D” [Online]. Available: http://edge.rit.edu/edge/P08452/public/Gas Properties and Compressor Data. [Accessed: 15-Jan-2018]. [60] Dr. Mahmood Moshfeghian, “How to Estimate Compressor Efficiency? | Campbell Tip of the Month” [Online]. Available: http://www.jmcampbell.com/tip-of-the-month/2015/07/how-to-estimate-compressor-efficiency/. [Accessed: 19-Jan-2018]. [61] Motta, S., Braga, S., and Parise, J., 1996, “A Study on the Polytrupic Exponent of Reciprocating Hermetic Compressors.” [62] Cavallini, A., Doretti, L., Longo, G. A., Rosseto, L., Bella, B., and Zanneiro, A., 1996, 133  “Thermal Analysis of a Hermetic Reciprocating Compressor,” Proc. 1996 Int. Compress. Eng. Conf. Purdue, pp. 535–540. [63] Dr. Mahmood Moshfeghian, “How to Estimate Compressor Efficiency? | Campbell Tip of the Month.” [64] Moore, J. J., Blieske, M., Delgado, H., Lerche, A., Alsup, C., Pacheco, J., Bough, M., and Byard, D., 2009, “Research and Development Needs for Advanced Compression of Large Volumes of Carbon Dioxide” [Online]. Available: https://www.nist.gov/sites/default/files/documents/pml/high_megawatt/6_1-Approved-Moore.pdf. [Accessed: 19-Jan-2018]. [65] Ribas, F. A., Deschamps, C. J., Fagotti, F., Morriesen, A., and Dutra, T., 2008, “Thermal Analysis of Reciprocating Compressors. A Critical Review,” Int. Compress. Eng. Conf., pp. 1–9. [66] Pérez-Segarra, C. D., Rigola, J., Sòria, M., and Oliva, A., 2005, “Detailed Thermodynamic Characterization of Hermetic Reciprocating Compressors,” Int. J. Refrig., 28(4), pp. 579–593. [67] Mantri, P., and Tamma, B., 2014, “Friction Model Development for a Reciprocating Compressor,” Int. Compress. Eng. Conf. [68] Stouffs, P., Tazerout, M., and Wauters, P., 2001, “Thermodynamic Analysis of Reciprocating Compressors,” Int. J. Therm. Sci., 40(1), pp. 52–66. [69] 2009, ISO 1217: 2009(E) - Displacement Compressors- Acceptance Tests. [70] US Department of Energy, 1997, “Determining Electric Motor Load and Efficiency.” [71] Carlisle, 2015, “Energy Loss and Efficiency of Power Transmission Belts,” Third World Energy Eng. Congr. 134  [72] Mctaggart-Cowan, G. P., Bushe, W. K., Hill, P. G., and Munshi, S. R., 2003, “A Supercharged Heavy-Duty Diesel Single-Cylinder Research Engine for High-Pressure Direct Injection of Natural Gas,” Int. J. Engine Res., 4(4), pp. 315–330. [73] Patychuk, B. D., and Rogak, S. N., 2012, “Particulate Matter Emissions from a Natural Gas Fueled High Pressure Direct Injection Engine Tested over the SCRE 9 Mode Test Cycle.,” Proceedings of the Combustion Institute Canadian Section Spring Technical Meeting. [74] McTaggart-Cowan, G., 2001, “The Application of Exhaust Gas Recirculation to a Single Cylinder Compression Ignition Engine Fuelled with Natural Gas,” University of British Columbia, Vancouver, Canada. [75] “Northshore Artic 1000 Series Owners Manual: Edition 1.4” [Online]. Available: https://store-84dac.mybigcommerce.com/content/manuals/Arctic/Arctic 1000 User Manual - 4-2012.pdf. [Accessed: 14-Jan-2018]. [76] 2016, ELITE pro XC TM ELOG TM 15 Operator ’ s Guide, Dent Instruments. [77] Faghani, E., and Kheirkhah, P., 2014, “Transient Slice Model ( TSM ): A Phenomenological Model for Simulation of HPDI Engine Including PM,” pp. 1–48. [78] Holman, J. P., J.P. Holman. Experimental Methods for Engineers. [79] Patychuk, B., Wu, N., Mctaggart-cowan, G., Hill, P., and Munshi, S., 2015, “Intake and Exhaust Valve Timing Control on a Heavy-Duty , Direct-Injection Natural Gas Engine.” [80] Chmela, F. G., and Orthaber, G. C., 1999, “Rate of Heat Release Prediction for Direct Injection Diesel Engines Based on Purely Mixing Controlled Combustion,” SAE Tech. Pap. 1999-01-0186. [81] Goodwin, D. G., Moffat, H. K., and Speth, R. L., 2018, “Cantera 2.3: An Open-Source 135  Suite of Tools for Problems Involving Chemical Kinetics, Thermodynamics, and Transport Processes” [Online]. Available: https://cantera.org/. [Accessed: 15-Sep-2018]. [82] Faghani, E., Patychuk, B., McTaggart-Cowan, G., and Rogak, S., 2013, “Soot Emission Reduction from Post Injection Strategies in a High Pressure Direct-Injection Natural Gas Engine,” SAE Int. 2013-24-0114. [83] Caton, J. A., 1982, “Comparisons of Averaged and Mass-Averaged Exhaust Gas Temperatures for a Spark-Ignited Engine,” SAE Tech. Pap. Ser., (820050). [84] Chen, S. K., and Flynn, P. F., 1965, “Development of a Single Cylinder Compression Ignition Research Engine,” p. 650733. [85] “Reciprocating Compressor -PetroWiki” [Online]. Available: http://petrowiki.org/Reciprocating_compressor. [Accessed: 21-Jan-2018]. [86] Ludwig, E., 2015, “Applied Process Design for Chemical and Petrochemical Plants: Vol 3 (Pages for Reciprocating Compression),” Gulf Professional Publishing, pp. 368–455. [87] “Interactive Response Surface Modeling - MATLAB Rstool” [Online]. Available: https://www.mathworks.com/help/stats/rstool.html. [Accessed: 04-Jun-2018]. [88] Myers, R.H; Montgomery, D.C; Anderson-Cook, C. ., 2009, Response Surface Methodology: Process and Product Optimization Using Designed Experiments, John Wiley & Sons, Inc., Hoboken. [89] Montgomery, D. T., and Reitz, R. D., 2000, “Optimization of Heavy-Duty Diesel Engine Operating Parameters Using a Response Surface Method,” SAE Tech. Pap. 2000-01-1962. [90] Razmjooei, K., and Nagarajan, G., 2010, “Optimization of DI Diesel Engine Operating Parameters Using a Response Surface Method,” SAE Int. 2010-01-1262. [91] Wickman, D. D., Senecal, P. K., and Reitz, R. D., 2001, “Diesel Engine Combustion 136  Chamber Geometry Optimization Using Genetic Algorithms and Multi-Dimensional Spray and Combustion Modeling,” 2001(724). [92] “Emission Standards: Europe: Heavy-Duty Truck and Bus Engines” [Online]. Available: https://www.dieselnet.com/standards/eu/hd.php. [Accessed: 03-Jul-2018]. [93] Gregory P. Smith, David M. Golden, Michael Frenklach, Nigel W. Moriarty, Boris Eiteneer, Mikhail Goldenberg, C. Thomas Bowman, Ronald K. Hanson, Soonho Song, William C. Gardiner, J., Vitali V. Lissianski, and and Zhiwei Qin, “GRI-Mech 3.0” [Online]. Available: http://www.me.berkeley.edu/gri_mech/. [Accessed: 23-Aug-2018].  137  Appendices Appendix A Calculation of Uncertainty  The calculated uncertainty presented in Section 3.3 is based on the propagation of error methodology described in the experimental methods textbook by Holman [78]:  𝑈𝑅 = [∑ (𝜕𝑅𝜕𝑥𝑖𝑈𝑥𝑖)2𝑖]12 (7.1) where 𝑅 = 𝑓(𝑥𝑖) is equation of the parameter for which the uncertainty needs to be calculated; 𝑈𝑥𝑖 is the absolute uncertainty in 𝑥𝑖, often determined from instrument specifications; and 𝑈𝑅 is the absolute uncertainty in the parameter R, referred to as the calculated uncertainty in  this thesis. A.1 Uncertainty Analysis: Instrument Specifications   Table 7.1 Uncertainty relevant information: Exhaust Gas Analyzer.  