"Applied Science, Faculty of"@en . "Mechanical Engineering, Department of"@en . "DSpace"@en . "UBCV"@en . "St Hill, Simon"@en . "2008-09-23T17:17:29Z"@en . "1993"@en . "Master of Applied Science - MASc"@en . "University of British Columbia"@en . "Due to increasing fuel cost and emphasis on energy conservation as well as pollution control, there has been considerable interest in improving propulsive efficiency of road vehicles. Reduction in aerodynamic resistance is one aspect of it. Although aerodynamically contoured automobiles has become a standard design practice. Trucks have changed little over the past three decades.\r\nThe thesis presents results of an organized and extensive wind tunnel test-program, complemented by full-scale road tests, aimed at assessing the effectiveness of two boundary-layer control procedures for reduction of the pressure drag of a cube-van. Wind tunnel results, obtained using 1/6th scale models, at a subcritical Reynolds number of 105, suggest that both the Moving Surface Boundary-layer Control (MSBC) as well as the tripping of the boundary-layer using fences reduce the pressure drag coefficient. Although both the concepts are promising, application of the entirely passive fence procedure appears more attractive from an economic consideration as well as the ease of implementation.\r\nThe road tests with a full-size cube-van substantiated the trends indicated by the fence data although the actual drag reduction observed was lower (yet quite significant, = 16.6%) than that predicted by the wind tunnel tests. This may be attributed to a wide variety of factors including the differences in the geometry (models; fences and their orientation),operating conditions (Reynolds number; yaw; wind variations in magnitude and direction; turbulence; road boundary-layer; road surface condition), and measurement errors. However, the objective of the study was to assess potential of the concepts which, indeed, is quite promising.\r\nFuel consumption results also substantiated the drag reduction trend. As expected they depend on the gearing condition and hence no general expression applicable to all speed ranges is available. As anticipated the data show a rapid increase in the fuel consumption efficiency at the top end of the speed range.\r\nIt is concluded that fences can lead to a significant improvement in drag reduction and fuel consumption when applied to flat-faced trucks if positioned correctly. They represent a more elegant, versatile, and cheaper alternative to the 'nose cones' and deflectors available in the market. It is recommended that further road tests should be conducted using both boundary-layer control devices."@en . "https://circle.library.ubc.ca/rest/handle/2429/2358?expand=metadata"@en . "9595369 bytes"@en . "application/pdf"@en . "We accept this thesis as conforming to the required standardDRAG REDUCTION OF CUBE-VAN THROUGH BOUNDARY-LAYERCONTROL : WIND TUNNEL EXPERIMENTS AND PROTOTYPE ROADTESTS.bySimon St HillB. Eng, Royal Melbourne Institute of Technology, 1988A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THEREQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIEDSCIENCEinTHE FACULTY OF GRADUATE STUDIESDepartment of Mechanical EngineeringTHE UNIVERSITY OF BRITISH COLUMBIADecember, 1992\u00C2\u00A9 Simon St HillIn presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)Department of /i, / ---(1.1-71-A)/Cf)g ^,/k)The University of British ColumbiaVancouver, CanadaDate ^\u00E2\u0080\u0094^- 1992-DE-6 (2/88)ABSTRACTDue to increasing fuel cost and emphasis on energy conservation aswell as pollution control, there has been considerable interest inimproving propulsive efficiency of road vehicles. Reduction inaerodynamic resistance is one aspect of it. Although aerodynamicallycontoured automobiles has become a standard design practice. Truckshave changed little over the past three decades.The thesis presents results of an organised and extensive windtunnel test-program, complemented by full-scale road tests, aimed atassessing the effectiveness of two boundary-layer control procedures forreduction of the pressure drag of a cube-van. Wind tunnel results,obtained using 1/6th scale models, at a subcritical Reynolds number of 10 5 ,suggest that both the Moving Surface Boundary-layer Control (MSBC) aswell as the tripping of the boundary-layer using fences reduce thepressure drag coefficient. Although both the concepts are promising,application of the entirely passive fence procedure appears more attractivefrom an economic consideration as well as the ease of implementation.The road tests with a full-size cube-van substantiated the trendsindicated by the fence data although the actual drag reduction observedwas lower (yet quite significant, = 16.6%) than that predicted by the windtunnel tests. This may be attributed to a wide variety of factors includingthe differences in the geometry (models; fences and their orientation),operating conditions (Reynolds number; yaw; wind variations inmagnitude and direction; turbulence; road boundary-layer; road surfacecondition), and measurement errors. However, the objective of the studywas to assess potential of the concepts which, indeed, is quite promising.11Fuel consumption results also substantiated the drag reductiontrend. As expected they depend on the gearing condition and hence nogeneral expression applicable to all speed ranges is available. Asanticipated the data show a rapid increase in the fuel consumptionefficiency at the top end of the speed range.It is concluded that fences can lead to a significant improvement indrag reduction and fuel consumption when applied to flat-faced trucks ifpositioned correctly. They represent a more elegant, versatile, andcheaper alternative to the 'nose cones' and deflectors available in themarket.It is recommended that further road tests should be conducted usingboth boundary-layer control devices.111TABLE OF CONTENTSABSTRACT^ iiTABLE OF CONTENTS^ ivLIST OF FIGURES viLIST OF TABLESNOMENCLATURE^ xiACKNOWLEDGEMENT xii1^INTRODUCTION^ 11.1 Background 11.2 Theory of Fences^ 21.3 A Brief Review of the Relevant Literature^41.4 Scope of the Present Investigation^ 142 TEST PROCEDURE^ 172.1 Wind Tunnel Test Procedure^ 172.1.1 Model specifications 172.1.2 Wind tunnel^ 182.1.3 Model support system and instrumentation^182.2 Full Scale Test Procedures^ 222.2.1 Truck specifications 222.2.2 Torque and speed measurements^302.2.3 Fuel flow measurement^ 332.2.4 On-road test procedure 34iv3^RESULTS AND DISCUSSION^ 453.1 Wind Tunnel Model tests 453.1.1 Wind tunnel models with fences^ 453.1.2 Wind tunnel model with a rotating cylinder and fences513.2 Full Scale Tests^ 563.3 Comparison Between Model and Full Scale Test^644 CONCLUDING REMARKS^ 694.1 Summary of Results 694.1.1 Wind tunnel model tests^ 694.1.2 Full scale tests^ 704.2 Recommendation for future study^ 71REFERENCES^ 73APPENDIX A : PROGRAM LISTINGS^ 78APPENDIX B : FORCE REGRESSION PLOTS^ 88APPENDIX C : FUEL CONSUMPTION PLOTS 113viLIST OF FIGURESFigure1-1 A schematic diagram showing the effect of fences on the front face ofa bluff body, on pressure distribution: (a) without fences; (b) withfeces 31-2 The practical application of moving wall for boundary layer control.101-3 Schematic diagrams explaining principles of the MSBC and theboundary layer trip devices in reducing drag on bluff bodies^132-1 Photograph of the 1/6 scale cube-van model^ 172-2 University of British Columbia boundary-layer wind tunnel^192-3 Model Arrangement in the Wind Tunnel^ 212-4 The truck as tested with an increased frontal area. This wasachieved by enlarging the box through raising the roof. Theextension was in a removable modular form. This is referred to asthe 'extended roof case. 232-5 The extended roof truck with a horizontal fence.^242-6 The Truck with the extended roof, and all fences^252-7 A block diagram of the test set up.^ 282-8 An example of the computer display. 29vii2-9 Equipment layout in the truck.^ 302-10 Torquemeter as installed in the truck.^ 312-11 Force calibration method^322-12 Photograph of the fuel meter installed on the truck.^342-13 Test route map (n.t.s.).^ 393-1 Wind tunnel test results for the cube-van model (1/6 scale) todetermine optimum locations of vertical fences 1 and 2.^463-2 Wind tunnel test results for the cube-van model to determineoptimum location of horizontal fence 3.^ 473-3 Wind tunnel test results for the cube-van model with one horizontaland tow vertical fences. Note the horizontal fence extends to theentire width of the van. 483-4 Wind tunnel test results for the cube-van model with two horizontaland two vertical fences.^ 493-5 Wind tunnel test results for the cube-van model showing optimumfour fence configuration.^ 503-6 Effect of momentum injection and kit position on the drag reductionof the cube-van model.^ 523-7 Effect of the height and spacing of the two vertical fences on thedrag of the cube-van model. Note, the kit is in its optimumorientation, however, the cylinder is not rotating. Of course, therounded upper leading edge presented by the cylinder 543-8 Effect of the cylinder rotation on the drag of the cube-van in thehybrid MSBC/fence configuration.^ 553-9 Comparative force regression analysis showing the effect of fences.613-9b Comparative force regression analysis showing the effect of fencesin terms of CD.^ 623-10 Reduction in fuel consumption due to the presence of fences asaffected by the speed.^ 633-11 Wind tunnel test results for the cube-van model with roundedcorners and vertical fences.^ 673-12 Wind tunnel test results for the cube-van model with roundedcorners, vertical, and horizontal fences.^ 68B-1 Regression analysis of the force data for the modified truck withextended roof (baseline or reference case).^ 89B-3 Analysis of the force data obtained with one horizontal and twovertical fences installed on the cube-truck during the road testsTest Numbers 1-4.^ 109C-1 Fuel consumption data for the reference (baseline) configurationTest Numbers 1-12.^ 114C-2 Fuel consumption data with horizontal fenceTest Numbers 1-8.^ 126viiiC-3 Fuel consumption as affected by the truck speed in the presence ofone horizontal and two vertical fencesTest Numbers 1-4.^ 134ixLIST OF TABLESTable2-1 Geometry of horizontal and vertical fences used during the road-tests.^ 262-2 Details of the road tests conducted with the full-scale truckconfigurations.^ 373-1 Results based on the road test data^ 59xNOMENCLATUREA^frontal projected area, m2CD^coefficient of drag, D / (1/2) p U2 Ad^hydraulic diameter, (4.100 1/2, mF^force exerted by the wheels to propel the truck, NHf fence height, mHk height of MSBC kit above the truck box, mp.^standard air viscosity, 17.8 x 10 -6 , kg/msP^barometric pressure, kPaPf^pressure on the front face of a body, PaPr^pressure on the rear face of a body, Pap^air density, kg/m3Rn Reynolds number, pUd/ptT^absolute temperature, \u00C2\u00B0KU^free stream velocity, m/sU,^surface velocity of the cylinder, m/sxf^horizontal coordinate of the vertical fence, myf^vertical coordinate of the horizontal fence, mxiACKNOWLEDGEMENTSpecial thanks is extended to my supervisor, Dr V.J. Modi for his time andguidance throughout this project. His insight and amiable nature hasmade this project a thoroughly enjoyable experience.The assistance of Oliver B. Ying, M.A.Sc Graduate and Mr Jimmy Ng,Project Engineer in the installation of instrumentation, and advice aregratefully acknowledged.The investigation reported here was supported by the Science Council ofBritish Columbia, Grant Nos. AGAR 5-53628, 5-53698, Natural Sciencesand Engineering Research Council of Canada, Grant No. A-2181, and C.P.Express Transport Ltd.xii11^INTRODUCTION1.1 BackgroundWith the dwindling fossil fuel reserves, and our ever increasingdependence on them, it is no longer sufficient to have a machine or devicethat simply 'works', it must work at its optimum efficiency. As fossil fuelreserves dry up, the cost in obtaining them will increase, and hence itrightly makes economic sense to improve efficiency to as high a level aspossible. Of course, the economist will also asses the cost involved inmaking the device more efficient, which, in fact this must be taken in toconsideration in any engineering research. The proposed project dealswith one such situation, that of minimization of the aerodynamic drag oftrucks.The truck transportation industry is currently very large in Canada,and is continuing to grow. Presently two thirds of all goods in NorthAmerica are transported by trucks, and the average truck travelsapproximately 130,000 - 150,000 km per year. At highway speeds (80 -100 kph) approximately 50 % of the power is used to overcomeaerodynamic drag. With this in mind several researchers and truckmanufacturers have produced various fairings with moderate success.Existing fairings have achieved good reductions in fuel consumption.An example of this is the \"Nose Cone\". Its manufacturers claim areduction in fuel consumption of 5% under ideal circumstances [1]. The\"Nose Cone\" fairing costs up to $555, excluding the installation charge.Similar types of devices, such as the \"Air Shield\", make claims of the sameorder of saving, but are more expensive.Ying [2] indicated that fences could effectively reduce aerodynamicdrag on trucks by 23 %. These devices are passive, simple to manufacture,and install, and seem are quite promising. The drag reduction correlateswith a fuel consumption reduction of approximately 17% at a speed of 100kph. This is a marked improvement from the claimed results of \"NoseCone\", and hence warrants further investigation. Thus encouragingresults of Ying on fences as drag reducing elements forms the basis of thepresent study.A specific model of a cube-van was tested in the wind tunnel to arriveat an optimum configuration of the fences. The corresponding prototypewas road tested to determine the actual drag reduction and fuel savingrealized.1.2 Theory of FencesThe operating principle of fences is rather elegantly simple. Theobjective is to trip the boundary-layer and thus interfere with the pressurerecovery. The resulting pressure is lower. Fences, when carefully placedon the front face of a bluff body would thus lead to a reduction in drag.The principal is illustrated in Figure 1-12Figure 1-1Di = jj P i dx dyDa > DbA schematic diagram showing the effect of fences on the frontface of a bluff body, on pressure distribution: (a) Withoutfences; (b) with fences.As can be seen from the above figure the effective frontal area hasbeen reduced, the pressure distribution over the front of the body has alsobeen reduced, and hence there is a significant reduction in the drag of thebody.1.3 A Brief Review of the Relevant LiteratureA comprehensive literature review of road vehicle aerodynamicssuggest that although the aerodynamically contoured car design hasbecome a standard practice lately, trucks and buses have changed littleduring the past 30 years [3-6]. Most of the modifications have beenlimited to rounded edges with provision for vanes, skirts, and flowdeflectors. The benefit due to some of the \"add-on\" devices is still a matterof controversy and, at best, marginal under conditions other than thespecific ones used in their designs. Bearman [7] has presented anexcellent review of the subject (with 54 references cited). The thesis byWacker [8] also discusses the limited influence of the \"add-on\" deviceswith a possibility of increasing the drag under non-optimal conditions. Onthe other hand, it was found that judicious choice of ground clearance, gapsize between the tractor and the trailer, and back inclination can reducethe drag coefficient by a significant amount.The comprehensive investigation, aimed at assessing the effect of theboundary-layer modification on truck aerodynamics, is presented by Ying[2]. He investigated both the influence of judiciously placed fences as wellas the rating elements for Moving Surface Boundary-layer Control(MSBC). His findings have led to the present investigation where a modelof a smaller cube-truck was tested with the boundary-layer controlachieved through both the procedures. The model study was4complemented by extensive road tests with the more attractive, passive,fence concept.A word concerning model and full scale testing of trucks would beappropriate. Aerodynamic testing of trucks began in the mid 1970's.Model tests were initially conducted with deflector type devices, andinvolved simple modifications to existing vehicles. A wide range of deviceswere tested, with an equally scattered range of results, fortunately theywere mostly positive. This led to the installation of deflectors to largetractor trailer units, on long distance hauls. As testing of the devices andresulting modifications progressed, the measured drag reductioncontinued to grow. However, as late as 1987, fleet managementcompanies were not noting significant reduction in their fuel consumption[9]. This sparked interest in full scale testing of the trucks, in an attemptto see why the results did not translate.Buckley [10] developed a new test method, and suggested it as animprovement of the standard S.A.E. J1321 procedure. The approachinvolves two trucks, and a chase car. This method seems to be quiteaccurate, however it is costly. Saunders et. al. [11] used the method inAustralia, where he found the correlation between the wind tunnel andprototype results to be quite poor. In an attempt to determine the cause ofdiscrepancy, his graduate student, Simon Watkins, started a series oftests, which mainly concentrated on the local turbulence levels. Theresults [12] suggest that the wind tunnel studies, in smooth laminar flowconditions (turbulence level < 0.1 %), tend to give more favourable dragreduction results. A turbulence level of 3-4 %, which would increase withyaw, should be more representative of the road conditions. When thispaper was published the author was completing his undergraduate degree5at the Royal Melbourne Institute of Technology, and his final year project,under Watkins, was concerned with the effects of longitudinal turbulenceintensity on saloon cars. His findings indicated that a 15% variation inthe measured drag can be attributed to the turbulence which turned out tobe of the same order as the maximum expected gain due to the devices.Full scale aerodynamic testing of vehicles is an established practice.Many automobile manufacturers conduct extensive tests of their vehiclesat the full scale level. General Motors has carried out systematicexperiments to ensure that the full scale tests are compatible with thewind tunnel results [13]. Generally these tests involve \"coastdown\" whichis relatively accurate, however, significant modifications had to beintroduced to the procedure in order to achieve close correlation betweenthe wind tunnel data and the prototype road results. This modificationmostly involved attending to the simplifying assumptions made inconducting the road-test. Their analysis also concluded that there was novariation of error with C D .A. Although the largest value of the (C D.A) theytested was 1.69 m2, whereas in this report it was greater than 3.0 m 2 .Other coastdown methods take into account yaw angle, and gradewith carefully surveyed test-tracks. Remenda et al. [14] conducted aseries of experiments using a very small vehicle in order to prove that theycould determine the drag coefficient for a worst case scenario. Since theforces on their vehicle were relatively small their accurate measurementwas difficult. The reported results are quite good, however, they dorequire an extensive survey of the test-track prior to starting.Although most of the reported full-scale tests involve the coastdownmethod, there is another approach of equal value based on the steady6state torque method. In this, which is mostly used in Europe, the vehicleis instrumented with torquemeters, and its speed and applied torque aremonitored. This alternative approach to coastdown method is believed tobe just as accurate. Passmore et al. [15] conducted a comparison of thetwo methods. They concluded that for a similar amount of data thecoastdown method provided a slightly more accurate means ofdetermining the drag. It should be noted, however, that it is easier toobtain significantly more data with the steady state torque method thanwith the coast down approach. Passmore et al. took 20 seconds of data, at4 Hz for each speed, which amounts to 80 points for each speed. Testswere conducted at 12 speeds making a data pool of 960 points. In thepresent study a sample rate of 1 Hz was used as no additional informationor accuracy can be obtained by employing a higher sampling rate. Thisslowing of the sampling rate, whilst increasing the number of data pointsapproximately 6 fold would extend the data acquisition period by a factorof 23. By extending the test the variable conditions have more time toaverage out, and hence increase the general accuracy.Test results have also been reported have been conducted which givea 'rule of thumb' as to the amount of power required to overcome variousdrag forces resisting the vehicle motion, and the importance of yaw angleanalysis. Drollinger [16] found that at 58 mph (94 kph) the aerodynamicdrag is equal to the rolling resistance. It was also reported that there isonly a 10% probability of yaw the angle being grater than 40, and only a3% probability of yaw the angle being greater than 100. Hence, yaw anglemeasurement was not considered vital. Earlier Inagawa [17] also foundthat aerodynamic drag was equivalent to rolling drag at approximately100 kph. A small discrepancy in the predicted speed may be attributed to7differences in the shape of the trucks. He went further and stated thatthe rolling resistance was approximately 1% of the truck weight (loaded orunloaded) with the recommended tyre pressure. In the present study, thiswould amount to a total Coulomb friction force of around 400 N.Ever since the introduction of the boundary-layer concept by Prandtl,there has been a constant challenge faced by scientists and engineers tominimize its adverse effects and control it to advantage. Methods such assection, blowing, vortex generators, turbulence promoters, etc. have beeninvestigated at length and employed in practice with a varying degree ofsuccess. The vast body of literature accumulated over years has beenreviewed rather effectively by several authors including Goldstein [18],Lachmann [19], Rosenhead [20], Schlichting [21], Chang [22], and others.However, judicious tripping of the flow on large bodies using fences hasreceived relatively less attention.Irrespective of the method used the main objective of a controlprocedure is to prevent, or at least delay, separation of the boundary layer.A moving surface attempts to accomplish this in two ways :(i) it retards growth of the boundary layer by minimizingrelative motion between the surface and the free stream;(ii) it injects momentum into the existing boundary layer.A practical application of the moving wall for boundary layer controlwas demonstrated by Favre [23]. Using an airfoil with the upper surfaceformed by a belt moving over two rollers (Figure 1-2), he was able to delayseparation until the angle of attack reached 55 0 where the maximum liftcoefficient CLmax = 3.5 was realised. Alvarex-Calderon and Arnold [24]8carried out test on a rotating cylinder flap to develop a high lift airfoil forSTOL type aircraft. The system was tested in flight on a single enginehigh-wing research aircraft, and appeared quite promising.Of some interest is the North American Rockwell designed OV-10Aaircraft with was flight tested by NASA's Ames Research Center (Cichy etal. [25], Weiberg et al. [26], and Cook et al. [27]). Cylinders located at theleading edges of the flaps were rotated at high speeds with the flaps inthe lowered position. The main objective of that test-program was toassess the handling qualities of the propeller powered STOL type aircraftat higher lift coefficients. The aircraft was flown at speeds of 29 - 31 m /s,along approaches up to - 8\u00C2\u00B0, which correspond to a lift coefficient CL=4.3.In the pilot's opinion any further reductions in approach speed werelimited by the lateral directional stability and control characteristics.9Figure 1-2 The practical application of moving wall for boundary layercontrol.I-40In terms of trying to understand the phenomenon at the fundamentallevel Tennant's contribution to the field is significant. Tennant et al. [28]have conducted tests with a wedge shaped flap having a rotating cylinderas the leading edge. Flap deflection was limited to 15\u00C2\u00B0 and the criticalcylinder velocity necessary to suppress separation was determined. Theeffect of increasing the gap-size (between the cylinder and the flap surface)was also assessed. No effort was made to observe the influence of anincrease in the ratio of cylinder surface speed (U) to the free streamvelocity (U) beyond 1.2.Through a comprehensive wind tunnel test-program involving afamily of airfoils with one or more cylinders forming moving surfaces,complemented by the surface singularity numerical approach and flowvisualisation, earlier studies by Modi et al.