D: drift; L: linearity error; N: noise; R: repeatability; mv: measured value; FS: Full scale; u= relative uncertainty; U=absolute uncertainty Emission name Manufacturer | Model | Principle Uncertainty specifications  THC  Pierburg | FID 4000 hhd | FID D= 0.5% mv; L= 1% mv; N= 1% mv; R=0.5% mv CH4 Pierburg | FID 4000 hhd | FID D= 0.5% mv + 1% THC mv/hr; L= 1% mv;  N= 1% mv; R=0.5% mv + 2% of THC mv CO ABB | Uras 14 EGA | NDIR D= 1% span/24 hr; N= 0.2% span; L= 1% mv CO2 exhaust ABB | Uras 14 EGA | NDIR D= 1% span/24 hr; N= 0.2% span; L= 1% mv NOx Pierburg | CLD 4000 hhd | Chemiluminescent D= 0.5% mv; L= 1% mv; N= 1% mv; R=0.3% mv  Table 7.2 Uncertainty relevant information: PM measurements.  D: Drift; Res: resolution; mv: measured value; FS: full scale; u= relative uncertainty; U= absolute uncertainty Emission name Manufacturer | Model | Principle Uncertainty specifications  PM TSI | DustTrak DRX Aerosol Monitor 8533| Light scattering mass  0.1% mv or 0.001 mg/m3, whichever is greater; 0.002 mg/m3 in 24 hrs CO2 low range California Analytical Instruments| Model 100| NDIR 2% FS; FS= 1000 ppm 138   Table 7.3 Uncertainty relevant information: Miscellaneous measurements.  D: Drift; Res: resolution; mv: measured value; FS: Full scale; u= relative uncertainty; U= absolute uncertainty; L: linearity error; R: repeatability Measured parameter Manufacturer | Model | Principle Uncertainty specifications  CNG flowrate Endress+Hauser| Promass 80A | Coriolis 0.5% of mv Air addition flowrate Endress+Hauser| Promass 80A | Coriolis 0.5% of mv Diesel flowrate Adam Equipment PGW 4502e precision balance Res= 0.01 g; R= 0.01g;  L= 0.02 g  A.2 Sample Uncertainty Calculations This section shows the detailed calculations involved in the determination of calculated uncertainty of specific emissions and gross indicated efficiency. The calculation methodology of the uncertainty of the merit function value at the determined optimum solution in Section 6.4 is also shown.  Specific PM Emissions The expression for determining the specific PM emissions is given below: 𝑃𝑀𝑠 ∝[𝑃𝑀].𝐷𝑅.?̇?𝑒𝑥ℎ𝑃  ∝[𝑃𝑀].𝐷𝑅.?̇?𝑒𝑥ℎ𝐺𝐼𝑀𝐸𝑃.𝑁 where  𝑃𝑀𝑠 is specific PM emissions in g/kWhr, [𝑃𝑀] is the PM concentration in mg/m3, DR is the exhaust dilution ratio, ?̇?𝑒𝑥ℎ is the exhaust flowrate, P is gross indicated power consumption of the engine, 𝐺𝐼𝑀𝐸𝑃 is gross indicated mean effective pressure, and N is the speed of the engine. The relative uncertainty in power specific PM emissions (𝑢𝑃𝑀𝑠) was calculated as follows:  𝑢𝑃𝑀𝑠 = √(𝑢[𝑃𝑀])2 + (𝑢𝐷𝑅)2 + (𝑢?̇?𝑒𝑥ℎ)2 + (−𝑢𝐺𝐼𝑀𝐸𝑃)2 + (−𝑢𝑁)2 (7.2) 139  Individual expressions were considered for the calculation of relative uncertainties of the components in the right-hand side of the Equation (7.2) and is described below.  For relative uncertainty calculation in DR measurement, a simplified expression which neglected the effect of water vapor in exhaust was considered as follows:    𝐷𝑅 =[𝐶𝑂2,𝑒𝑥ℎ] − [𝐶𝑂2,𝑑𝑖𝑙][𝐶𝑂2,𝑑𝑖𝑙−𝑠] − [𝐶𝑂2,𝑑𝑖𝑙]=𝑋𝑌 (7.3) where [𝐶𝑂2,𝑒𝑥ℎ] is the measured CO2 concentration in the exhaust stream from raw emissions sample, [𝐶𝑂2,𝑑𝑖𝑙] is the concentration of the CO2 in the exhaust sample diluent (air) and was assumed constant,  [𝐶𝑂2,𝑑𝑖𝑙−𝑠] is the measured CO2 concentration in the diluted exhaust sample. Based on the dependency of DR on its constituent components (Equation (7.3)), the uncertainty calculation of  𝑢𝑟,𝐷𝑅 therefore, can be expressed as follows:  𝑢𝐷𝑅 = √(𝑢𝑥)2 + (𝑢𝑦)2 (7.4) where 𝑢𝑥 =𝑈𝐶𝑂2,𝑒𝑥ℎ[𝐶𝑂2,𝑒𝑥ℎ]−[𝐶𝑂2,𝑑𝑖𝑙] , 𝑢𝑦 =𝑈𝐶𝑂2,𝑑𝑖𝑙−𝑠[𝐶𝑂2,𝑑𝑖𝑙−𝑠]−[𝐶𝑂2,𝑑𝑖𝑙], 𝑈𝐶𝑂2,𝑒𝑥ℎ and 𝑈𝐶𝑂2,𝑑𝑖𝑙−𝑠 are absolute uncertainties in [𝐶𝑂2,𝑒𝑥ℎ] and [𝐶𝑂2,𝑑𝑖𝑙−𝑠], respectively.  The relative uncertainty of ?̇?𝑒𝑥ℎ (𝑢?̇?𝑒𝑥ℎ), which is taken as equal to the sum of fresh intake air flow, pilot diesel flow, natural gas fuel flow and air addition flow, was determined by uncertainty in fresh intake air flow (4% of measured value). This is because instrument uncertainty from natural gas flow, air addition flow and pilot diesel flow are much lower compared to that from fresh intake air flow. Relative uncertainty in the GIMEP (𝑢𝐺𝐼𝑀𝐸𝑃) for a particular operating point  was represented by the coefficient of variance (COV= ratio of 1 standard deviation to mean) of 140  measured GIMEP from 45 consecutive cycles. For all tested operating points in the current study, the COV varied from 0.5 to 2.5% with 1% being the mean of COV of GIMEP at the primary repeat points 𝑈𝑁 was represented by the operator tolerance in setting the engine speed i.e. 10 rpm, and therefore 𝑢𝑁 can be calculated from it.   Gaseous Emissions Relative uncertainty in specific gaseous emissions (𝑢𝑔𝑎𝑠𝑠; 𝑔𝑎𝑠 = 𝑁𝑂𝑥, 𝑇𝐻𝐶, 𝐶𝐻4 𝐶𝑂, 𝐶𝑂2) are calculated in a similar methodology as shown for PMs in Section A.2: 𝑢𝑔𝑎𝑠𝑠 = √(𝑢[𝑔𝑎𝑠])2 + (𝑢?̇?𝑒𝑥ℎ)2 + (−𝑢𝐺𝐼𝑀𝐸𝑃)2 + (−𝑢𝑁)2 where 𝑢[𝑔𝑎𝑠] is the relative uncertainty in the concentration of the gaseous emission (𝑁𝑂𝑥, 𝑇𝐻𝐶, 𝐶𝐻4 𝐶𝑂, 𝐶𝑂2) and is determined based on the instrument specifications of the corresponding gas analyzers tabulated in Table 7.1.   Gross Indicated Efficiency The expression for determining gross indicated efficiency (𝜂𝑖,𝑔) is given below: 𝜂𝑖,𝑔= 𝑃𝑖,𝑔(?̇?𝑁𝐺.𝐿𝐻𝑉𝑁𝐺 +?̇?𝑑.𝐿𝐻𝑉𝑑)= 𝐺𝐼𝑃𝑌  where 𝑃𝑖,𝑔, ?̇?𝑁𝐺, 𝐿𝐻𝑉𝑁𝐺, ?̇?𝑑, and 𝐿𝐻𝑉𝑑 are gross indicated power (kW), natural gas mass flowrate, lower heating value of natural gas, pilot diesel mass flowrate, and lower heating value of diesel, respectively. Relative uncertainty in 𝜂𝑖,𝑔, represented as 𝑢𝜂𝑖,𝑔 can be calculated as follows: 𝑢𝜂𝑖,𝑔  = √(𝑢𝐺𝐼𝑃)2 + (𝑢𝑌)2 141  where 𝑢𝑃𝑖,𝑔= √(𝑢𝐺𝐼𝑀𝐸𝑃)2 + (𝑢𝑁)2, 𝑢𝐺𝐼𝑀𝐸𝑃, 𝑢𝑁 are as determined in Section A.2, 𝑢𝑌 =√(𝑈?̇?𝑁𝐺 .𝑄𝑁𝐺 )2 + (𝑈?̇?𝑑 .𝑄𝑑 )2 (?̇?𝑁𝐺.𝑄𝑁𝐺  + ?̇?𝑑.𝑄𝑑) , 𝑈?̇?𝑁𝐺  is absolute uncertainty in the mass flowrate measurement of natural gas and 𝑈?̇?𝑑 is the absolute uncertainty in the mass flowrate measurement of diesel flowrate. 𝑈?̇?𝑁𝐺 , and 𝑈?̇?𝑑  were determined from the instrument specifications in Table 7.3.   Merit function used for Optimization Merit function (𝑓𝑚) used in Section 6.4, is of the following form:   𝑓𝑚 = 1000𝑌−1 Where 𝑌 = (𝑃𝑀𝑃𝑀𝑡)2 + (𝑁𝑂𝑥𝑁𝑂𝑥𝑡)2 + (𝜂𝑖,𝑔𝑡𝜂𝑖,𝑔) with the parameters as described in Section 6.4. Using Equation (7.1), 𝑈𝑓𝑚  results in the following: 𝑈𝑓𝑚 = √(−1000𝑈𝑌𝑌2)2 =𝑓𝑚2𝑈𝑌1000 Again, using Equation (7.1), 𝑈𝑌 can be further expanded as follows:   𝑈𝑌 = √(2𝑃𝑀. 𝑈𝑃𝑀𝑃𝑀𝑡)2 + (2𝑁𝑂𝑥. 𝑈𝑁𝑂𝑥𝑁𝑂𝑥𝑡)2 + (−1(𝜂𝑖,𝑔𝑡𝜂𝑖,𝑔)2𝑈𝜂𝑖,𝑔)2 Individual uncertainties: 𝑈𝑃𝑀, 𝑈𝑁𝑂𝑥, and 𝑈𝜂𝑖,𝑔 are then evaluated from the standard errors of their respective response surface fits (0) to give a confidence interval of 95% (i.e. 1.96 × 𝜎𝑠𝑡𝑑). Measurement uncertainties are also included whenever the standard errors are comparable to the measurement uncertainties.  The uncertainty at the value of merit function at the optimum solution (𝑓𝑚∗) is represented as 𝑈𝑓𝑚∗ .142  Appendix B On-Engine Air Addition Experiments: Additional Information B.1  Determination of Realistic Engine Operating Mode A standard determination of the validity of an engine operating strategy to reduce emissions, requires it to be tested in various operating modes, with a mode defined as an engine speed and load combination. The emissions characteristics will be dependent on more than 9 mode parameters, resulting in more than 512 operating points which just two levels of variation in each parameter. Such large testing campaigns are usually undertaken by engine calibration engineers in established industrial engine testing centers and was not possible within the scope and duration of this research project, therefore, a single ‘suitable’ mode with just 3 variable operation parameters was determined for characterization of air addition in SCRE. The ‘suitable’ mode chosen was determined with the following objectives in mind:  • Determine a suitable gas rail pressure upper limit for high load modes to prevent gas rail pressure fluctuations that is inherent to the facility when the gas rail pressure is close to supply pressure • Determine a mode representative of realistic heavy-duty operation i.e. high load (> 70% of maximum engine load) and slightly lower engine speeds (1000- 1500 rpm) • Assess the ‘Mode flexibility’ i.e. the mode should provide sufficient operating space to complete planned test matrix without exceeding constraints like SCRE’s warning RESD limits (in-cylinder peak pressure, maximum exhaust back pressure, exhaust temperatures, misfires etc.) • The mode should have sufficient PM emissions in order to demonstrate the PM reduction effectiveness of air addition  143  The gas rail pressure (GRP) stability in the experimental facility depended on the flowrate of fuel consumed by the SCRE and the difference of the natural gas supply pressure to outlet pressure (nearly equal to GRP) of the natural gas dome loaded regulator. The mass flowrate through a nozzle, the injector nozzle is proportional to the upstream pressure, therefore, any drift in the GRP translates to drift in the natural gas fuel flow rate. Since, it is critical to minimize GRP drifts to operate in the same load, therefore, GRP stability was ensured through the following steps: • The engine was operated in B75 mode (GIMEP= 16.5 bar; Speed= 1500 rpm) with a GRP of 20 MPa. This mode had been traditionally used by previous SCRE researchers for their measurements and typically consumed a NG flowrate ~7.68 kg/hr [22,37] • The GRP was varied from 20 MPa to 24 MPa (lower limit of NG supply pressure ~24.8 MPa) in increments of 0.5 MPa, while maintaining the same load via reduced gas pulse widths (GPW). Higher GRPs allow achievement of high loads with smaller GPWs. However, GRPs closer to the natural gas supply pressure would have a higher tendency to drift. With each of these incremented GRPs, the engine was operated for 15 minutes to have a good estimate of the mean drift.  • 22 MPa was found to be the highest GRP which could be operated at mode B75, without a drift in mean GRP and therefore, was chosen as the GRP for future multi-mode tests at high loads.   The determination of a heavy-duty and ‘flexible’ mode with sufficient PM output, was assessed through a range of multi-mode tests conducted on heavy loads i.e. loads >16.5 bar at engine speeds of 1200, 1350 and 1400 rpm. The ‘flexibility’ assessment required determination of the highest loads that could be achieved at the chosen engine speeds with Φ= 0.55 and EGR= 25% without reaching SCRE’s operation constraints like emergency shutdown warnings (peak pressure, total 144  hydro-carbon emissions) or set point constraints (maximum intake pressure regulator set point). The parameter chosen besides FDR for air addition characterization were Φ and EGR. The lowest Φ (0.58) and the highest EGR (25%) point in the planned test matrix, requires the highest intake pressures which further leads to high peak in-cylinder pressures, and thereby, the highest possibility of exceeding SCRE’s operation constraints. Achieving Φ= 0.55 and EGR= 25% in a mode is therefore an appropriate determinant of the ‘mode flexibility’ i.e. whether the mode would be able to cover the entire range of Φ and EGR to be tested. As can be seen from Figure 7.1, the highest load achievable (GIMEP values encircled in the plot) was found to be limited by peak in-cylinder pressures (SCRE warning limit 175 bar) at the lowest speeds while at higher speeds, the limiting factor was intake pressure regulator setting. Since, peak in-cylinder pressure is a much critical limiting factor from the safety perspective, therefore, loads with GIMEP= 16.75 bar, 16.5 bar at speeds 1350 and 1400 rpm respectively were chosen as ‘flexible modes’. Subsequently, these ‘flexible modes’ were operated with Φ= 0.6 and EGR= 12.5%, and their PM levels were compared to traditionally operated B75 (GIMEP= 16.5 bar, speed=  1500 rpm) known for high levels of PM [21,37]. It can be seen from Figure 7.2, that the PM emissions in the determined ‘flexible mode’ were lower than B75 but were high enough to demonstrate the effectiveness of air addition. Finally, the ‘flexible mode’ of GIMEP= 16.75 bar and speed= 1350 rpm was chosen as the mode for testing because of relatively higher levels of PM and will be subsequently referred to as ‘Mode Z’.      145   Figure 7.1 Determination of ‘mode flexibility’. Note: The GIMEP values in the encircled boxes are the highest loads achievable within SCRE’s operation constraints at the corresponding speeds with EGR= 25%; Φ= 0.55; IHR50= 10oATDC; PSEP= 0.3 ms; GRP= 22 MPa; Diesel= 8.5 mg/inj  Figure 7.2 Determination of high PM ‘flexible modes’.  Note: The operating parameters are: EGR= 12.5%; Φ= 0.58; IHR50= 10oATDC; PSEP= 0.3 ms; GRP= 22 MPa; Diesel= 8.5 mg/inj      146  B.2  Effect of EGR and Φ on AHRR          Figure 7.3 Effect of EGR on AHRR at various Φ and FDR Increasing Φ Increasing FDR 147            Figure 7.4 Effect of Φ on AHRR at various EGR and FDRIncreasing FDR Increasing EGR 148  B.3  Calculation of Adiabatic Flame Temperatures Similar to that described in Hill and McTaggart-Cowan [49] but with modifications made to account for fuel dilution with air: 𝑓. 𝐶𝐻4 + 𝑑. (𝑂2 + 3.76𝑁2) + 𝐴. (𝑂2 + 3.76𝑁2) + 𝑍. (𝛼1𝑂2 + 𝛼2𝑁2 + 𝛼3𝐶𝑂2 + 𝛼4𝐻2𝑂)→ (𝛼1𝑂2 + 𝛼2𝑁2 + 𝛼3𝐶𝑂2 + 𝛼4𝐻2𝑂) where f, d, A, and Z are moles of fuel (considered as only CH4 for ease of calculations), oxygen in diluent air, oxygen in fresh intake air, and fraction of total exhaust that is recirculated, respectively  The mole fractions of primary gases (i.e. O2, N2, CO2, and H2O) in the intake mixture were determined by atomic balance of a mixture consisting of fuel, diluent, and intake air, undergoing complete combustion.  