[30-33] have shown spectaculareffectiveness of the concept, which increased the maximum lift coefficientby move than 200 % and delayed stall angle to 48\u00C2\u00B0.Yet another approach to boundary-layer control can be through itstripping by judiciously located fences on the front face of a bluff body.This interferes with the pressure recovery thus promising to reduce drag.The basic concepts involved in the boundary-layer control throughthe above two methods are illustrated in Figure 1-3. It shows a bluff body,a tow dimensional prism, located in a fluid stream at zero angle of attack.Pf and Pb are pressures on the front and rear faces, respectively. They areassumed to be uniform over the faces, in this illustrative example, forsimplicity. Obviously by increasing Pb and /or decreasing Pf we can reducethe pressure drag. MSBC tries to increase Pb by keeping the flowattached. On the other hand, fences reduce Pf by tripping the boundary-11layer. These principles are explained through diagrams of flow past acircular cylinder in the same figure. At the stagnation point the pressureis the largest and the pressure coefficient is 1. The boundary-layerseparates at es forming the wake. In the wake, the pressure is essentiallyuniform at a lower value. This is what fences try to achieve. If theseparation is prevented, ideally the pressure will reach the stagnationvalue. This is what the MSBC tries to accomplish.12It is apperent that by increasing Pb or reducing Pf we canreduce the pressure drag.* MSBC tends to increase Pb by keeping the flow attached* Fences tend to reduce Pf by tripping the boundary-layerand preventing the pressure recovery. Potential Flow\u00E2\u0080\u00A2Figure 1-3 Schematic diagrams explaining principles of the MSBC andthe boundary layer trip devices in reducing drag on bluffbodies1.4 Scope of the Present InvestigationThe present study builds on this background and assesses theeffectiveness of the above mentioned two boundary-layer controlprocedures in reducing drag of a cube-van. The carefully planned projecthas two phases:(a)^Wind tunnel tests using 1/6 scale model of a cube-van with:(i) the MSBC applied at the top front edge of the van;(ii) the trip fences mounted on the front face of the van(b)^full scale prototype road tests using the passive fence withpromised to be quite effective, easy to install and maintain,as well as economical.It would be appropriate to point out that eventual application offences is intended for semi-trailers, i.e. tractor-trailer truckconfigurations. Long distance journeys at a relatively high but essentiallyuniform speed would make such configurations ideal candidates foraerodynamic drag reduction. However, at this stage of the developmentthe focus is not on the acquisition of precise data but on the validity of theconcept and gaining some appreciation as to the extent of benefit that canbe realized. A relatively inexpensive cube-van, with appropriateinstrumentation and well throughout test methodology as well as datareduction procedure can provide the necessary information. Hencecommitment to a high cost associated with acquisition, operation andmaintenance of semi-trailer was considered unwarranted at this stage ofdevelopment.14However, factors associated with implementation of the MSBC(accomplished through a kit, comprising of a bearing mounted poweredcylinder, that can be bolted to the front face of the van) were thoroughlyexplored. At the outset, it was recognized that installation of fences wouldbe quite simple, on the other hand implementation of the MSBC devicewould require careful planning. Construction of a 600 mm diameter, lightbut sufficiently strong cylinder, supported by a pair of self aligningbearings, rotating at around 5000 rpm and dynamically balanced, doespresent a challenge. In fact, local inquiries, including the BCIT MachineShop (perhaps the best in B.C.) suggested that no such facility wasavailable in greater Vancouver. Furthermore, although there wereseveral options as to the drive system for the cylinder rotation, they allwould require careful study. For example, power required by an electricmotor, though small compared to the truck power, was far in excess of thecurrent electrical system rating. Hence installation of a larger alternatorwould be required. One option would be to power the cylinderhydraulically. A pump driven by the truck engine may power thehydraulic motor. However, this would entail additional equipment andinstallation cost of a significant amount. A possibility of compressed air-motor drive system was also explored to gain some appreciation as to therelative merit and problems involved.The scope of the investigation and various phases involved may besummarized as follows:1516a) Model tests in the wind tunnel using both MSBC and fence typedevices. Acquisition of the measuring instrumentation and theirinstallation.b) Development of a test methodology that would accurately determineaerodynamic drag as well as rolling resistance, and internal viscouslosses.c) Establish a methodology for data processing, to get reliable,repeatable, and accurate results. Development of the softwarenecessary for interrogation of instrumentation in an on-roadenvironment, as well as all post processing of data.d) Road tests with several configurations of a cube-van to assessperformance of fences as a drag reducing device. Configurations ofa GMC van tested were:i) standard box, as supplied by Grumman Olsen Ltd;ii) standard box, with sharp corners and raised roof;iii) same as ii) with horizontal fence added;iv) same as iii) with two vertical fences added.The fence locations were determined from wind tunnel tests. It maybe pointed out that the cube-van model used in the wind tunnel tests,though similar, differed significantly in detail from the prototype.Furthermore , the model was based on a older version which had a muchlarger hood.2 TEST PROCEDUR.E2.1 Wind Tunnel Test Procedure2.1.1 Model specificationsA 1/6 scale cube truck model was constructed using Plexiglass. Themodel is based on a mid 1980's GMC truck, with van body mounted on it.It has a hydraulic diameter of 469.2 mm, and can be used to assesseffectiveness of both the MSBC device and the boundary-layer trip fences,or combinations of the two.17Figure 2-1 Photograph of the 1/6 scale cube-van model2.1.2 Wind tunnelThe truck model was tested in the boundary-layer wind tunnel at theUniversity of British Columbia (Figure 2-2). The tunnel is an open circuittype powered by an 80 kW three phase motor, which drives an axial flowfan at a constant 700 rpm. The tunnel wind speed is varied using apneumatic controller to alter either the rotating frequency of the fan orthe blade pitch. The settling section contains a honeycomb and fourscreens to smooth the flow as it enters a 4.7:1 contraction section. Thetunnel has a test-section of 2.44 m width by 1.6 m height, and in 24.4 mlength, consisting of eight 3.05 m long bays with a variable height roof toallow for boundary-layer correction. The stable wind speed of the tunnelis in the range of 2.5 to 25 mis. The adjustable test-section roof was set forzero pressure gradient. The present set of experiments were carried out inthe second bay which provided smooth flow with a turbulence level lessthan 0.4 %. The typical test Reynolds number based on the hydraulicdiameter was 2 x 10 5 .2.1.3 Model support system and instrumentationThe truck model was supported by four steel guy wires which weresuspended from the ceiling and carried turnbuckles to help level themodel. As the length of the wire (1450 mm) is much larger than themaximum horizontal displacement of the truck model (< 50 mm) the draginduced displacement was essentially linear in the downstream direction.18Figure 2-2 University of British Columbia boundary-layer wind tunnelVariation in the drag due to the fences was relatively small, andrequired the use of a sensitive transducer for its measurements. Themodel was suspended from the ceiling by four wires, as describedpreviously, to minimise the effect of friction. The drag induceddownstream motion of the model was transmitted via an inelastic cable toa cantilever beam with a pair of strain gages near its root. The gagesformed a part of the Wheatstone Bridge (of the Bridge Amplifier Meter,BAM) and the amplified, filtered output was recorded using a DISAVoltmeter. The sensitivity of the drag measurements was approximately4 N/ V. The calibration of the cantilever using static loads was performedtwice during a test-session, before and after the tests; and the averagecalibration value was adopted to account for any drift. Figure 2-3 showsschematically the model support and drag measurement system.20Figure 2-3 Model Arrangement in the Wind Tunnel2.2 Full Scale Test Procedures2.2.1 Truck specificationsThe truck tested was a 1991, Chevy, one tonne Cube Van. This typeof vehicle is generally used for small deliveries, and transport aroundused this vehicle at a wide range of speeds. The vehiclesbelow:town. Our testspecification appears\u00E2\u0080\u00A2 Model G30 Commercial Cut-away,146\" WB, 2WD\u00E2\u0080\u00A2 Body And Trim Standard Model\u00E2\u0080\u00A2 Package Code Level 1\u00E2\u0080\u00A2 GVWR 4536 Kg\u00E2\u0080\u00A2 Engine 5.7 litre V8\u00E2\u0080\u00A2 Transmission 4 Speed Automatic\u00E2\u0080\u00A2 Mirrors Camper Type, Painted\u00E2\u0080\u00A2 Final Drive 4.1:1\u00E2\u0080\u00A2 Tyre Sizes LT225/75R - 16/D\u00E2\u0080\u00A2 Tyre Pressure 290 kPa\u00E2\u0080\u00A2 Box Height 2.150 m / 2.740 m\u00E2\u0080\u00A2 Box Width 2.420 m\u00E2\u0080\u00A2 Frontal Area 6.016 m2 / 7.444 m2\u00E2\u0080\u00A2 Hydraulic Dia. 2.768 m / 3.908 m\u00E2\u0080\u00A2 Reynolds No. 5 x 10 5 To 8 x 10622A photograph of the truck with its additional height is presented inFigure 2-4. As can be seen the removable top section is in the plainaluminium colour, thus illustrating its relative size.Figure 2-4 The truck as tested with an increased frontal area. This wasachieved by enlarging the box through raising the roof. Theextension was in a removable modular form. This is referredto as the 'extended roof case.2.2.2 Test configurations and instrumentationFour different configurations of the truck were used during the road-tests: the plain cube van as purchased; the one with the extended roofadded; the extended roof truck with a horizontal fence, and finally the onewith one horizontal and two vertical fences installed at their optimum23locations as determined by Ying [2]. The extended roof configurationserved as the baseline reference. Each of the configurations was subjectedto extensive road tests to ensure repeatability and accuracy of data.Figure 2-5 and 2-6 show the extended box configuration with a horizontalfence, and horizontal-vertical combination respectively. The heights andlocations of the fences appear in Table 2-1.24Figure 2-5 The extended roof truck with a horizontal fence.Figure 2-6 The Truck with the extended roof, and all fences25Table 2-1 Geometry of horizontal and vertical fences used during theroad-tests.Fence VerticalLocation(Yf)Yf /dHorizontalLocation(Xf)Xf IdlengthSizeheight thicknessHorizontal 1125 mm 0 288 - - Full width2240 mm152/0.039* 1/0.00026 *Vertical - - 968 mm 0.333 1420 mmFrom top152/0.039* 1/0.00026*Note : The box dimensions were 2.740 x 2.240 m (height x width)Hydraulic diameter, d, equals 3.908 m.* Nondimentionalized with respect to d.The instrumentation consisted of two transducers: a torquemeter,and a fuel meter. They were connected to a Dycor DA \M 100 analog todigital converter, which in turn was interfaced with a 386 SX lap topcomputer. Figure 2-7 shows a block diagram indicating the general set-upof the equipment on the truck. The computer logged data from eachtransducer, sampled at a rate of 1 Hz, and displayed it in real time. Theon-screen output was designed to display the drive train force versusvehicle speed, an example of which can be seen in Figure 2-8. The steadystate torque method permits an accurate determination of the dragcoefficient, its variation as a function of speed. This results in the preciseestimate of drag reduction, effectively incorporating the yaw anglesensitivity and Reynolds number variations in a polynomialrepresentation of the force as a function of speed.26A photograph of the test set-up appears in Figure 2-9. As can beseen, the A/D card is mounted on the front of the case, the torque metersignal processor is inside, and the computer is fixed on the top.27Figure 2-7 A block diagram of the test set up.28Figure 2-8 An example of the computer display.29Figure 2-9^Equipment layout in the truck.2.2.2 Torque and speed measurementsThe torque and speed were measured by a torquemeter which wasinstalled in the drive shaft of the truck. It was a 6-02T noncontacttransducer manufactured by S.M. Himmelstein in the U.S.A. Aphotograph of the torquemeter as installed is presented in Figure 2-10.30Figure 2-10 Torquemeter as installed in the truck.Torque calibration was conducted by putting the truck on a constantslope road in the 'park' setting and releasing the brake. This effectivelyloads the torquemeter with the force acting down the hill. Since torque isproportional to force ( the radius and gearing remaining constant ), adirect calibration of force can be obtained. The set-up is schematicallyshown by Figure 2-11. To ensure that the truck is placed on an evengrade, the angle was measured at each of the four wheels. It is notpossible to determine the normal loads at each wheel, since the height ofthe center of mass is not known. However, this is of no consequence as forcalibration the normal force can be readily estimated for the constantgradient at the wheels. This was ensured using a spirit level whichshowed a deviation of less than 1:400 at the wheels.31Velocity calibration was carried out by driving the truck over aknown distance at a constant speed. The velocity of the truck was loggedby the computer every 0.5 seconds. The average of this result was thencompared to the average velocity of the truck, which was determined bythe distance divided by the time. Tyre pressure was also noted as toensure the same setting during future tests.Since the torque and angular velocity of the drive shaft were notmeasured directly, accuracy of the manufacturer's calibration data cannotbe assessed. It was decided to obtain this result independently. To thatend, the truck was driven down the road at a constant speed, and thetorque and drive shaft angular velocity recorded. Simultaneously thecomputer logged the velocity and truck drag force. From basic physics,power is equal to both the product of force and velocity, and the product oftorque and angular velocity, thereby giving a means of validation. Thischeck showed excellent correlation between the manufacturer's calibrationdata, and the results given by the present test procedure.2.2.3 Fuel flow measurementThe fuel flow meter that was installed was a micro oval type. It wascalibrated by pumping fuel through it and measuring the number ofpulses generated for a fixed volume. The calibration confirmed themanufacturers specifications of one pulse per millilitre. Since the truckhas a carburettor type fuel system, and other engine options are availablefor fuel injection, the pressure of the pump exceeds the float buoyancy.Hence a return fuel-line is installed in the vehicle. To account for the fuelflow in the return line for accurate estimation of the fuel consumptionthere were two options; to install another fuel to measure the amount of33fuel being sent back to the tank; or alternatively, the head loss in thesupply line could be decreased, so that the pressure supplied to thecarburettor was low enough to allow proper functioning of the needle andseat valve. The second option was chosen in the present study. The headloss due to installation of an additional in-line fuel filter and the fuel flowmeter was sufficiently high to eliminate the need for a return line atnormal operating speeds. It should be noted, however, that at idle thetruck does tend to run a little rich. This is deemed to be insignificant.34Figure 2-12^Photograph of the fuel meter installed on the truck.2.2.4 On-road test procedureA procedure was developed to ensure that the results obtained duringthe road-tests were accurate and repeatable. The following is an outline ofthe procedure used to collect the data on the route described in Table 2-2.It was found that the repeatability and accuracy of the tests were farimproved when the tests were performed at night. This gives theadvantage of both lower wind levels and reduced road vehicle interference.In addition, a device should be compared with a baseline referencedirectly, i.e. the baseline test, and the configuration of interest should betested on the same night. The final four tests of each configuration weredone on four nights, with all the three configurations each night:i) Measure the ambient air temperature and pressure.Note : This information can be obtained after the test, bynoting the time at the start and end of each run, andthen calling Environment Canada [(604) 664 9156].The Vancouver International Airport site is the closestreference point to the test site.ii) Switch on the power, and ensure that the equipment isin working order using DAMCAL1.EXE. The truckshould be parked on level ground, in neutral, with thepark brake disengaged. Thus the file created byDAMCALLEXE will also confirm the offsets of thetorquemeter.iii) Run the data logging program (TRUCK.EXE) on thecomputer. It may be pointed out that mass is arequired input for this program. It is designed to havean accelerometer installed in the vehicle to account forchanges in velocity and gradient. Accelerationcompensation is beyond the scope of this project, but isanticipated in the future work. The mass input for theprogram, for this stage of the project, is not used, so it35does not have to be determined prior to the test. Anarbitrary value (say 3400 kg) should be entered toenable the program to run trouble free.iv) Commence driving, accelerating to a desired speedgiven in Table 2-2, and hold it constant.v) Commence Logging data (note the first satisfactorypoint). Keep account of any interferences that may beencountered, such as vehicles overtaking and pullingin front, changes in gradient, large trucks eitherpassing, or being passed etc. There should be nosignificant grade changes in the route indicated inTable 2-2.vi) After the desired number of points have been logged,usually 250, note the last recorded point, and adjustthe vehicle speed to the next desired level.vii) Repeat steps (iv) - (vi) until the test route has beencompleted as per Table 2-2Speed zones are listed in the table, and grid references can be locatedon the map (Figure 2-13). It may be pointed out that no effort was madeto obtain a statistical distribution of drag as this would involve a largenumber of tests with associated fuel costs and time. However, as pointedout later the standard deviation of the CD was within 2.5 %.36Table 2-2 Details of the road tests conducted with the full-scale truckconfigurations.Test &File No.Vehicle Speedand DirectionInitialCoord.End.Coord.Time[s]CommentsFile #11.1 100 km/hr E 968387 032874 250 Start test after rise.1.2 90 km/hr E 032874 099874 2501.3 80 km/hr E 106873 129357 125 Start test after rise.1.4 80 km/hr W 129357 106873 1251.5 70 km/hr W 099874 050874 250 Start test after rise.1.6 60 km/hr W 050874 000874 2501.7 50 km/hr W 000874 968387 250 Exit, Stop, and Save.File #22.1 40 km/hr E/N 964397 976414 2502.2 10 km/hr N 976414 976423 2502.3 10 km/hr S 976423 976414 2502.4 40 km/hr S/W 976414 964397 2502.5 30 km/hr E/N 964397 976407 2502.6 20 km/hr N 976407 976423 250 Stop and Save.File #33.1 20 km/hr S 976423 976407 2503.2 30 km/hr S/W 976407 964397 250 Stop and Save.3738(Table 2-2^continue)Test &File No.Vehicle Speedand DirectionInitialCoord.End.Coord.Time[s]CommentsFile #44.1 50 km/hr E 968387 000874 2504.2 60 km/hr E 000874 042874 2504.3 70 km/hr E 050874 099874 2504.4 80 km/hr E 106873 129357 125 Start test after rise.4.5 80 km/hr W 129357 106873 1254.6 90 km/hr W 099874 032874 250 Start test after rise.4.7 100 km/hr W 032874 968387 250 Exit, Stop, and Save.File #55.1 110 km/hr E 968387 050874 2505.2 110 km/hr W 050874 968387 250END OF TEST FORTHIS SET-UPNote: Grid References with an x coordinate greater than 900 refer to map92G2 and Grid references with an x coordinate of less than 200 refer tomap 92G3 of the \"Canadian Topographical Maps\" 1:50 000Figure 2-13^Test route map (n.t.s.).Approximately 4000 data points were required to determine the dragcharacteristics of a given configuration. The estimate was arrived at byexperience. Tests undertaken in this thesis have approximately 5500points to ensure greater accuracy than was obtained by Passmore [16] andalso allow a greater time for random factors to even out. This representslogging of 250 points at each of the 11 different speeds in 2 directions.The force on a vehicle is a function of velocity. The function is taken tohave the form:F = ao + a iV + a2V2 ,where :^ac, = rolling resistance;a l = viscous drag coefficient in the final drive, combinedwith the tire drag;a2 = aerodynamic drag coefficient, pAC D/2Now, from the generalized gas law, p = P/RTWhere R = 0.287 kj/kg.KIn the present study the torquemeter was calibrated directly for forceand velocity as mentioned before. Thus all steady state torquemeasurements were made with the same device. This combination provedto be the most accurate as zero load transmission losses wereautomatically accounted for.To process the final data set five different methods were used. Theseare summarized below:4041a) In the first method a polynomial regression program waswritten (POLYREG.EXE) which used the least squaresmethod of curve fitting. This program gave the velocitycoefficients of the force function. A listing of this programappears in Appendix A. The regression analysis wasconducted using all the data points, without any pre-averaging procedure. Since the test procedure had 22distinct stages, at an evenly spaced velocity, the regressionanalysis should be weighted evenly. This was the firstprocedure used in the data analysis, and it is the simplest.Surprisingly, it is also one of the most accurate.b) The second method averaged the force at each of the discretevelocities, thus ensuring far less weighting of a least squaresanalysis. A program was written to perform thediscretisation which is called DSCRT2CL.EXE' and appearsin Appendix A. A wide range of speeds must also be tested,with a