Atom balance yields: 𝛼1 =𝐴+𝐷−2𝑓1−𝑍   𝛼2 =3.76(𝐴+𝐷)1−𝑍      𝛼3 =𝑓1−𝑍  𝛼4 =2𝑓1−𝑍  For a mixture of intake air and exhaust, moles of various species:  𝑛𝑜2,𝑖𝑛 =𝐴+𝑍𝐷−2𝑓𝑧1−𝑍  𝑛𝑁2,𝑖𝑛 =(𝐴+𝑍𝐷)3.761−𝑍  𝑛𝐶𝑜2,𝑖𝑛 =𝑓𝑧1−𝑍  𝑛𝐻2𝑂,𝑖𝑛 =2𝑓𝑧1−𝑍 𝑛𝑡,𝑖𝑛 =4.76(𝐴+𝑍𝑑)+𝑍𝑓1−𝑍 (total moles of intake air)  Mole fractions of various species in this mixture:  𝑋𝑜2,𝑖𝑛 =𝐴+𝑍𝐷−2𝑓𝑧4.76(𝐴+𝑍𝑑)+𝑍𝑓  𝑋𝑁2,𝑖𝑛 =(𝐴+𝑍𝐷)3.764.76(𝐴+𝑍𝑑)+𝑍𝑓  𝑋𝐶𝑜2,𝑖𝑛 =𝑓𝑧4.76(𝐴+𝑍𝑑)+𝑍𝑓 𝑋𝐻2𝑂,𝑖𝑛 =2𝑓𝑧4.76(𝐴+𝑍𝑑)+𝑍𝑓 149   Like diesel engines, PIDING engine combustion is also taken as stoichiometric diffusive burning. Let 𝑛𝑥 moles of diluted fuel combust stoichiometrically with 𝑛𝑡 moles of O2 from the intake mixture: 𝑛𝑥 (𝑓𝑓 + 𝑑. 𝐶𝐻4 +𝑑𝑑 + 𝑓. (𝑂2 + 3.76𝑁2))+ 𝑛𝑡(𝑋𝑜2,𝑖𝑛𝑂2 + 𝑋𝑁2,𝑖𝑛𝑁2 + 𝑋𝐶𝑜2,𝑖𝑛𝐶𝑂2 + 𝑋𝐻2𝑂,𝑖𝑛𝐻2𝑂)→ 𝑎𝑑𝑖𝑎𝑏𝑎𝑡𝑖𝑐 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 𝑝𝑟𝑒𝑠𝑠𝑢𝑟𝑒 𝑒𝑞𝑢𝑖𝑙𝑖𝑏𝑟𝑖𝑢𝑚 𝑐𝑜𝑚𝑏𝑢𝑠𝑡𝑖𝑜𝑛 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑠  1CH4 requires 2O2 molecules for stoichiometric combustion.  𝑛𝑥𝑓𝑓+𝑑  moles of CH4 requires 2. 𝑛𝑥𝑓𝑓+𝑑 moles of O2   2. 𝑛𝑥𝑓𝑓+𝑑=  𝑛𝑥𝑑𝑑+𝑓+ 𝑛𝑡  𝑋𝑜2,𝑖𝑛  𝑛𝑡 =𝑛𝑥𝑓+𝑑(2𝑓−𝑑𝑋𝑜2,𝑖𝑛)  Equivalently, the above equation reduces to:    (𝑓𝐶𝐻4 + 𝑑(𝑂2 + 3.76𝑁2)) + (2𝑓−𝑑𝑋𝑜2,𝑖𝑛)(𝑋𝑜2,𝑖𝑛𝑂2 + 𝑋𝑁2,𝑖𝑛𝑁2 + 𝑋𝐶𝑜2,𝑖𝑛𝐶𝑂2 + 𝑋𝐻2𝑂,𝑖𝑛𝐻2𝑂) →𝑎𝑑𝑖𝑎𝑏𝑎𝑡𝑖𝑐 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 𝑝𝑟𝑒𝑠𝑠𝑢𝑟𝑒 𝑒𝑞𝑢𝑖𝑙𝑖𝑏𝑟𝑖𝑢𝑚 𝑐𝑜𝑚𝑏𝑢𝑠𝑡𝑖𝑜𝑛 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑠 The temperature of the adiabatic constant pressure equilibrium combustion was calculated based on Cantera 2.3 [81], an open-source suite of tools for problems involving chemical kinetics, transport, and thermodynamics. The mechanism used by GRI-Mech 3.0 [93]. The sensitivity of this representative AFT to reactant pressure and temperature is shown in Figure 7.5. (b) 150                            Figure 7.5 Sensitivity of AFT to (a) reactant pressure and (b) reactant temperature (c) fuel jet temperature  (a) (b) (c) 151  B.4  Sensitivity Analysis of Expansion Power of Diluent For choosing a suitable temperature at BDC for expansion power of diluent (?̇?𝑑𝑖𝑙) calculations involved in Section 4.4, a sensitivity analysis was conducted as follows: Table 7.4 Sensitivity of expansion power of diluent to the assumed temperature at BDC.  Note: measured values correspond to engine operated at Mode Z with Φ = 0.68; EGR% = 25%; FDR% = 100% where the measured intake pressure at BDC= 2.6 bar. 𝑇𝑚: measured manifold temperature and 𝑇𝑐: coolant temperature.   Calculated values  T= 𝑇𝑚 = 56oC T= 𝑇𝑚+𝑇𝑐2 = 68oC T= 𝑇𝑐= 80 oC Intake p at BDC (bar) 2.01 2.08 2.16 ?̇?𝑑𝑖𝑙  (kW) 0.38 0.38 0.38 Peak cylinder T (oC) 668 697 726  Sensitivity of expansion power of diluent to the assumed temperature at BDC is tabulated in Table 7.4. The variation in assumed temperature at BDC resulted in no significant impact on the calculation of ?̇?𝑑𝑖𝑙. The actual temperature at BDC for a motoring cycle would be lower than that measured in manifold due to expansion, however, for a fired cycle, the temperature at BDC would also be affected by the heating effect of hot walls and residual gases. Therefore, 𝑇𝑚 from the corresponding fired cycle seemed to be practical temperature for BDC and was chosen for subsequent calculations. Since 𝑇𝑑𝑖𝑙 is also an unknown parameter, another sensitivity analysis shown in Table 7.5 was conducted to determine the effect of chosen 𝑇𝑑𝑖𝑙 on the expansion power of diluent and was subsequently chosen as 20oC for the model.  Table 7.5 Sensitivity of expansion power of diluent to the assumed stagnation temperature of the diluent jet Note: measured values correspond to engine operated at Mode Z with Φ = 0.68; EGR% = 25%; FDR% = 100%, where initial conditions at BDC are calculated from 𝑇𝑚 and measured mass Calculated parameter 𝑇𝑑𝑖𝑙 = 10oC 𝑇𝑑𝑖𝑙 = 20oC 𝑇𝑑𝑖𝑙 = 30oC ?̇?𝑑𝑖𝑙  (kW) 0.36 0.38 0.39  152  The effect of FDR on the motored pressure and temperature on the modelled motored in-cylinder pressure and temperature is shown in Figure 7.6. Diluent addition resulted in a rise of nearly 1 bar while the gas averaged temperature was found to decrease by nearly 10oC.   Figure 7.6 Effect of air addition on modelled motored in-cylinder pressure and temperature Note: The initial conditions at BDC for this case corresponds to that of a measured fired cycle at Φ = 0.68; EGR% = 25%; FDR% = 100%   153  B.5  Summary Tables Summary Table: FDR sweep at EGR= 25% | Φ = 0.6 nr 6 3 1 1 1 Parameter ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ FDR (%) 0 0.24 9.85 0.76 25.93  48.00  98.86  Gross Indicated Efficiency (%) 44.83 0.54 45.10 0.70 45.43  45.82  46.45  Combustion Efficiency (%) 97.73 0.28 97.90 0.55 98.28  98.75  99.49  GISFC - diesel equivalent (g/kW-hr) 187.77 2.28 186.62 2.91 185.26  183.69  181.21  NOx (g/kW-hr) 1.03 0.05 1.11 0.17 1.41  1.66  2.35  DustTrak (g/kW-hr) 0.13 0.02 0.10  0.07  0.03  0.00  CH4 (g/kW-hr) 0.62 0.09 0.58 0.16 0.53  0.42  0.24  tHC (g/kW-hr,C1) 0.76 0.12 0.71 0.18 0.65  0.51  0.31  CO (g/kW-hr) 9.55 1.53 8.83 3.06 6.93  4.86  1.50  CO2 (kg/kW-hr) 0.44 0.01 0.44 0.01 0.44  0.44  0.44  Cylinder exhaust temperature (°C) 462.18 3.08 458.79 9.43 449.57  447.97  447.97  Peak Mixture Averaged T (Celsius) 1276.34 9.81 1279.64 25.93 1276.67  1266.26  1246.46  EGR Temperature 92.36 3.84 95.80 2.72 94.45  93.71  93.53  Diesel injection mass (mg/inj) 8.33 0.18 8.50 0.65 8.39  8.37  8.63  Diesel flow (kg/hr) 0.34 0.01 0.34 0.03 0.34  0.34  0.35  CNG flow (kg/hr) 7.44 0.05 7.42 0.13 7.31  7.30  7.16  Intake surge tank pressure (kPag) 211.76 3.77 212.98 6.24 209.48  210.04  210.22  Air flow (kg/hr) 197.70 1.83 197.47 4.93 196.29  195.84  195.89  Engine speed (rpm) 1343.89 4.39 1344.49 12.96 1340.02  1340.87  1341.78  Gross IMEP (bar) 16.68 0.08 16.73 0.07 16.67  16.78  16.70  Carbon balance (corrected for Air Addition) 0.99 0.01 0.99 0.01 0.98  0.98  0.99  Inlet Oxygen mass fraction  0.19 0.00 0.19 0.00 0.19  0.19  0.20  PSOI set [deg] -26.33 0.81 -26.75 1.64 -28.00  -29.50  -33.50  PPW [ms] 0.64 0.04 0.65 0.03 0.64  0.64  0.64  GPW [ms] 2.31 0.08 2.60 0.38 3.11  3.71  4.71  Pilot Ignition Delay (ms) 1.45 0.08 1.48 0.21 1.55  1.62  1.86  Gas Ignition Delay (ms) 1.50 0.16 1.37 0.11 1.27  1.36  1.45  CA at peak cylinder pressure (bar) 11.09 0.10 10.73 0.61 10.00  9.19  7.42  Peak cylinder pressure (bar) 159.55 1.26 161.01 3.64 162.63  165.42  172.77  Peak AHRR ((kJ/m^3-deg)) 109.39 2.09 105.01 5.11 98.51  90.07  76.96  dP/dCA (bar/deg) 5.79 0.07 5.90 0.14 6.14  6.48  7.04  Combustion duration: IHR10 to IHR90 (deg) 38.83 0.64 39.00 1.24 40.50  42.50  46.00  Early cycle duration: IHR10 to IHR50 (deg) 12.40 0.15 13.00 0.35 14.06  15.60  20.01  Late cycle duration: IHR50 to IHR90 (deg) 26.43 0.55 26.00 1.01 26.44  26.90  25.99  IHR5 (deg ATDC) -4.42 0.21 -5.00 1.24 -6.50  -8.50  -13.00  IHR10 (deg ATDC) -2.50 0.00 -3.17 0.72 -4.50  -6.00  -10.50  IHR90 (deg ATDC) 36.33 0.64 35.83 0.72 36.00  36.50  35.50  System Efficiency: tested 3-stage (%) 44.84 0.54 44.80 0.70 44.63  44.34  43.41    154  Summary Table: FDR sweep at EGR=25% | Φ =0.7 nr 3 1 1 1 1 Parameter ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ FDR (%) 0 0.61 10.91  23.62  48.52  100.02  Gross Indicated Efficiency (%) 43.20 0.10 43.22  43.45  44.18  45.74  Combustion Efficiency (%) 95.97 0.29 96.49  97.03  98.01  99.24  GISFC - diesel equivalent (g/kW-hr) 194.83 0.46 194.74  193.69  190.51  184.00  NOx (g/kW-hr) 0.77 0.05 0.80  0.89  1.08  1.54  DustTrak (g/kW-hr) 0.24 0.07 0.19  0.13  0.07  0.01  CH4 (g/kW-hr) 0.96 0.05 0.80  0.65  0.42  0.21  tHC (g/kW-hr,C1) 1.21 0.10 0.99  0.80  0.51  0.26  CO (g/kW-hr) 18.36 1.61 16.43  14.10  9.44  3.24  CO2 (kg/kW-hr) 0.44 0.00 0.44  0.45  0.45  0.45  Cylinder exhaust temperature (°C) 495.95 2.99 498.22  496.75  494.75  486.17  Peak Mixture Averaged T (Celsius) 1328.19 12.28 1339.78  1336.75  1345.32  1330.31  EGR Temperature 92.09 4.09 95.94  95.68  95.77  95.35  Diesel injection mass (mg/inj) 8.59 0.55 8.90  8.88  9.36  9.13  Diesel flow (kg/hr) 0.35 0.02 0.36  0.36  0.38  0.37  CNG flow (kg/hr) 7.74 0.06 7.78  7.67  7.55  7.35  Intake surge tank pressure (kPag) 188.07 7.80 186.48  184.56  181.57  179.09  Air flow (kg/hr) 181.41 2.57 180.30  178.43  175.20  173.42  Engine speed (rpm) 1344.86 7.53 1344.15  1344.35  1347.04  1347.76  Gross IMEP (bar) 16.71 0.05 16.83  16.68  16.72  16.84  Carbon balance (corrected for Air Addition) 0.99 0.00 0.98  0.98  0.98  0.99  Inlet Oxygen mass fraction  0.19 0.00 0.19  0.19  0.19  0.19  PSOI set [deg] -27.33 0.95 -27.50  -28.00  -29.25  -31.50  PPW [ms] 0.66 0.05 0.68  0.68  0.68  0.68  GPW [ms] 2.43 0.24 2.67  2.94  3.50  4.40  Pilot Ignition Delay (ms) 1.55 0.13 1.49  1.61  1.58  1.67  Gas Ignition Delay (ms) 1.57 0.11 1.44  1.44  1.49  1.31  CA at peak cylinder pressure (bar) 10.98 0.41 10.78  10.37  9.78  8.53  Peak cylinder pressure (bar) 153.66 3.94 153.78  154.04  155.39  158.00  Peak AHRR ((kJ/m^3-deg)) 109.65 6.71 107.51  101.59  93.91  83.34  dP/dCA (bar/deg) 5.97 0.15 5.89  6.03  6.31  6.57  Combustion duration: IHR10 to IHR90 (deg) 40.67 1.43 41.50  42.00  43.00  45.00  Early cycle duration: IHR10 to IHR50 (deg) 12.54 0.43 13.31  13.57  14.98  18.08  Late cycle duration: IHR50 to IHR90 (deg) 28.13 1.01 28.19  28.43  28.02  26.92  IHR5 (deg ATDC) -4.83 0.72 -5.00  -6.00  -7.50  -10.50  IHR10 (deg ATDC) -2.83 0.72 -3.50  -4.00  -5.50  -8.00  IHR90 (deg ATDC) 37.83 0.72 38.00  38.00  37.50  37.00  System Efficiency: tested 3-stage (%) 43.21 0.10 42.88  42.73  42.69  42.67    155  Summary Table: FDR sweep at EGR=25% | Φ =0.8 nr 3 2 1 1 1 Parameter ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ FDR (%) 0 0.00 10.16  24.75  49.42  98.40  Gross Indicated Efficiency (%) 41.09 0.58 41.51  42.03  42.72  43.78  Combustion Efficiency (%) 93.57 0.71 94.76  95.67  97.06  98.39  GISFC - diesel equivalent (g/kW-hr) 204.82 2.88 202.75  200.27  197.02  192.26  NOx (g/kW-hr) 0.59 0.09 0.67  0.72  0.83  1.00  DustTrak (g/kW-hr) 0.40 NaN 0.28  0.18  0.09  0.02  CH4 (g/kW-hr) 1.53 0.37 1.05  0.82  0.47  0.23  tHC (g/kW-hr,C1) 1.85 0.32 1.35  1.06  0.61  0.30  CO (g/kW-hr) 31.37 5.96 26.83  22.14  15.34  8.32  CO2 (kg/kW-hr) 0.44 0.00 0.45  0.45  0.45  0.45  Cylinder exhaust temperature (°C) 516.12 4.94 520.31  521.99  524.36  522.56  Peak Mixture Averaged T (Celsius) 1374.10 24.03 1377.14  1391.19  1379.51  1382.50  EGR Temperature 94.96 4.28 96.32  97.77  97.89  97.98  Diesel injection mass (mg/inj) 8.87 0.68 9.02  8.95  9.11  9.27  Diesel flow (kg/hr) 0.36 0.03 0.36  0.36  0.37  0.37  CNG flow (kg/hr) 8.14 0.16 8.11  8.01  7.89  7.71  Intake surge tank pressure (kPag) 171.37 8.10 173.48  170.98  168.44  163.03  Air flow (kg/hr) 170.36 4.27 171.82  168.43  165.86  159.20  Engine speed (rpm) 1347.87 6.46 1348.93  1349.97  1349.81  1350.16  Gross IMEP (bar) 16.68 0.16 16.79  16.78  16.81  16.85  Carbon balance (corrected for Air Addition) 0.98 0.01 0.98  0.98  0.98  0.98  Inlet Oxygen mass fraction  0.18 0.00 0.19  0.19  0.18  0.18  PSOI set [deg] -27.08 0.95 -27.25  -28.00  -29.00  -32.00  PPW [ms] 0.65 0.03 0.66  0.66  0.66  0.66  GPW [ms] 2.50 0.20 2.74  3.10  3.63  4.50  Pilot Ignition Delay (ms) 1.56 0.11 1.61  1.54  1.67  1.85  Gas Ignition Delay (ms) 1.51 0.20 1.37  1.39  1.39  1.48  CA at peak cylinder pressure (bar) 10.90 0.21 10.76  10.10  9.61  8.31  Peak cylinder pressure (bar) 148.42 3.03 149.89  150.20  151.23  154.31  Peak AHRR ((kJ/m^3-deg)) 115.80 4.67 107.34  101.72  93.83  85.97  dP/dCA (bar/deg) 6.01 0.02 5.98  6.12  6.31  6.78  Combustion duration: IHR10 to IHR90 (deg) 41.67 0.72 42.50  43.50  45.00  47.50  Early cycle duration: IHR10 to IHR50 (deg) 12.45 0.60 13.10  13.89  15.42  18.35  Late cycle duration: IHR50 to IHR90 (deg) 29.21 0.50 29.40  29.61  29.58  29.15  IHR5 (deg ATDC) -4.50 0.00 -5.25  -6.00  -7.50  -10.50  IHR10 (deg ATDC) -2.67 0.72 -3.25  -4.00  -5.50  -8.50  IHR90 (deg ATDC) 39.00 0.00 39.25  39.50  39.50  39.00  System Efficiency: tested 3-stage (%) 41.11 0.58 41.20  41.26  41.20  40.76    156  Summary Table: FDR sweep at EGR=12.5% | Φ =0.6 nr 16 1 8 1 1 Parameter ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ FDR (%) 0 0.10 10.16  25.03 0.43 49.74  102.65  Gross Indicated Efficiency (%) 44.65 0.20 44.77  45.70 0.18 46.11  46.88  Combustion Efficiency (%) 98.74 0.04 98.95  99.20 0.03 99.51  99.78  GISFC - diesel equivalent (g/kW-hr) 188.52 0.84 187.99  184.17 0.71 182.53  179.55  NOx (g/kW-hr) 2.62 0.07 2.80  3.00 0.07 3.41  4.81  DustTrak (g/kW-hr) 0.06 0.00 0.04  0.02 0.00 0.01  0.00  CH4 (g/kW-hr) 0.34 0.02 0.33  0.27 0.01 0.24  0.17  tHC (g/kW-hr,C1) 0.43 0.03 0.41  0.35 0.02 0.31  0.23  CO (g/kW-hr) 6.54 0.35 5.21  3.73 0.26 1.82  0.37  CO2 (kg/kW-hr) 0.45 0.00 0.45  0.45 0.00 0.44  0.44  Cylinder exhaust temperature (°C) 490.24 2.28 489.01  478.03 5.41 471.22  463.98  Peak Mixture Averaged T (Celsius) 1356.60 5.39 1373.62  1373.04 8.51 1362.69  1382.59  EGR Temperature 81.67 0.97 83.50  82.18 1.62 83.47  82.01  Diesel injection mass (mg/inj) 8.65 0.22 8.86  8.98 0.23 8.83  9.12  Diesel flow (kg/hr) 0.35 0.01 0.36  0.36 0.01 0.36  0.37  CNG flow (kg/hr) 7.50 0.03 7.49  7.32 0.05 7.29  7.23  Intake surge tank pressure (kPag) 185.43 1.23 184.39  181.79 2.38 184.78  177.15  Air flow (kg/hr) 211.21 0.93 208.73  207.10 1.76 207.85  205.70  Engine speed (rpm) 1344.43 1.88 1350.92  1345.67 2.28 1352.19  1354.29  Gross IMEP (bar) 16.75 0.03 16.71  16.77 0.05 16.76  16.89  Carbon balance (corrected for Air Addition) 1.00 0.00 0.99  0.99 0.01 0.99  1.00  Inlet Oxygen mass