fairly even distribution of speeds, to maintainaccuracy of the regression analysis. Care should be taken toensure that the same number of points are taken in eachspeed range, within a 10% tolerance. The preprocessed dataserved as input to the regression program.The method proved to be somewhat unreliable in its outcome,and is not considered as accurate. It is not recommended foranalysing data generated in this form of tests.c)^As an improvement on the method in (b), all velocities withless than 50 data points were removed prior to the regressionanalysis. The removal of points was undertaken using aspreadsheet program. This resulted in some improvement,however, the results were still not as reliable as thosegenerated in the test procedure (a).d) The next analysis procedure was to take a time averagedsample to get an average force for a particular average speed.This procedure was implemented with 10, 25, and 250 secondtime averaged samples. Each of these then served as input tothe polynomial regression program. The 250 point timeaverage proved to be the most accurate. It gave 2 sets of 11equally spaced points from 10 to 110 km/hr. To preform thisprocedure 'TAVG.EXE' program was written, which is listedin Appendix AMethod (d) provided the best results compared to the earlierprocedures. The consistency of the test can be shown bycomparing the 10, 25, and 250 s time averaged results.Ideally, if the tests were perfect, these results shouldcoincide. In most cases, the discrepancy was less than onepercent.e) The final method of analysis, and the one which is used tohere is a modification of the one given in (d). To begin with,the average force coefficients for both rolling resistance andinternal brake, and viscous drag were averaged for 24 tests.These nonaerodynamic forces were then removed from eachdata point, leaving only the aerodynamic force. Since theaerodynamic resistance is proportional to the square of the42velocity, the velocity terms were converted to m/s andsquared. Following this a linear regression analysis, withzero offset, was performed to determine the coefficient for theaerodynamic term. The worst standard deviation over 4 testswas 2.5%. This procedure was implemented using the'TAVGRM.EXE' program which appears in Appendix AThis method was found to be the most accurate. It gave astandard deviation for most configurations of approximately2.5 %. It involves the assumption, made at the start of thetest-program, that a change in the shape of the truck, onlyaffects the aerodynamic term. Thus it gives a realisticcomparison of forces between the two cases.During the road tests, difficulties at times did arise in separating theviscous term from the aerodynamic drag. For example if the tests arebeing conducted in a relatively windy condition, the viscous dragcontribution appears to be larger. This is because the effective yaw angleof the vehicle decreases as the vehicles speed increases. Now the CD valuedecreases with an increase in speed, thus making the forces more linearwith respect to the velocity. Care was taken to avoid high winds as muchas possible. To minimize the effects of turbulence, the last set of four testsfor each configuration was carried out at night, between 10 p.m. and 5a.m., and a qualitative measure of the wind was noted by several flags onthe test-route.The fence performance proved to be somewhat speed dependant.This resulted in a much lower, and some times negative viscous term,whilst the aerodynamic drag term actually increased. It seems logical to43expect that the CD of the truck depends on the velocity. The normalsecond order polynomial regression does not take this into account. Theonly way that the regression analysis can account for this variation is bymodifying both the viscous, and aerodynamic drag terms of the equation.To reduce the effect of the speed dependency, the method described in (e)was adopted.443 RESULTS AND DISCUSSION3.1 Wind Tunnel Model tests3.1.1 Wind tunnel models with fencesThe wind tunnel model was tested with various arrays of fences. Themodel had a baseline drag coefficient CD0 = 1.43. Initially only verticalfences were installed on the model, these gave a drag reductions of 12.5 %for an optimal height (hf Id) of 0.054. When a horizontal fence is added tothe two vertical fences configuration the drag coefficient decreased to 1.03,a 28.0 % reduction based on the reference case. Extensions were thenadded to the horizontal fence, so that it reached the full truck width, andagain there was improvement in the performance (C D of 0.999, a totalreduction from the baseline of 30.1 %). A second horizontal fencecontinued the trend and further reduced the drag slightly, to 0.981, i.e. adecrease of 31.4 % !. Figures 3-1 through 3-5 detail the progressionthrough these sequences of fence configurations. It was decided that theadditional fourth fence would not be used on the full-scale truck as it gaveonly a 1.3 % improvement over the three fence case. Furthermore the roadtest error would be, perhaps, larger than this.450.450.40.351.41.351.3hf/d \u00E2\u0080\u00A2 0.081h f /d \u00E2\u0096\u00A0 0.054hf/d \u00E2\u0080\u00A2 0.030bfv/d \u00E2\u0096\u00A0 0.43Rn \u00E2\u0096\u00A0 2x10 5CDos 1.430.3Co1.251.21.150.25xf /dFigure 3- 1 Wind tunnel test results for the cube-van model (1/6 scale) todetermine optimum locations of vertical fences 1 and 2.0.16^0.18 0.20.12 0.14 0.221.3(hf /d)cr = 0.054(xf /d)cr = 0.38bfv/d = 0.43R n a 2x10 5CDo\" 1.431.151.11.0511.251.2CDyif3 /dFigure 3-2 Wind tunnel test results for the cube-van model to determineoptimum location of horizontal fence 3.Figure 3-3 Wind tunnel test results for the cube-van model with onehorizontal and tow vertical fences. Note the horizontal fenceextends to the entire width of the van.CDYf4 idFigure 3-4 Wind tunnel test results for the cube-van model with twohorizontal and two vertical fences.3.1.2 Wind tunnel model with a rotating cylinder and fencesA rotating cylinder was installed in kit form, so as to be comparableto a similar configuration suitable for the full-scale truck, on the front ofthe cube. The kit contained the cylinder, drive motor, and the speedmonitoring and control system. Being located outside it does not infringeon the payload space. The kit position, which controls the projection of thecylinder in to the free stream, and the cylinder speed were variedsystematically to establish favourable operating conditions. The results,presented in Figure 3-6, clearly show that there is an optimum position ofthe kit (lik /d = -0.021). As expected, for a given position of the kit,injection of momentum through the cylinder rotation is to delay separationof the boundary-layer from the top face of the van which, in turn, isreflected in the drag reduction. For U, /U = 3 and the optimum kitlocation corresponding to li k /d = -0.021, the change in CD from 1.24 to 1.03represents a drag reduction of around 28%, essentially the same as thangiven by the fences. Note, any further increase in the U, /U does notresult in significant improvement in the performance.513MN1.41.3CD1.21.11^ Hkici \u00E2\u0080\u00A2 -0.021 Li Hk/d \u00E2\u0080\u00A2 0 Hk/d \u00E2\u0080\u00A2 0.021Hk/d \u00E2\u0096\u00A0 0.042 A Hk /d \u00E2\u0096\u00A0 0.064^Hk/d \u00E2\u0096\u00A0 0.085Uc z11111111111\u00E2\u0080\u00A2 1\u00E2\u0080\u00A2\u00E2\u0096\u00A0 \u00E2\u0080\u00A2 \u00E2\u0080\u00A2^\u00E2\u0080\u00A2 \u00E2\u0080\u00A21116.S MOM.Rn \u00E2\u0080\u00A2 2x10 5U0^0.5^1^1.5^2^2.5^3^3.5^4^4.5/U Figure 3-6 Effect of momentum injection and kit position on the dragreduction of the cube-van model.As the kit provided a flat area exposed to the free stream, it wastempting to introduce fences. With two vertical fences and U, /U = 2, areduction in CD by 41.3 % (based on the original nonkit configuration) israther astounding. Of course, MSBC is not a passive device as the fenceand would involve expenditure of energy through the drive system. Theseresults are presented in Figures 3-7 and 3-8.531.3h f /d \u00E2\u0096\u00A0 0.02 \u00E2\u0080\u00A2h f /d R 0.04h f /d R 0.06h f /d \u00E2\u0096\u00A0 0.08h f /d \u00E2\u0096\u00A0 0.10CD1.21.110.90.3^0.32^0.34^0.36^0.38^0.4^0.42^0.44Xf idFigure 3-7 Effect of the height and spacing of the two vertical fences on the drag of the cube-van model. Note,the kit is in its optimum orientation, however, the cylinder is not rotating. Of course, the roundedupper leading edge presented by the cylinder would contribute to the drag reduction.0.90.8431^2UciUFigure 3-8 Effect of the cylinder rotation on the drag of the cube-van inthe hybrid MSBC/fence configuration.\u00E2\u0080\u00A2^ hf/d \u00E2\u0080\u00A2 0.08h f /d \u00E2\u0080\u00A2 0.10R n = 2=105H k/d \u00E2\u0096\u00A0 -0.02CDo 8 1 . 19U0.7 ^01.31.21.1C.3.2 Full Scale TestsThe baseline (i.e. reference configuration devoid of fences and MSBC)tests were conducted on the cube-van as supplied by GMC, with aGrumman Olsen body and the extended roof. The extension was built byIntercontinental Truck Body (B.C.) Inc. and raised the roof of the truck by600 mm. The raised roof simulates a vast number of larger, customised,trucks, with bigger chassis, that often use drag reducing devices.The truck without its extended roof had a CD of 0.40. However, oncethe extension was placed on the vehicle the CD increased to 0.610, thisrepresented 75.25 % of the total force exerted by the engine to propel thetruck at 100 kph. With the a horizontal fence installed there was only asmall change in the drag coefficient (C D = 0.614). Note, this represents a0.65 % drag increase. Of course, with the standard deviation of the datafor the horizontal fence configuration being 2.5 %, and that for theextended roof at 2.4 %, it is clear that the result represents no net change.However, the situation changed dramatically with the addition of twovertical fences. Now CD reduced to 0.524, a decrease of aerodynamicresistance by 16.6 %. Indeed this is quite significant, representing a dropin power demand of 11.1% and the corresponding reduction in the fuelconsumption of 9.6 % as an average over the speed range.The results showed that the reduction of fuel consumption was farmore noticeable at higher speeds. In the speed range of 90 - 110 kph, itamounted to around 25.1 %. Since this represents the highway speedrange at which most trucks travel, the financial benefits are enormous.The Fences enabled the vehicle to operate in a higher gear, and hence thefuel savings not only came from the fact that the engine was not required56to deliver as much force, but it could deliver the required force at muchlower engine speeds.A factor which may contribute to the uneven reduction in the fuelconsumption is the Reynolds number. CD may be a function of velocityover the range tested, and hence the change in force would vary withspeed.The time averaged force data with polynomial regression arecompiled in Appendix B. Each plot has the test name and number. Thetests are numbered in the reverse chronological order. All tests with agiven number were performed on the same day/night. Time averaged fueldata appears in the same format in Appendix C. A summary of the forceregression in Table 3-1.It should be noted that tests 1 to 4 were performed at night, andhence as expected the baseline (extended roof case) drag values are highercompared to those obtained for the tests performed during the day.Watkins [13] found the CD to decrease with and increase in thelongitudinal turbulence intensity.It is of interest to recognize that performance of the fencesdeteriorates at lower speeds (< 30 kph, Figure 3-10). In fact, the resultsshowed increase in CD. This would suggest that, at a lower speed, thefluid is unable to clear the fences, i.e. it is blocked by them and remainsstagnant, resulting in an increase in pressure and a correspondingincrease in drag. Fortunately, the increase in fuel consumption is rathersmall as the nominal demand itself is rather low at lower speeds.57The force was normalized to the weight of the vehicle in Figure 3-9a.This was done because the drag forces had two nonaerodynamic termswhich contribute approximately one third of the total force at 100 kph andsignificantly more at lower speeds. Alternatively the force can benormalized to the dynamic pressure times the frontal projected area (q.A), and this result appears in Figure 3-9b. The horizontal lines in figure3-9b represent the CD value that the curves asymptotically approach.Each line-type represents one set-up, both in the curves, and theasymptotes.58Table 3-1 Results based on the road test dataTestNo.Test typeV2Coefficientkg/mAmbientTemperature0 CBarometricPressurekPaCD1 ext. roof 2.90 8.2 102.12 0.62hor. fence 2.87 9.4 101.98 0.61all fences 2.43 11.4 101.91 0.522 ext. roof 2.75 7.3 101.30 0.59hor. fence 2.97 9.0 101.27 0.64all fences 2.53 7.9 101.19 0.543 ext. roof 2.85 10.4 102.25 0.61hor. fence 2.85 11.3 102.21 0.61all fences 2.47 10.4 102.25 0.534 ext. roof 2.90 12.5 101.75 0.63hor. fence 2.76 12.6 102.25 0.60all fences 2.30 14.5 101.64 0.505 ext. roof 2.48 20 101.84 0.55hor. fence 2.47 20 101.84 0.556 ext. roof 2.49 22 101.35 0.56hor. fence 2.46 22 101.35 0.557 ext. roof 2.47 23 101.84 0.55hor. fence 2.47 23 101.84 0.5559Table 3-1 Results based on the road test dataTestNo.Test typeV2Coefficientkg/mAmbientTemperature0 CBarometricPressurekPaCD8 ext. roof 2.79 20 102.29 0.62hor. fence 2.59 20 102.29 0.589 ext. roof 2.30 28 101.67 0.5310 ext. roof 3.07 20 101.64 0.6811 ext. roof 2.92 22 102.11 0.6512 ext. roof 2.93 15 101.34 0.6460^0.12^0.10-0.08-02 0.06-L0.04-0.02-^0.00^0 20 40^do^80^lb()Velocity [km/hr]120Figure 3-9a Comparative force regression analysis showing the effect offences.Figure 3-9b Comparative force regression analysis showing the effect offences in terms of CD.3.3 Comparison Between Model and Full Scale TestThe wind tunnel tests with a 1/6 scale model of the cube-truckprovided encouraging performance of fences in reducing the drag throughtripping of the boundary-layer. Of course, ultimate validity of the conceptis assessed through tests in the operating environment as accomplishedhere through the road tests using a full-scale prototype truck. Due toidealized character of the wind tunnel tests, discrepancy between the windtunnel and the road tests results can be expected. In the present study,the wind tunnel tests predicted the drag reduction of around 31.4 % underoptimum fence configuration, however, during the road tests the peakreduction of 16.6 % was realized. Suck a deviation was anticipated andcan be explained quite readily:(i) At the outset, as pointed out before, the rather idealizedcharacter of the wind tunnel tests must be recognized. Itdoes not account for the free stream turbulence, roadboundary-layer, and yaw condition of the real-life situation.Furthermore, the Reynolds number is smaller by around twoorders of magnitude compared to that of the full-scale truck.(ii) The wind tunnel model, being based on an earlier version, issignificantly different from the actual cube-van used duringthe road tests. The main differences lie in the hood design,and sharp corners present in the plexiglass wind tunnelmodel, while all the corners and edges of the truck wererounded. Obviously this would give the prototype truck arelatively small reference (i.e. baseline) drag coefficient. Infact, wind tunnel tests were conducted incorporating rounded64edges in the model with a baseline drag dropping form 1.43 to1.09 (Figures 3-11, 3-12). Note, with the rounded corners andhorizontal-vertical fences, the CD of 0.82 for optimumconfiguration (Figure 3-12) corresponds to a reduction inaerodynamic resistance of 24.8 %. This is closer to the road-tests result of 16.6 % in spite of the difference in geometryand idealized test conditions as mentioned before.(iii) Scaling of fences and their orientation for full-scaleapplication were also approximate. Dimensions of the fenceswere geometrically scaled, so was their orientation. Forpreciseness, they should have been based on the hydraulicdiameter. Furthermore, the truck geometry being different,the optimum location of the fences is still not arrived at.Thus there is a possibility of improvement in the dragreduction through optimized positioning of the fences.(iv) The wind tunnel tests, in the optimized situation involvedfour fences, two horizontal and two vertical. On the otherhand, so far, the road tests have been conducted only withthree fences, two vertical and one horizontal. In an earlierstudy by Ying [2], with a different geometry of the truck,addition of the force fence made further significant reductionin the drag (by around 4-5 %).(v) Yet another factor that would affect correlation between themodel and the full-scale road test results would be due todifferences in the flow under the truck, which can contributesignificantly to the drag. In the present case, the bottom of65the model was made of smooth plexiglass and the tunnel floorwas of wood. On the other hand, the undercarriage of thecube-van has a complex geometry with drive shaft,torquemeter and other components of the transmissionassembly.Considering the above factors, the trends suggested by the windtunnel tests are indeed valuable and accurate.660.90.26 0.34Xfid0.3 , 0.38 0.421.2hf/d = 0.057h f /d = 0.085h f /d \u00E2\u0080\u00A2 0.106R n \u00E2\u0080\u00A2 2.0*10 5CDO \u00E2\u0080\u00A2 1.0910.951.151.1CD1.05Figure 3-11^Wind tunnel test results for the cube-van model withrounded corners and vertical fences.0)as4 CONCLUDING REMARKS4.1 Summary of Results4.1.1 Wind tunnel model testsBased on a rather fundamental study of both the Moving SurfaceBoundary-layer Control and trip-fences with a cube-van model thefollowing general conclusions can be made:(i) MSBC and trip fences are both successful at reducing dragwhen applied to a cube-van model.(ii) With the fences, the drag reductions were 31.4 % for a modelwith sharp edges and 24.8 % for the model with roundededges. This suggested considerable promise for the conceptand need for full-scale tests.(iii) Fences appear to be a promising, cost effective way ofreducing drag on the baseline model, regardless of whether ithas curved or straight edges.(iv) Application of the MSBC through a rotating cylinder kitresulted in a drag reduction of around 28 %. Its hybridapplication in conjunction with the fences led to a surprisingresult of a decrease in aerodynamic resistance by 41.3 %.(v) The power demanded by the MSBC procedure is relativelysmall. In the present study with the 1/6 scale models, itamounted to around 90 W.69(vi) The entirely passive character of the fences, their simplicityand ease of application are indeed quite attractive featuresconducive to full-scale application.4.1.2 Full scale tests(i) In general, the full-scale tests with the fences substantiate thepositive influence of the fences in reducing drag.(ii) Although the trends are accurate, there is a discrepancy betweenthe prototype and the model results. This is attributed tosignificant differences between the model and road-tests conditions.(iii) A decrease in drag coefficient by 16.6 % with two vertical and onehorizontal fences is indeed impressive. It represents a reduction inthe power demand by 11.1 % with a decrease in fuel consumption ofaround 9.6 %, the average over the speed range used during thetests.(iv) Due to differences between the prototype and the model, the fencegeometry and orientations are not yet optimum. Hence, there is ascope for further improvement in the performance.(v) As the fuel consumption would depend on the speed of operation,the 11.1 % reduction in power with the application of fences wouldnow permit use of the overdrive instead of 3rd gear. For the trucknegotiating flat terrain at 100 kph, with overdrive, the reduction infuel consumption amounts to around 39.6 % (Figure 3-10). Ofcourse in actual practice, the net benefit would be significantly70smaller due to energy expended during acceleration, braking,negotiation of the gradient, decent,etc.(vi) The tests account for the yaw effect only in an average fashion.Performance of the fences at various yaw orientation has not beenestablished systematically.^Fortunately, as pointed out byDrollinger [16], most trucks spend a large portion of their time atyaw angles of less than 4 0 (only a 10 % probability of yaw anglesgreater than 40 for commercial trucks). Hence the results are notlikely to be affected significantly by the presence of yaw.(vii) The fuel consumption results showed considerable fluctuations andpresented some difficulty in the curve-fitting. Engagement anddisengagement of gears would contribute to it, however, they maynot completely account for the variations.4.2 Recommendation for future studyThe investigation reported here represents only the first step instudying this exciting, challenging and promising phenomenon of the flowcontrol over a three-dimensional truck. There are a number of aspectswhich demand further attention. Only a few of the more significant areasrequiring further study are indicated below:(i) Wind tunnel test arrangement needs improvement withreference to the model support system, instrumentation,simulation of the ground boundary-layer as well as geometricaccuracy of the model. A systematic study to asses the effectof the Reynolds number, free stream turbulence and yawangle of the model would also provide useful information.71(iii) Hybrid fence configurations with deflectors and MSBC needmore systematic further study.(iv) A cruise control unit should be installed in the cube-van.This would improve measurement of force as well as reducefluctuations in the fuel consumption data.(v) An accelerometer should be installed to account for variationin the speed during acceleration, braking, negotiation ofgradients, etc. Provision for it has been made in the presenthardware as well as software, so its incorporation would bequite simple. This would also permit precise determinationof the energy expended during braking and the effect oftruck-load on it.72REFERENCES[1] Van Setters, R., Service Manager, Intercontinental Truck Body Inc.,B.C., Canada, 1992.[2] Ying, 0.B., \"Boundary-layer Control of Bluff Bodies withApplication to Drag Reduction of Tractor-Trailer TruckConfigurations\", M.A.Sc Thesis, University of British Columbia,December 1991.[3] Sovaran, G., Morel, T., and Mason, T. W. Jr., \"Aerodynamic DragMechanisms of Bluff Bodies and Road Vehicles\", Proceedings of theSymposium held at the General Motors Research Laboratories,Plenum press, New York, 1978.[4] Koernig-Facsenfeld, F.R., \"Aerodynamik des kraftfahrzeugs: Verlayder Motor-Rundschau\", Umshau Verlag, Frankfurt, West Germany,First Edition, 1951, Reprinted 1980.[5] Kramer, C., and Gerhardt, H. J., \"Road Vehicle Aerodynamics\",Proceedings of the 4th Colloquium on Industrial Aerodynamics,Aachen, June 1980.[6] Kurtz, D.W., \"Aerodynamic Design of Electric and Hybrid Vehicles:A Guidebook\", U. S. Department of Energy, Report No. 5030-471,September 1980.[7] Bearman, P.W., \"Review of Bluff Body Flows Applicable to VehicleAerodynamics\", Transactions of the ASME, Journal of FluidEn , Vol. 102, September 1980, pp 265-274.7374[8] Wacker, T., \"A Preliminary Study of Configuration Effects on theDrag of a Tractor-trailer Combination\", M.A.Sc. Thesis, Universityof British Columbia, Vancouver, Oct. 1985.[9] Tyrrell C. L., \"Aerodynamics and Fuel Economy - On Highwayexperience.\" Proceedings of the Truck and Bus Meeting andExposition, Dearborn, Michigan, November 16 - 19, 1987, Society ofAutomotive Engineers, No. 872278[10] Buckley, F. T. Jr, \"An Improved over the road test method forDetermining the Fuel Savings benefit of a truck aerodynamicdevice\", Proceedings of the International Congress and Exposition,Detroit, Michigan, February 25 - March 1, 1985, Society ofAutomotive Engineers, No. 850285[11] Saunders, J.W., Watkins, S., Hof man, P., and Buckley, F. T. Jr,\"Comparison of On-Road and Wind Tunnel Tests for Tractor TrailerAerodynamic Devices, and Fuel Savings Prediction.