fraction  0.21 0.00 0.21  0.21 0.00 0.21  0.22  PSOI set [deg] -25.36 0.54 -25.75  -25.84 0.25 -27.00  -30.50  PPW [ms] 0.64 0.03 0.66  0.66 0.00 0.66  0.66  GPW [ms] 2.22 0.07 2.40  2.65 0.06 3.22  4.23  Pilot Ignition Delay (ms) 1.38 0.03 1.33  1.43 0.01 1.48  1.54  Gas Ignition Delay (ms) 1.29 0.04 1.24  1.20 0.02 1.20  1.17  CA at peak cylinder pressure (bar) 11.77 0.11 11.47  11.11 0.14 10.32  8.61  Peak cylinder pressure (bar) 151.40 0.82 152.67  152.09 1.18 152.73  157.28  Peak AHRR ((kJ/m^3-deg)) 114.83 1.66 110.50  108.69 1.88 98.13  84.13  dP/dCA (bar/deg) 5.61 0.10 5.75  5.75 0.10 5.75  6.23  Combustion duration: IHR10 to IHR90 (deg) 38.34 0.54 38.00  37.94 0.47 38.50  40.50  Early cycle duration: IHR10 to IHR50 (deg) 12.46 0.24 12.92  13.21 0.13 14.46  17.61  Late cycle duration: IHR50 to IHR90 (deg) 25.88 0.32 25.08  24.73 0.39 24.04  22.89  IHR5 (deg ATDC) -4.53 0.31 -5.00  -5.44 0.27 -6.50  -10.50  IHR10 (deg ATDC) -2.66 0.29 -3.50  -3.63 0.19 -4.50  -8.00  IHR90 (deg ATDC) 35.69 0.34 34.50  34.31 0.50 34.00  32.50  System Efficiency: tested 3-stage (%) 44.66 0.20 44.46  44.93 0.17 44.58  43.73    157  Summary Table: FDR sweep at EGR=12.5% | Φ =0.7 nr 5 4 2 4 1 Parameter ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ FDR (%)   0 0.23 10.72 0.36 25.44  49.53 2.40 100.84  Gross Indicated Efficiency (%) 42.98 0.24 43.49 0.23 44.04  44.77 0.54 45.42  Combustion Efficiency (%) 98.07 0.23 98.50 0.24 98.79  99.14 0.16 99.71  GISFC - diesel equivalent (g/kW-hr) 195.83 1.08 193.55 1.05 191.12  188.00 2.27 185.31  NOx (g/kW-hr) 2.04 0.13 2.29 0.19 2.62  2.88 0.23 3.78  DustTrak (g/kW-hr) 0.12 0.02 0.06 0.02 0.03  0.01 0.01 0.00  CH4 (g/kW-hr) 0.29 0.03 0.24 0.04 0.22  0.19 0.01 0.13  tHC (g/kW-hr,C1) 0.36 0.05 0.30 0.06 0.28  0.24 0.01 0.19  CO (g/kW-hr) 11.83 1.41 9.11 1.53 7.17  4.80 1.14 1.08  CO2 (kg/kW-hr) 0.46 0.00 0.46 0.00 0.46  0.45 0.00 0.45  Cylinder exhaust temperature (°C) 533.57 5.85 531.07 8.21 525.24  518.73 7.87 490.62  Peak Mixture Averaged T (Celsius) 1436.68 12.14 1441.53 12.51 1457.42  1454.34 9.65 1463.50  EGR Temperature 82.98 2.09 83.77 0.65 83.53  83.81 0.87 79.48  Diesel injection mass (mg/inj) 9.03 0.39 9.02 0.09 8.99  9.01 0.25 10.24  Diesel flow (kg/hr) 0.37 0.02 0.37 0.00 0.36  0.36 0.01 0.41  CNG flow (kg/hr) 7.83 0.07 7.75 0.09 7.63  7.53 0.13 7.31  Intake surge tank pressure (kPag) 163.10 2.44 161.74 3.84 157.86  156.06 4.26 150.38  Air flow (kg/hr) 191.57 1.83 190.88 2.46 188.11  185.25 1.69 181.92  Engine speed (rpm) 1350.53 4.59 1352.97 6.86 1352.40  1351.75 6.85 1347.37  Gross IMEP (bar) 16.77 0.09 16.77 0.07 16.75  16.80 0.17 16.72  Carbon balance (corrected for Air Addition) 0.99 0.01 0.99 0.01 0.99  0.99 0.02 0.99  Inlet Oxygen mass fraction  0.21 0.00 0.21 0.00 0.21  0.21 0.00 0.21  PSOI set [deg] -26.05 0.86 -26.06 0.38 -26.50  -27.31 0.38 -29.50  PPW [ms] 0.66 0.01 0.66 0.00 0.66  0.66 0.00 0.66  GPW [ms] 2.28 0.21 2.45 0.12 2.73  3.24 0.12 4.08  Pilot Ignition Delay (ms) 1.45 0.07 1.42 0.07 1.39  1.46 0.08 1.61  Gas Ignition Delay (ms) 1.32 0.07 1.30 0.02 1.30  1.25 0.11 1.14  CA at peak cylinder pressure (bar) 11.58 0.21 11.56 0.19 11.12  10.44 0.24 8.86  Peak cylinder pressure (bar) 145.82 0.65 144.66 1.64 144.49  144.97 1.91 147.17  Peak AHRR ((kJ/m^3-deg)) 113.61 3.95 111.23 2.13 107.83  99.92 2.25 88.76  dP/dCA (bar/deg) 5.71 0.25 5.56 0.20 5.71  5.73 0.18 5.95  Combustion duration: IHR10 to IHR90 (deg) 40.40 0.81 39.88 1.00 39.75  40.38 0.76 40.50  Early cycle duration: IHR10 to IHR50 (deg) 12.61 0.32 12.86 0.51 13.36  14.64 0.33 16.89  Late cycle duration: IHR50 to IHR90 (deg) 27.79 0.50 27.02 0.76 26.39  25.73 0.50 23.61  IHR5 (deg ATDC) -4.80 0.56 -5.13 0.40 -5.75  -6.88 0.40 -10.00  IHR10 (deg ATDC) -3.00 0.44 -3.13 0.40 -3.75  -4.88 0.40 -7.50  IHR90 (deg ATDC) 37.40 0.52 36.75 0.80 36.00  35.50 0.65 33.00  System Efficiency: tested 3-stage (%) 42.99 0.24 43.16 0.24 43.26  43.25 0.56 42.34    158  Summary Table: FDR sweep at EGR=12.5% | Φ =0.8 nr 5 3 3 3 1 Parameter ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ FDR (%) 0 0.43 10.04 0.37 24.36 1.52 50.02 2.85 98.41  Gross Indicated Efficiency (%) 40.84 0.60 41.25 0.40 41.71 0.12 43.00 1.09 44.17  Combustion Efficiency (%) 97.93 0.30 98.19 0.24 98.22 0.09 98.60 0.24 99.45  GISFC - diesel equivalent (g/kW-hr) 206.13 3.04 204.04 1.98 201.81 0.58 195.75 4.92 190.55  NOx (g/kW-hr) 1.68 0.07 1.79 0.07 1.96 0.16 2.38 0.19 2.84  DustTrak (g/kW-hr) 0.18 0.02 0.14 0.03 0.09 0.02 0.03 0.01 0.00  CH4 (g/kW-hr) 0.13 0.04 0.10 0.04 0.10 0.01 0.09 0.04 0.06  tHC (g/kW-hr,C1) 0.17 0.05 0.12 0.05 0.12 0.00 0.12 0.06 0.09  CO (g/kW-hr) 14.38 2.08 12.64 1.21 12.47 0.75 9.54 2.19 3.47  CO2 (kg/kW-hr) 0.48 0.01 0.48 0.01 0.48 0.01 0.47 0.02 0.46  Cylinder exhaust temperature (°C) 563.94 5.26 565.77 9.93 566.73 1.64 558.12 10.82 541.05  Peak Mixture Averaged T (Celsius) 1473.43 26.55 1489.84 32.49 1501.75 20.71 1508.95 20.82 1502.45  EGR Temperature 82.91 2.65 84.61 1.65 84.50 1.99 82.93 4.54 81.09  Diesel injection mass (mg/inj) 9.06 0.30 9.25 0.30 9.37 0.48 9.09 0.51 9.14  Diesel flow (kg/hr) 0.37 0.01 0.37 0.01 0.38 0.02 0.37 0.02 0.37  CNG flow (kg/hr) 8.22 0.12 8.16 0.15 8.11 0.04 7.84 0.12 7.58  Intake surge tank pressure (kPag) 146.35 2.94 144.39 1.08 142.05 2.65 136.42 8.57 129.23  Air flow (kg/hr) 179.19 2.16 177.10 0.96 174.68 1.22 170.89 4.45 165.83  Engine speed (rpm) 1349.84 6.61 1353.75 8.90 1353.14 10.45 1352.01 16.43 1347.26  Gross IMEP (bar) 16.73 0.12 16.73 0.27 16.83 0.12 16.80 0.13 16.76  Carbon balance (corrected for Air Addition) 0.99 0.01 1.00 0.02 1.00 0.02 0.99 0.02 0.99  Inlet Oxygen mass fraction  0.21 0.00 0.21 0.00 0.21 0.00 0.21 0.00 0.21  PSOI set [deg] -25.55 0.77 -26.50 0.00 -27.17 0.72 -27.50 1.86 -29.25  PPW [ms] 0.62 0.07 0.66 0.00 0.66 0.00 0.62 0.16 0.55  GPW [ms] 2.32 0.02 2.59 0.03 2.92 0.15 3.36 0.27 4.18  Pilot Ignition Delay (ms) 1.45 0.06 1.42 0.01 1.54 0.01 1.54 0.07 1.64  Gas Ignition Delay (ms) 1.34 0.09 1.27 0.22 1.31 0.21 1.27 0.14 1.27  CA at peak cylinder pressure (bar) 11.62 0.13 11.24 0.69 10.79 0.54 10.43 0.81 9.14  Peak cylinder pressure (bar) 139.69 0.68 139.94 0.57 140.62 3.89 139.34 4.25 139.08  Peak AHRR ((kJ/m^3-deg)) 113.76 3.13 109.06 0.94 105.71 2.15 100.01 6.11 87.08  dP/dCA (bar/deg) 5.73 0.10 5.78 0.21 5.84 0.38 5.74 0.17 5.87  Combustion duration: IHR10 to IHR90 (deg) 41.30 0.71 41.83 0.72 42.67 0.72 42.33 0.72 44.50  Early cycle duration: IHR10 to IHR50 (deg) 12.52 0.41 13.26 0.35 13.86 0.23 14.76 0.73 17.59  Late cycle duration: IHR50 to IHR90 (deg) 28.78 0.40 28.57 0.40 28.81 0.55 27.57 0.53 26.91  IHR5 (deg ATDC) -4.60 0.28 -5.50 0.00 -6.17 0.72 -7.33 1.43 -10.00  IHR10 (deg ATDC) -2.70 0.34 -3.50 0.00 -4.17 0.72 -5.00 1.24 -7.50  IHR90 (deg ATDC) 38.60 0.68 38.33 0.72 38.50 0.00 37.33 0.72 37.00  System Efficiency: tested 3-stage (%) 40.84 0.60 40.94 0.41 40.96 0.07 41.46 1.05 41.15    159  Summary Table: FDR sweep at EGR=0% | Φ =0.6 nr 6 1 1 2 1 Parameter ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ FDR (%) 0 0.19 10.01  25.41  50.57  99.62  Gross Indicated Efficiency (%) 44.62 0.31 44.56  45.22  46.21  46.68  Combustion Efficiency (%) 99.34 0.05 99.44  99.56  99.74  99.86  GISFC - diesel equivalent (g/kW-hr) 188.62 1.33 188.87  186.15  182.13  180.32  NOx (g/kW-hr) 5.29 0.15 5.48  6.11  6.47  7.50  DustTrak (g/kW-hr) 0.03 0.00 0.02  0.01  0.00  0.00  CH4 (g/kW-hr) 0.18 0.01 0.17  0.16  0.15  0.13  tHC (g/kW-hr,C1) 0.25 0.02 0.23  0.22  0.21  0.19  CO (g/kW-hr) 4.05 0.42 3.38  2.39  0.99  0.18  CO2 (kg/kW-hr) 0.46 0.01 0.46  0.46  0.45  0.44  Cylinder exhaust temperature (°C) 526.05 11.39 530.61  517.39  500.97  478.36  Peak Mixture Averaged T (Celsius) 1429.71 9.30 1447.44  1444.71  1455.40  1459.63  EGR Temperature 56.33 6.29 53.15  50.73  51.34  51.27  Diesel injection mass (mg/inj) 8.84 0.20 8.94  9.04  8.88  9.26  Diesel flow (kg/hr) 0.36 0.01 0.36  0.36  0.36  0.37  CNG flow (kg/hr) 7.50 0.10 7.53  7.37  7.25  7.14  Intake surge tank pressure (kPag) 167.53 1.66 164.17  162.76  159.32  156.85  Air flow (kg/hr) 224.52 1.24 220.79  219.71  216.73  213.94  Engine speed (rpm) 1346.05 2.66 1344.75  1344.67  1346.41  1348.00  Gross IMEP (bar) 16.75 0.07 16.82  16.70  16.77  16.72  Carbon balance (corrected for Air Addition) 1.00 0.01 1.00  1.00  1.00  0.99  Inlet Oxygen mass fraction  0.23 0.00 0.23  0.23  0.23  0.23  PSOI set [deg] -24.75 0.64 -24.75  -25.50  -25.88  -28.00  PPW [ms] 0.66 0.00 0.66  0.66  0.66  0.66  GPW [ms] 2.16 0.16 2.30  2.60  3.05  3.91  Pilot Ignition Delay (ms) 1.35 0.03 1.33  1.36  1.35  1.48  Gas Ignition Delay (ms) 1.20 0.05 1.12  1.15  1.10  0.93  CA at peak cylinder pressure (bar) 12.19 0.18 11.91  11.31  10.78  9.44  Peak cylinder pressure (bar) 145.65 1.21 146.00  146.25  145.04  147.04  Peak AHRR ((kJ/m^3-deg)) 118.86 3.01 116.91  112.04  105.03  88.51  dP/dCA (bar/deg) 5.23 0.21 5.27  5.24  5.16  5.70  Combustion duration: IHR10 to IHR90 (deg) 37.75 0.92 37.50  37.50  36.50  36.50  Early cycle duration: IHR10 to IHR50 (deg) 12.19 0.41 12.65  12.96  13.53  16.25  Late cycle duration : IHR50 to IHR90 (deg) 25.56 0.52 24.85  24.54  22.97  20.25  IHR5 (deg ATDC) -4.42 0.40 -5.00  -5.50  -6.25  -9.00  IHR10 (deg ATDC) -2.33 0.43 -3.00  -3.50  -3.75  -6.50  IHR90 (deg ATDC) 35.42 0.70 34.50  34.00  32.75  30.00  System Efficiency: tested 3-stage (%) 44.63 0.32 44.26  44.44  44.66  43.62    160  Summary Table: FDR sweep at EGR=0% | Φ =0.7 nr 4 1 1 2 2 Parameter ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ FDR (%) 0 0.49 10.44  24.13  50.51  100.71  Gross Indicated Efficiency (%) 42.49 0.41 42.91  43.66  44.61  45.04  Combustion Efficiency (%) 99.17 0.10 99.30  99.38  99.62  99.79  GISFC - diesel equivalent (g/kW-hr) 198.08 1.91 196.16  192.80  188.70  186.89  NOx (g/kW-hr) 4.64 0.26 5.06  5.49  6.02  6.64  DustTrak (g/kW-hr) 0.05 0.01 0.02  0.01  0.00  0.00  CH4 (g/kW-hr) 0.08 0.02 0.07  0.08  0.09  0.08  tHC (g/kW-hr,C1) 0.10 0.03 0.10  0.11  0.12  0.12  CO (g/kW-hr) 6.32 0.84 5.31  4.49  2.50  1.14  CO2 (kg/kW-hr) 0.48 0.01 0.48  0.47  0.46  0.46  Cylinder exhaust temperature (°C) 589.02 15.56 588.43  572.29  551.49  531.49  Peak Mixture Averaged T (Celsius) 1520.90 9.11 1546.97  1549.65  1563.20  1554.50  EGR Temperature 54.17 10.63 52.20  52.50  50.83  48.78  Diesel injection mass (mg/inj) 9.19 0.38 9.02  8.96  9.18  9.62  Diesel flow (kg/hr) 0.37 0.01 0.36  0.36  0.37  0.39  CNG flow (kg/hr) 7.91 0.09 7.83  7.73  7.52  7.44  Intake surge tank pressure (kPag) 143.02 2.00 137.50  137.47  133.72  132.14  Air flow (kg/hr) 202.01 0.82 197.06  196.60  192.97  190.34  Engine speed (rpm) 1350.17 4.17 1348.01  1347.45  1348.45  1349.86  Gross IMEP (bar) 16.77 0.04 16.77  16.87  16.78  16.79  Carbon balance (corrected for Air Addition) 1.00 0.01 1.01  1.01  1.00  0.99  Inlet Oxygen mass fraction  0.23 0.00 0.23  0.23  0.23  0.23  PSOI set [deg] -25.38 1.19 -25.25  -25.75  -26.50  -29.50  PPW [ms] 0.66 0.01 0.66  0.66  0.66  0.66  GPW [ms] 2.32 0.31 2.40  2.67  3.15  4.09  Pilot Ignition Delay (ms) 1.34 0.09 1.39  1.39  1.42  1.48  Gas Ignition Delay (ms) 1.22 0.05 1.29  1.27  1.14  1.1  CA at peak cylinder pressure (bar) 11.94 0.31 11.93  11.50  10.95  9.14  Peak cylinder pressure (bar) 139.27 0.84 137.42  138.76  138.24  142.01  Peak AHRR ((kJ/m^3-deg)) 118.82 6.35 114.50  112.79  103.83  90.41  dP/dCA (bar/deg) 5.21 0.10 5.25  5.26  5.19  5.98  Combustion duration: IHR10 to IHR90 (deg) 40.13 1.51 39.50  39.00  38.75  40.75  Early cycle duration: IHR10 to IHR50 (deg) 12.47 0.95 12.72  13.06  14.12  16.93  Late cycle duration: IHR50 to IHR90 (deg) 27.65 0.59 26.78  25.94  24.63  23.82  IHR5 (deg ATDC) -4.75 0.80 -5.00  -6.00  -6.75  -10.00  IHR10 (deg ATDC) -2.75 0.80 -3.00  -3.50  -4.50  -7.50  IHR90 (deg ATDC) 37.38 0.76 36.50  35.50  34.25  33.25  System Efficiency: tested 3-stage (%) 42.50 0.42 42.59  42.91  43.05  41.95    161  Summary Table: FDR sweep at EGR=0% | Φ =0.8 nr 2 1 2 1 2 Parameter ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ ?̅? ∆ FDR (%) 0  10.37  24.96  50.63  98.90  Gross Indicated Efficiency (%) 40.83  41.38  41.74  42.66  43.58  Combustion Efficiency (%) 99.45  99.44  99.37  99.43  99.67  GISFC - diesel equivalent (g/kW-hr) 206.15  203.41  201.63  197.30  193.14  NOx (g/kW-hr) 3.90  4.27  4.40  4.92  5.82  DustTrak (g/kW-hr) 0.06  0.03  0.02  0.01  0.00  CH4 (g/kW-hr) 0.01  0.01  0.01  0.01  0.03  tHC (g/kW-hr,C1) 0.02  0.02  0.02  0.02  0.05  CO (g/kW-hr) 4.60  4.67  5.26  4.70  2.50  CO2 (kg/kW-hr) 0.50  0.49  0.49  0.48  0.47  Cylinder exhaust temperature (°C) 644.62  632.98  620.82  607.28  579.59  Peak Mixture Averaged T (Celsius) 1606.88  1621.52  1621.03  1641.55  1641.53  EGR Temperature 53.68  52.04  49.67  51.13  50.85  Diesel injection mass (mg/inj) 9.44  9.46  9.42  9.47  9.78  Diesel flow (kg/hr) 0.38  0.38  0.38  0.38  0.40  CNG flow (kg/hr) 8.23  8.14  8.05  7.88  7.70  Intake surge tank pressure (kPag) 122.04  119.81  118.36  114.04  112.47  Air flow (kg/hr) 182.98  180.84  179.44  174.68  172.47  Engine speed (rpm) 1350.79  1352.11  1351.66  1352.08  1351.46  Gross IMEP (bar) 16.76  16.79  16.76  16.77  16.78  Carbon balance (corrected for Air Addition) 0.99  0.99  0.99  0.99  0.99  Inlet Oxygen mass fraction  0.23  0.23  0.23  0.23  0.23  PSOI set [deg] -25.38  -26.00  -26.00  -27.00  -29.50  PPW [ms] 0.66  0.66  0.66  0.66  0.66  GPW [ms] 2.32  2.56  2.81  3.30  4.14  Pilot Ignition Delay (ms) 1.37  1.36  1.39  1.48  1.54  Gas Ignition Delay (ms) 1.25  1.20  1.20  1.29  1.14  CA at peak cylinder pressure (bar) 11.96  11.60  11.37  10.69  9.34  Peak cylinder pressure (bar) 132.82  133.54  132.27  131.79  134.47  Peak AHRR ((kJ/m^3-deg)) 117.65  115.74  112.80  103.88  91.30  dP/dCA (bar/deg) 5.38  5.32  5.31  5.28  5.84  Combustion duration: IHR10 to IHR90 (deg) 41.50  41.00  40.75  41.00  42.25  Early cycle duration: IHR10 to IHR50 (deg) 12.76  12.90  13.24  14.38  17.13  Late cycle duration: IHR50 to IHR90 (deg) 28.74  28.10  27.51  26.62  25.12  IHR5 (deg ATDC) -4.75  -5.50  -6.00  -7.00  -10.00  IHR10 (deg ATDC) -3.00  -3.50  -3.50  -4.50  -7.25  IHR90 (deg ATDC) 38.50  37.50  37.25  36.50  35.00  System Efficiency: tested 3-stage (%) 40.84  41.06  40.98  41.10  40.55    162  Appendix C System Level Study of Air Addition: Additional Information C.1  Response Surface Relevant Information 𝑅𝑒𝑠𝑖𝑑𝑢𝑎𝑙 % = (𝑀𝑒𝑎𝑠𝑢𝑟𝑒𝑑 𝑉𝑎𝑙𝑢𝑒 − 𝑅𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑆𝑢𝑟𝑓𝑎𝑐𝑒 𝑉𝑎𝑙𝑢𝑒𝑀𝑒𝑎𝑠𝑢𝑟𝑒𝑑 𝑉𝑎𝑙𝑢𝑒) . 