\" Proceedings ofthe International Congress and Exposition, Detroit, Michigan,February 25 - March 1, 1985, Society of Automotive Engineers, No.850286.[12] Watkins, S., Saunders, J.W., and Hoffman, P.H., \"Wind TunnelModelling of Commercial Vehicle Drag Reducing Devices: 3 CaseStudies\", 1987, Society of Automotive Engineers, No. 870717.[13] Eaker, G.,\"Wind Tunnel-to-Road Aerodynamic Drag Correlation\",Proceedings of the International Congress and Exposition, Detroit,Michigan, February 29 - March 4, 1988, Society of AutomotiveEngineers, No. 880250.[14] Remenda, B.A.P., Krause, A.E., and Hertz, P.B., \"CoastdownResistance Analysis Under Windy and Grade-Variable Conditions\",Proceedings of the International Congress and Exposition, Detroit,Michigan, February 27 - March 3, 1989, Society of AutomotiveEngineers, No. 890371.[15] Passmore, M.A., Jenkins, E.G., \"A Comparison of the Coastdownand Steady State Torque Methods of Estimating Vehicle DragForces\", Proceedings of the International Congress and Exposition,Detroit, Michigan, February 29 - March 4, 1988, Society ofAutomotive Engineers, No. 880475[16] Drollinger, R.A., \"Heavy Truck Aerodynamics\", Kenworth TruckCo., 1987, Society of Automotive Engineers, No. 870001.[17] Inagawa, M., Ohta, M., \"Contribution of Vehicular Parameters onFuel Economy of Trucks\", 1985, Society of Automotive Engineers,No. 852257[18] Goldstein, S, \"Modern Developments in Fluid Mechanics\", Vols. Iand II, Oxford University Press, 1938[19] Lachmann, G. V., \"Boundary-layer and Flow Control\", Vols. I andII, Pergamon Press, 1961[20] Rosenhead, L., \"Laminar Boundary-layers\", Oxford UniversityPress, 1966[21] Schlichting, H., \"Boundary-layer Theory\", McGraw Hill BookCompany, 196875[22] Chang, P. K., \"Separation of Flow\", Pergamon Press, 1970[23] Favre, A., \"Contribution a l'Etude Experimentale des MouvementsHydrodyamiques a Duex Dimensions\", Thesis presented to theUniversity of Paris, 1938.[24] Alvares-Calderon, A., and Arnold, F.R., \"A Study of theAerodynamic Characteristics of a High Lift Device Based onRotating Cylinder Flap\", Stanford University Technical ReportRCF-1, 1961.[25] Cichy, D.R., Harris, J.W., and MacJat, J.K., \"Flight Tests of aRotating Cylinder flap on a North American Rockwell YOV-10AAircraft\", NASA CR-2135, November 1972.[26] Weiberg, J.A., Giulianettij, D., Gambucci, B., and Innis, R.C.,\"Take-off and Landing Performance and Noise Characteristics of aDeflected STOL Airplane with Interconnected Propellers andRotating Cylinder Flaps\", NASA TM X-62, 320, December 1973.[27] Cook, W.L., Mickey, D.M., and Quigley, H.G., \"Aerodynamics of JetFlap and Rotating Cylinder Flap STOL Concepts,\" AGARD FluidDynamics Panel on V/STOL Aerodynamics, Delft, N \Netherlands,April 1974, Paper No. 10.[28] Johnson, W.S., Tennant, J.S., and Stamps, R.E., \"Leading EdgeRotating Cylinder for Boundary-layer Control on Lifting Surfaces\",Journal of Hydrodynamics, Vol 9, No. 2, April 1975, pp. 76-78.[29] Modi, V.J., Sun, J.L.C., Akutsu, T., Lake, P., McMillan, K., Swinton,P.G., and Mullins, D., \"Moving Surface Boundary-layer Control for76Aircraft Operation at High Incidence\", Journal of Aircraft, AIAA,Vol.18, No. 11, November 1981, pp. 963-968.[30] Mokhtarian, F., and Modi, V.J., \"Fluid Dynamics of Airfoil withMoving Surface Boundary-layer Control\", AIAA Atmospheric FlightMechanics Conference, August 1986, paper No. 86-2184-CP; alsoJournal of Aircraft, Vol 25, No. 2, February 1988, pp 163-169.[31] Mokhtarian, F., Modi V.J., and Yokomizo, T., \"Rotating Air Scoop asAirfoil Boundary-layer Control\", Journal of Aircraft, AIAA, Vol. 25No. 10 October 1988, pp. 973-975.[32] Mokhtarian, F., Modi V.J., and Yokomizo, T., \"Effect of MovingSurfaces on the Airfoil Boundary-layer Control\", AIAA AtmosphericFlight Mechanics Conference, Minneapolis, Minnesota, August1988, Paper No. AIAA-88-4303CP; also Proceedings of the Conference, Editors: R. Holdway and B.Kaufman, AIAA Publisher,pp. 660-668; also Journal of Aircraft, AIAA, Vol 27, No. 1, January1990, pp 44-60.77APPENDIX A : PROGRAM LISTINGSData Logging Program#include #include #include #include #include #include #include #include #include /* GLOBAL VARIABLES */float mass;/* CONSTANTS */const float rho =1.2;const float fuelrho= 0.00075;const int xmax=120, ymax=4000;char *comm=\"CE0A1,3 ;/* setport() is used to set up the port and initiatecommunications with the dam */setport(){char o[100];int p;bioscom(0,227,0);p=0;do{bioscom(1,0x0d,0);p++;o[p]=bioscom(2,NULL,0);}while(o[p]!=0x3e && p <200);}/* getdata() asks the dam for data when called */getdata(data)int data[20];{int q,p;char o[100],*r;p=0;for(q=0;q\");scanf(\"%s\",ans);if((fpraw=fopen(ans,\"a+t\"))==NULL){prinff(\"cannot open file\n\");exit(1);}prinff(\"input vehicle mass =>\");scanf(\"%f\",&mass);i=0; fuel=0; xpixel=-20; ypixel=-20;axis(xmax,ymax);outtextxy(100,10,\"Hit any key to start data logging\");getchar();getchar();setcolor(BLACK);outtextxy(100,10,\"Hit any key to start data logging\");do{do(if(i>(2001-120))printf(\"\a\");}while(biostime(0,0)<18);biostime(1,0);getdata(data);setcolor(BLACK);outtextxy(107,72,f1);outtextxy(107,87,ms);outtextxy(107,102,dp);velv=data[7]; fv=data[6]; fuel+=data[5]; accl=data[8];acc1=0; /* REMOVE WHEN ACCELEROMETER INSTALLED */totfuel = fuel*fuelrho;rawfuel[i] =data[5];rawtorque[i] =data[6];rawvel[i] =data[7];rawaccl[i] =data[8];time[i] =data[1]*3600.0+data[2]*60.0+data[3]+data[4]/1000.0;force[i]=fvical+foff-acct* (mass-totfuel);speed[i]=velv*velcal+veloff;box(xpixel,ypixel,BLACK);xpixel=speed[i]*580/xmax+30;ypixel=450-force[i]*420/ymax;box(xpixel,ypixel,WHITE);putpixel(xpixel,ypixel,WHITE);setcolor(WHITE);gcvt(fue1,6,f1);outtextxy(107,72,f1);gcvt(mass-toffue1,6,ms);outtextxy(107,87,ms);gcvt(i,5,dp);outtextxy(107,102,dp);i++;}while(!bioskey(1));outtextxy(100,10,\"Hit any key to store data\");getchar();getchar();for(j=0;j<=i;j++){fprintf(fpraw,\"%i^%f^%d^%d^%d^%d^%f^%f\n\",j,time[j],rawfuel[j],rawtorque[j],rawvel[j],rawaccl[j],speed[j],force[j]);}fprintf(fpraw,\"\n\n%d %f\",fuel,mass-totfuel);coefdet(force,speed,i);setcolor(BLACK);outtextxy(100,10,\"Hit any key to store data\");setcolor(WHITE);outtextxy(100,10,\"Hit any key to termintate this program\");getchar();getchar();fclose(fpraw);}Polynomial Regression Program#include \"stdio.h\"int size;float force, speed;float a[8],d,cd,cv,cr;int i,size;char fn[30];main(){FILE *fpin;clrscr();printf(\"\n\n^POLYREG : 2ND ORDER POLYNOMIAL REGRESSION\n\n\");printf(\"This program is designed to take velocity-force data and perform a\n\");printf(\"regression on it. The velocity data should be in km/hr as this program\n\");printf(\"will automatically convert it to m/s. This velocity conversion makes\n\");priptf(\"this program unsuitable for other curve fitting.\n\n\");printf(\"Input data file name =>\");81scanf(\"%s\",fn);if((fpin=fopen(fn,\"r\"))==NULL){printf(\"cannot open input file\n\");exit(1);}fscanf(fpin,\"%i\",&size);printf(\"\n\nThe data field contains %i points\n\n\n\",size);for(i=1;i<=7;i++) a[i]=0;for(i=1;i<=size;i++){fscanf(fpin,\"%f %f\",&speed,&force);a[1]=a[1]+speed/3.6;a[2]=a[2]+speed*speed/12.96;a[3]=a[3]+speed*speed*speed/46.656;a[4]=a[4]+speed*speed*speed*speed/167.9616;a[5]=a[5]+force;a[6]=a[6]+force*speed/3.6;a[7]=a[7]+force*speed*speed/12.96;)d= size*(a[2]*a[4]-a[3]*a[3])-a[1]*(a[1]*a[4]-a[2]*a[3])+a[2]*(apra[3]-a[2]*a[2]);cr=(a[5]*(a[2]* a[4]-a[3]* a[3])-a[1]*(a[6]*a[4]-a[7]*a[3])+a[2]* (a[6]*a[3]-a[7]* a[2]))/d;cv=(size*(a[6]*a[4]-a[7]*a[3])-a[5]*(a[1]*a[4]-a[2]* a[3])+a[2]*(a[1]*a[7]-a[2]*a[6]))/d;cd=(size*(a[2]*a[7]-a[3]* a[6])-a[1]*(a[1]*a[7]-a[2]*a[6])+a[5]* (a[1]*a[3]-a[2]*a[2]))/d;printf(\"The result is of the form: R + k.v + D.v^2, where\n\n\");prinff(\"R = %15.5f ; k = %15.5f ; D = %15.5f\n\",cr,cv,cd);printf(\"\nTherefore at 30 m/s F = \u00C2\u00B0/08.0f\n\",cr+30*cv+900*cd);}Discretization Program#include #include float a, vs[131][2], fcal, foff, velcal, veloff;int b,vel;int i,size, v[131][2];char fn[30];main(){FILE *fpin, *fpout, *fpcal;clrscr();printf(\"\n\n^DSCRT2CL : AVERAGING FORCES AT DISCRETE SPEEDS\n\n\n\");prinff(\"Loading Calibraiton data\n\");if((fpcal=fopen(\"C:\\TC\\CALDAT \",\"r\"))==NULL){printf(\"cannot open calibration file\n\");exit(1);)fscanf(fpcal,\"%f %f %f %f\",&fcal,&foff,&velcal,&veloff);fclose(fpcal);printrnInput data file name =>\");scanf(\"%s\",fn);if((fpin=fopen(fn,\"r\"))==NULL){printf(\"cannot open input file\n\");exit(1);}prinff(\"\nInput sorted file name =>\");scanf(\"%s\",fn);if((fpout=fopen(fn,\"w\"))==NULL){printf(\"cannot open sorted file\n\");exit(1);)fscanf(fpin,\"%i\",&size);prinff(\"\n\nThe data field contains %i points\n\n\n\",size);82for(i=1;k131;i++){v[i][0]=0; v[i][1]=0; vs[i][0]=0; vs[i][1]=0;)for(i=1;i<=size;i++){fscanf(fpin,\"%i %i\",&b,&vel);v[vel][0]+=b;v[vel][1]+=1;)for(i=1;i<131;i++){if(v[i][1]!=0)vs[i][0]=1.0*v[i][0]/v[i][1];)rewind(fpin);printf(\"\nmeans calculated \n\");fscanf(fpin,\"%i\",&b);for(i=1;i<=size;i++){fscanf(fpin, Kyo, / &b,&vel);vs[vel][1]+=(b-vs[vel][0])*(b-vs[vel][0]);}for(i=1;k131;i++){if(v[i][1]!=0)vs[i][1]=sqrt(vs[i][1]/v[i][1]);vs[i][0]=vs[i][0]*fcal+foff;vs[i][1j=vs[i][1]*fcal;a=i*velcal+veloff;if(v[i][1]!=0)fprinff(fpout,\"%f %f %f \u00C2\u00B0/0i\n\",a,vs[i][0],vs[i][1],v[i][1]);)prinff(\"processing complete\n\");}Time Averaging Program#include #include #include float vavg,favg,fcal, foff, velcal, veloff;int vel, fr, sz, flag;int i, dcount,pcount, ocount,size;char fn[30];main(){FILE *fpin, *fpout, *fpcal;clrscr();printf(\"\n\n^TAVG : TIME AVERAGEING OF FORCE DATA\n\n\n\");prinff(\"Loading Calibraiton data\n\");if((fpcal=fopen(\"C:\\TC\\CALDAT \",\"r\"))==NULL){prinff(\"cannot open calibration file\n\");exit(1);}fscanf(fpcal,\"%f %f %f %f\",&fcal,&foff,&velcal,&veloff);fclose(fpcal);prinff(\"\nInput data file name =>\");scanf(\"%s\",fn);if((fpin=fopen(fn,\"r\"))\u00E2\u0080\u0094NULL){prinff(\"cannot open input file\n\");exit(1);}prinff(\"\nInput sorted file name =>\");scanf(\"%s\",fn);if((fpout=fopen(fn,\"w\"))==NULL){printf(\"cannot open sorted file\n\");exit(1);)fscanf(fpin,\"%i\",&size);prinff(\"\n\nThe data field contains %i points\n\",size);prinff(\"\nHow many data points per average =>\");scanf(\"\u00C2\u00B0/01\",&sz);83dcount=0; ocount=0;do{i=0;do{if(flag==0){fscanf(fpin,\"%i %i\",&fr,&vel);dcount++;}flag=0; i++;if(i> 1)(if(abs(vel-vavg/(i-1))<=4)(vavg+=vel; favg+=fr;}else {flag=1; i--;})else { vavg=vel ; favg = fr;})while(flag==0 && i< sz && dcount < size);printf(\".\"); ocount++;vavg=vavg/rvelcal+veloff;favg=favg/Ncal+foff;fprintf(fpout,\"%f %f \u00C2\u00B0/01\n\",vavg,favg,i);}while(dcount < size);prinff(\"\n\aTotal number of points generated = %i\n\n\",ocount);fclose(fpin); fclose(fpout);}Data Cleaning Program#include char ans[30];int i,j, a[5], point1 , point2;float b[3];main() {FILE *fpin, *fpout;clrscr();prinff(\"\n\n^CLEAN : REMOVING BAD DATA POINTS FROM DATA FILES\n\n\");prinff(\"Output file name =>\");scanf(\"%s\",ans);if((fpout=fopen(ans,\"a+t\"))==NULL)(prinff(\"Cannot open Output file\n\");exit(1);}1=1 ;point1=1 ;do{printf(\"\nlnput File #%i Name =>\",i);scanf(\"%s\",ans);iffflpin=fopen(ans,\"r\"))==NULL)(printf(\"Cannot open Input File #%i\n\",i);exit(1);}do{rewind(fpin);printf(\"\nStart Point =>\");scanf(\"%i\",&point1);printf(\"\nEnd Point =>\");scanf(\"/oi\",&point2);printf(\"^ \n\n\");if(point1>0)(for(j=0;j0);i++;fclose(fpin);}while(point1>-1);}Force Removal, and Regression Program#include #include #include float vavg,favg,fcal, foff, velcal, veloff, r, k;float v2sum, fsum;int vel, fr, sz, flag, regcount;int i, dcount,pcount, ocount,size;char fn[30];main(){FILE *fpin, *fpout, *fpcal, *fpoffset ;cIrscr();prinff(\"\n\nTAVGRM : TIME AVERAGEING OF FORCE DATA, REMOVING NON AEROFORCES\n\n\n\");prinff(\"Loading Calibraiton data from: c:\\project\\progs\\cal.dat\n \");if((fpcal=fopen(\"C:\\project\\progs\\cal.dat\",\"r\"))==NULL)(printf(\"cannot^open^calibrationfile\n\");exit(1);}fscanf(fpcal,\"%f %f %f %f\",&fcal,&foff,&velcal,&veloff);fclose(fpcal);prinff(\"\nInput data file name =>\");scanf(\"%s\",fn);if((fpin=fopen(fn,\"r\"))==NULL)(printf(\"cannot open input file\n\");exit(1);}printf(\"\nlnput sorted file name =>\");scanf(\"%s\",fn);if((fpout=fopen(fn,\"w\"))==NULL)fprintf(\"cannot open sorted file\n\");exit(1);}fscanf(fpin,\"%i\",&size);if((fpoffset=fopen(\"C:\\project\\progs\\offset.dat \",\"r\"))==NULL){prinff(\"cannot^open^offsetfile\n\");exit(1);}prinr\nLoading Offset data from : c:\\project\\progs\\offset.dat\n \");fscanf(fpoffset,\"%f %f\",&r,&k);printf(\"\nAverage Rolling Resistance = %f\n\",r);prinff(\"Average Brake and Internal Viscous Drag = %f\n\",k);printf(\"\nThe data field contains %i points, it is averaged over 250 point segments\n\n\",size);sz=250;dcount=0; ocount=0; regcount=0;do{i=0;do{if(flag==0){fscanf(fpin,\"\u00C2\u00B0/0i %i\",&fr,&vel);dcount++;}flag=0; i++;if(i> 1)(if(abs(vel-vavg/(i-1))<=4){vavg+=vel; favg+=fr;}else {flag=1; i--;}}else { vavg=vel ; favg = fr;})while(flag==0 && i< sz && dcount < size);printf(\" (%3i)\",i); ocount++;vavg=vavg/i*velcal+veloff;favg=(favg/ilcal+foff)-r-vavg*k/3.6;if(i>100){fprintf(fpout,\"%f^\u00C2\u00B0/0i\n\",vavg,favg,i);fsum+=favg; v2sum+=vavg*vavg/12.96; regcount++;prinff(\"x\");}else printf(\" \");if(ocount/1 0==ocount/1 .0)printf(\" \ n\");}while(dcount < size);prinff(\"\n\aTotal number of points generated = %i\n\n\",ocount);printf(\"The gradient m = %7.5f \n\",fsum/v2sum);printf(\"The number of points used =^\n\",regcount);fclose(fpin); fclose(fpout);}Time Averaging of Fuel Data Program#include #include #include float vavg,favg,fcal, Toff, velcal, veloff, r, k;float fsum,fc;int vel, fr, sz, flag, regcount;int i, dcount,pcount, ocount,size;char fn[30j;main(){FILE *fpin, *fpout, *fpcal;cIrscr();printf(\"\n\nTAVGRMF : TIME AVERAGEING OF FUEL CONSUMPTION DATA\n\n\n\");printf ^Calibraiton data from: c:\\project\progs\\cal.dat\n \");if((fpcal=fopen(\"C:\\project\\progs\\cal.dat \",\"r\"))==NULL){printf(\"cannot^open^calibrationfile\n\");exit(1);}fscanf(fpcal,\"%f %f %f cYof\",&fcal,&foff,&velcal,&veloff);fclose(fpcal);printf(\"\nlnput data file name =>\");scanf(\"%s\",fn);if((fpin=fopen(fn,\"r\"))==NULL){printf(\"cannot open input file\n\");exit(1);}printf(\"\nlnput sorted file name =>\");86scanf(\"%s\",fn);if((fpout=fopen(fn ,\"w\"))==N ULL){printf(\"cannot open sorted file\n\");exit(1);)fscanf(fpin,\"%i\",&size);printf(\"\nThe data field contains %i points, it is averged over 250 point segments\n\n\",size);sz=250;dcount=0; ocount=0; regcount=0;do{i=0;do{if(flag==0){fscanf(fpin,\"%i %i\",&fr,&vel);dcount++;}flag=0; i++;if(i> 1 ){if(abs(vel-vavg/(i- 1 ))<=4){vavg+=vel; favg+=fr;}else {flag=l; i--;}}else { vavg=vel ; favg = fr;}}while(flag==0 && i< sz && dcount < size);ocount++;vavg=vavg/i*ve lcal+ve loff ;favg=f avg/i; fc=vavg/(3.6*favg);if(i>100){fprintf(fpout,\"%15.3f %15.3f %15.3f %4i\n\",vavg,favg,fc,i);printf(\"%15.3f %15.3f %15.3f %4i\n\",vavg,favg,fc,i);regcount++;}}while(dcount < size);printf(\"\n\aTotal number of points generated = %i\n\n\",ocount);printf(\"The number of points used = %i \n\",regcount);fclose(fpin); fclose(fpout) ;}87APPENDIX B : FORCE REGRESSION PLOTSThe plots presented in this appendix are identified in the followingmanner. The letter refers to this appendix, and the digit to theconfiguration tested. The test numbers are indicated on the diagram.880.10-0.08-C)2 0.06-U-0.04-0.02-0.12Test 2......... 41p-0.000 20^40^60^80^1 00Velocity [km/hr]120Figure B-1 Regression analysis of the force data for the modified truck with extended roof(baseline or reference case).)!i0.12Test 4+ 7.0.10-0.08-0)2 0.06-u_0.04-0.02-.........0.000 40^do^80^100Velocity [km/hr]120Figure B-1 Regression analysis of the force data for the modified truck with extended roof(baseline or reference case).0.10- +0.12Test 54.0.08-0)2 0.06-U-.0.\u00E2\u0080\u00A2 'it,---- ' - -+.....0.04- .... \" ...+ .. ' '^\u00C2\u00B0IN....\".... - ++0.02-0.000 irJ^4o^do^io^100Velocity [km/hr]120Figure B-1 Regression analysis of the force data for the modified truck with extended roof(baseline or reference case).0.12Test 74s0.10-0.08-2 0.06-u-0.04-404s4040 ses4,^-sis0.02 -----------20^40^60^80^100^120Velocity [km/h r]Figure B-1 Regression analysis of the force data for the modified truck with extended roof(baseline or reference case).0.0000.12Test 8400.10-0.08-0)2 0.06-u_0.04-\u00E2\u0080\u009E4: -.--is .-'Aliis.-'.\u00E2\u0080\u00A2\".0.02-is.---.-3. - ' \u00E2\u0080\u0094 - - 4'... \u00E2\u0080\u0094 414\u00E2\u0080\u00A20.000 20^40^60^80^lOoVelocity [km/hr]120Figure B-1 Regression analysis of the force data for the modified truck with extended roof(baseline or reference case).0.08-410)2 0.06-U- NIN0.04-0.02-0.12Test 90.10- al\u00E2\u0080\u00A245----------0.000 20^40^60^80^160Velocity [km/hr]120Figure B-1 Regression analysis of the force data for the modified truck with extended roof(baseline or reference case).Figure B-1 Regression analysis of the force data for the modified truck with extended roof(baseline or reference case).C0)2 6%LLsissis12%Test 1110%-4080/0-4140/0-.\u00E2\u0080\u00A2..\"'2%-4s4s^ 4s-16----------0%0 20^40^60^80^1 00Velocity [km/hr]120\u00E2\u0096\u00A0Figure B-1 Regression analysis of the force data for the modified truck with extended roof(baseline or reference case).0.12Test 120.10-0.08-zC)0.06-U-0.04-4\u00E2\u0080\u00A2410.02------------4;^410.000 40^60^80^1 00Velocity [km/hr]120Figure B-1 Regression analysis of the force data for the modified truck with extended roof(baseline or reference case).^1-L012%Test 24010%-4.0)6\u00C2\u00B0/0-U-4%40\u00E2\u0080\u00A2\u00E2\u0080\u00A22%-----------4o^do^80^1 00^120Velocity [km/hr]Figure B-2 Regression analysis of the force data with the horizontal fence .+00/000)6\u00C2\u00B0/0-U-4P/0-80/0-12%Test 310%-.42%-..^....------- 4,0^20^40^60^80^100^120Velocity [km/hr]Figure B-2 Regression analysis of the force data with the horizontal fence,10%-4s8%-402 60/0-u_ 4s4%- sls12%Test 4402%----------- -0%0 0^4o^do^80^1 00Velocity [km/hr]120Figure B-2 Regression analysis of the force data with the horizontal fence.{^,10%-+80/0- 41../0)2 6cro-a_12%Test 6^ 4s40/0-\u00E2\u0080\u009EFP--'...--AP --^4s..... Ale ^+...-\".2%- ...-^, ^\u00E2\u0080\u0094 \"Si.^---------- \u00E2\u0080\u0094 44^sisirJ^4o^do^80^1 00^120Velocity [km/hr]Figure B-2 Regression analysis of the force data with the horizontal fence.C0\u00C2\u00B0/0040/ -0 +12%Test 710%-80/0-402%-00/00 4o^do^1 00Velocity [km/hr]120Figure B-2 Regression analysis of the force data with the horizontal fence.atr/0\u00E2\u0080\u009412%Test 84010%\u00E2\u0080\u00948\u00C2\u00B0/0+2%\u00E2\u0080\u0094 AIL -----------0\u00C2\u00B0/00 20^40^60^80^100Velocity [km/h r]120Figure B-2 Regression analysis of the force data with the horizontal fence,12%Test 1i\u00E2\u0080\u00A210%-80/0-cm2 6%-u_4\u00C2\u00B0/0-419'' 4't.. \u00E2\u0080\u00A2 '.'\u00E2\u0080\u00A2'.....\"4,,+2%----------- . -IC - - 410\u00C2\u00B0/00 0^40^60^80^100Velocity [km/hr]120Figure B-3 Analysis of the force data obtained with one horizontal and two vertical fencesinstalled on the cube-truck during the road tests .12%Test 210%-80/0-4\u00C2\u00B0/0-2%-----------^------20^40^60^80^lOoVelocity [km/hr]0\u00C2\u00B0/00 120Figure B-3 Analysis of the force data obtained with one horizontal and two vertical fencesinstalled on the cube-truck during the road tests .<12%Test 310%-80/0-.52)da EM-4.4c)/0-2%-......... -4F- ......\"Pr!I\u00E2\u0080\u00A2 4100/00 20^40^60^80^100Velocity [km/hr] 120Figure B-3 Analysis of the force data obtained with one horizontal and two vertical fencesinstalled on the cube-truck during the road tests12%Test 410%-80/0-C)2 6%-u..4\u00C2\u00B0/0--\"2%-----------40-0%0 20^40^60^80^1 00Velocity [km/hr]120Figure B-3 Analysis of the force data obtained with one horizontal and two vertical fencesinstalled on the cube-truck during the road tests.0 t C4APPENDIX C : FUEL CONSUMPTION PLOTS113e 1 <-V'e ( 48 ^7-t 6-o 5-E 4-0 3-2-=u_1 -Test 701040U,404\u00E2\u0080\u00A2 40404040404020^Lib^do^80Velocity [km/hr]100^120Figure C-1Fuel consumption data for the reference (baseline) configuration.^8 ^7-6-co 5-E 4-no 3-a) 2-m1 -0^0Test 8404020^40^60^80^100Velocity [km/h r]120Figure C-1Fuel consumption data for the reference (baseline) configuration\u00E2\u0080\u00A21 \u00E2\u0080\u00A2+*tk i \u00E2\u0080\u00A2 40+++E 4-=u)co 3-0DC D 2-u_itil\u00E2\u0080\u00A2IN++Test 9..! i (78 ^1201 -0^ 0 Lio^60^do^100Velocity [km/hr]Figure C-1Fuel consumption data for the reference (baseline) configuration ,v C8 ^7-6-co 5-Eo 3- 2-U-1 -Test 10404m10 20 30 40 50 60 70 80 90 100 110Velocity [km/h r]Figure C-1Fuel consumption data for the reference (baseline) configuration8^7-E 6-o 5-0.--E 4-mo 3-4:1)= 2-u_1 -Test 114140^4,404104040o o 20^40^60^80^160Velocity [km/hr]120Figure C-1Fuel consumption data for the reference (baseline) configuration.t8 ^7-6-o 5-E 4-D(/)0 3-a) 2-LL1 -Test 12sis4s4s40slsses 4140sissis0^20^40^60^80^100^120Velocity [km/hr]Figure C-1Fuel consumption data for the reference (baseline) configuration ,^8^7-6-c0 5-E 4-=o 3-(1) 2-1 -0^0Test 1\u00E2\u0080\u00A24P4ssis404o^do^Eio^100Velocity [km/hr]11.120Figure C-2 Fuel consumption data with horizontal fence .^8^7-E 6-_.c0 5-E 4-mcoc0 3-00 2-=ii.1 -^0^0 Test 2\u00E2\u0080\u00A2*4.llssles^IN++al\u00E2\u0080\u00A2+4*is+ 404.4o^60^60^100Velocity [km/hr]120Figure C-2 Fuel consumption data with horizontal fence .fr\-44-8 ^Test 37-E 6-05-(I)o 3-on 2-U-1 -0^0\u00E2\u0080\u00A245IN4o^do^go^100Velocity [km/h r]120Figure C-2 Fuel consumption data with horizontal fence .1-0^0 20^4o^do^go^100Velocity [km/hr]1208 ^7-6-o 5-Test 4\u00E2\u0080\u00A240 43E 4_o 3-a) 2-LL=1*Figure C-2 Fuel consumption data with horizontal fence .Test 57-E 6- 4140434,el\u00E2\u0080\u00A2NINsio0^20^40^60^80^160Velocity [km/hr]o 5-4=0_E 4-0 3-a) 2-U-=1 -0^45120,4`8 ^Figure C-2 Fuel consumption data with horizontal fence .8 ^Test 67-.t 6-0o 5-laE 4-mwco 3-U(L) 2-=u_1 -0^20^40^60^80^100Velocity [km/h r]42il\u00E2\u0080\u00A2424242424242^ 414. 42 4.42424242424242^ *42120Figure C-2 Fuel consumption data with horizontal fence .87-le 6-0o 5-\" CaE 4-mwco 3-U(Dm 2-u_ 1 -Test 74\u00E2\u0080\u00A240NIN 4.+4.slo 4. ++ +4.\u00E2\u0080\u00A2100^120oo^40^60^goVelocity [km/hr]Figure C-2 Fuel consumption data with horizontal fence .87-6-c0 5-E 4-D0 3-con 2-LL-Test 842\u00E2\u0080\u00A2\u00E2\u0080\u00A2\u00E2\u0080\u00A24.eleels20^40^60^80^100^120Velocity [km/hr]Figure C-2 Fuel consumption data with horizontal fence,1 ood20^40^do^80^100Velocity [km/hr]87-E 6-0 5-E 4-=o 3-75 2-=1 -0^0Test 144.4045^ 404.el\u00E2\u0080\u00A2120Figure C-3 Fuel consumption as affected by the truck speed in the presence of one horizontaland two vertical fences8^7-C 6-c05-E 4-(I)co 3-0cl)m 2-U-1 -Test 2+*+4.^I.4.4.20^40^60^80^1 00Velocity [km/h r]oo4, 4. IN*el\u00E2\u0080\u00A2++120Figure C-3 Fuel consumption as affected by the truck speed in the presence of one horizontaland two vertical fences.c,8 ^7-6-c05-E 4-C')=o 3-CD 2-m1 -Test 3sis^ ses^ 4sses^sis40 sis4s404s 4s404s sisissis4soo 20^40^60^80^100Velocity [km/hr]120Figure C-3 Fuel consumption as affected by the truck speed in the presence of one horizontaland two vertical fences .^8^7-E 6-o 5-E 4-ncno 3-a) 2-IL.1 -^0^0 20^40^60^80^160Velocity [km/hr]Figure C-3 Fuel consumption as affected by the truck speed in the presence of one horizontalTest 4uelele^ Alm4, 4140120and two vertical fences ."@en . "Thesis/Dissertation"@en . "1993-05"@en . "10.14288/1.0080872"@en . "eng"@en . "Mechanical Engineering"@en . "Vancouver : University of British Columbia Library"@en . "University of British Columbia"@en . "For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use."@en . "Graduate"@en . "Drag reduction of cube-van through boundary-layer control: wind tunnel experiments and prototype road tests"@en . "Text"@en . "http://hdl.handle.net/2429/2358"@en .