100 𝑅𝑒𝑠𝑖𝑑𝑢𝑎𝑙𝑠 = (𝑀𝑒𝑎𝑠𝑢𝑟𝑒𝑑 𝑉𝑎𝑙𝑢𝑒 − 𝑅𝑒𝑠𝑝𝑜𝑛𝑠𝑒 𝑆𝑢𝑟𝑓𝑎𝑐𝑒 𝑉𝑎𝑙𝑢𝑒)                          Standard error (𝜎𝑠𝑒) formula is Equation (3.15)  in Section 3.3  Table 7.6 Standard errors of various response surfaces Note: The standard errors were determined based on Equation (3.15) Response surface Standard error (𝝈𝒔𝒆) NOx (g/kWhr) 0.095 g/kWhr PM (g/kWhr)  0.014 g/kWhr Gross indicated efficiency (%) 0.34%-point Intake surge tank pressure (bar) 0.036 bar Compressor motor power (kW) 0.047 kW Compressor outlet mass flowrate (kg/hr) 0.094 kg/hr          163      Figure 7.7 Accuracy of response surface fit of NOx  y = 0.997x + 0.0088R² = 0.9970123456780 2 4 6 8NOx: Response Surface (g/kWhr)Measured NOx (g/kWhr)164     Figure 7.8 Accuracy of response surface fit of PM        y = 1.0104x - 0.0023R² = 0.9718-0.100.10.20.30.40.50 0.1 0.2 0.3 0.4 0.5PM: Response Surface (g/kWhr)Measured PM (g/kWhr)165     Figure 7.9 Accuracy of response surface fit of gross indicated efficiency       y = 0.9567x + 1.8947R² = 0.95673940414243444546474838 40 42 44 46 48Gross Indicated Efficiency:  Response Surface (%)Measured Gross Indicated Efficiency (%)166     Figure 7.10 Accuracy of response surface fit of intake surge tank pressure     y = 0.9804x + 0.052R² = 0.980422.22.42.62.833.23.42 2.5 3 3.5Intake Surge Tank Pressure:  Response Surface (bar)Measured Intake Surge Tank Pressure (bar)167     Figure 7.11 Accuracy of fit of response surface fit of compressor motor power     y = 0.9796x + 0.062R² = 0.97962.52.72.93.13.33.53.73.92.5 3 3.5 4Compressor Motor Power:  Response Surface (kW)Measured Compressor Motor Power (kW)168     Figure 7.12 Accuracy of response surface fit of compressor outlet mass flowrate  C.2 ANOVA Summary Table  p-value Parameters Intercept EGR Φ FDR EGR.Φ Φ.FDR EGR.FDR 𝜂𝑠𝑦𝑠 tested 3-stage 0.00 0.84 0.00 0.00 0.88 0.00 0.76 𝜂𝑠𝑦𝑠 hypothetical 3-stage  0.00 0.84 0.00 0.00 0.88 0.00 0.76 𝜂𝑠𝑦𝑠 tested 2-stage  0.00 0.32 0.00 0.10 0.20 0.00 0.00 𝜂𝑠𝑦𝑠 hypothetical 2-stage  0.00 0.86 0.00 0.01 0.86 0.01 0.88 y = 0.998x + 0.0073R² = 0.99801234567890 2 4 6 8 10Compressor Outlet Mass Flowrate:  Response Surface (kg/hr)Measured Compressor Outlet Mass Flowrate (kW)169  Appendix D P&ID  170   SCRE Facility 171  Experimental Facility: Additional Information Coolant System The facility had a closed loop heat exchanger system with glycol and water blend (1:1) as the coolant. With the engine not in operation, a heater and an external pump in the coolant system circulated warm coolant to keep the engine block warm (50oC) and therefore, facilitate easy startup by reducing oil viscosity. With the engine in operation, the coolant system dissipated excess heat from the engine block and EGR cooler to the city water, through the heat exchanger. Internal thermostats of the engine were controlled from the operator’s computer and were used to keep the coolant temperature at engine block outlet to 80±2o C.   Gravimetric diesel measurement system The measurement of pilot fuel- diesel is yet another parameter essential to defining and setting an operating point in an HPDI engine, where the energy contribution of pilot fuel is typically less than 10% of the total energy content. For the measurement of diesel, a closed loop gravimetric system was used for the measurement of pilot fuel-diesel flowrate. The diesel was drawn from a gravimetric pail and successively pumped at high pressure to the injector. Simultaneously, unused diesel from the injector and the high-pressure pump was returned continuously to the gravimetric pail. This constituted the closed loop system and a regression analysis over the rate of change of diesel mass in the gravimetric pail for more than 3 minutes, was therefore, taken as the diesel flowrate measurement. The specifications of the gravimetric pail as well as relevant uncertainty analysis are presented in Appendix A.   172  Engine Intake Flow measurement Engine fresh intake air flow measurement is critical to determine operating parameters like the EGR and global equivalence ratio. The accuracy of the measurement is also critical to determine unaccounted leaks through carbon balance. The measurement is carried is via a venturi flowmeter system. The system constituted of a UBC machined subsonic venturi flowmeters, a gauge pressure transducer, am ungrounded K-type thermocouple, and a differential pressure transducer (Omega PX2300-2DI). The gauge pressure transducer measured the absolute upstream pressure assuming atmospheric pressure of 1 bar. The thermocouple measured fluid temperature upstream and together with the absolute upstream pressure, determined the fluid density upstream. The differential pressure transducer in conjunction with the absolute upstream determined the pressure at the throat of the venturi and thereby the ratio of pressure at the restriction to the inlet of venturi. With these measurements, a known area of venturi cross sections (at the inlet and at the throat), and a Cv determined from calibration the air flow measurement was determined with a standard venturi equation.    Diluent (Air) Compression System: Additional Details of Data Acquisition System However, the LabVIEW 6.0 user interface for SCRE was slightly modified to add a separate page for ‘Compressor Characterization’ which allowed the monitoring of the compressor operation stability through the plots of pressure, temperature and flow. Once the stage outlet temperatures reached steady states (defined as rate of temperature change < 1oC per minute), ‘slow’ data was recorded for a duration of 180s.   

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

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"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0376211/manifest

Comment

Related Items