Licensed from the SAE Digital Library Copyright 2010 SAE International
E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM
Using a New Driveline Model to Define ResearchEngine Operating Conditions
G. P. McTaggart-Cowan and E.J. Pallett
2010-01-0002
Published04/12/2010
Wolfson School of Mechanical and Manufacturing Engineering, Loughborough University, Loughborough UK
Copyright © 2010 SAE International
ABSTRACT
Steady state engine dynamometer testing provides the highestlevel of detail for understanding fundamental enginecombustion. It can provide insight into pollutant formationmechanisms and methods for minimizing fuel consumption.However, steady-state dynamometer tests are normallycarried out at test conditions far removed from the actualconditions that a vehicle engine encounters. This paperdescribes the application of a simple powertrain model todefine steady-state engine test conditions that are morerepresentative of real-world engine operation.
The model uses a backward-facing, modular structure. Themodel is validated against two powertrain configurations: aconventional powertrain equipped with a continuouslyvariable transmission (CVT) and a parallel hybrid powertrain.Powertrain parameters and performance data for validationfor both cases are supplied from the literature. The model isshown to agree well with both sets of published experimentalresults. These two models, along with a conventionalpowertrain equipped with a manual gearbox, are then used toevaluate steady-state and transient engine operation forvehicles following specified drive cycles. Steady-state testconditions are identified for all three models on the basis ofboth time and fuel consumption. The model resultsdemonstrate that engine transients are typically on the orderof 2-4 second duration and involve relatively modest changesin engine speed and torque.
test-bed differ significantly from the conditions encounteredin a vehicle. This is especially true for steady-state enginetesting, which while providing the deepest level ofunderstanding of the engine and combustion processes,differs significantly from in-use engine conditions.Powertrain and vehicle architecture and the drive-cycle bothsignificantly influence the operating conditions of an in-service engine. This paper describes the development of asimplified modular vehicle powertrain model and its use indefining steady-state engine test conditions for a variety ofvehicle architectures and over different drive-cycles.Powertrain models are widely used in vehicle research. Manynumerical simulation models have been developed andproposed in the literature to assist the development ofpowertrains with a particular set of performance, emissionsand economy targets [1]. The models range from simpleobject orientated simulation codes that use simplified modelsof the relevant physical processes in each of the maincomponents in a conventional powertrain, through tocomplex fluid dynamic models which can, in principle, offera detailed prediction of engine performance and emissions.These models differ substantially in terms of computationalrequirements and required input data.
Previous powertrain models have been developed to runtransient simulations; these include models based onMATLAB/Simulink [2,3]; such models are often focused onmaking use of MATLAB's advanced controls capabilities.Common to these models is the use of a modular structure,which allows easy variation of powertrain geometries.AMESim (Advanced modelling environment for performingsimulation of Engineering Systems) has also been used tomodel vehicle powertrains using a similar modular structure[4,5]. Parametric models have also been developed forspecific powertrain geometries [6]. Dedicated commerciallyavailable codes also provide powertrain simulations,
INTRODUCTION
Dynamometer-based engine testing is central tounderstanding fundamental engine combustion processes anddeveloping advanced engines that minimize fuel consumptionand emissions. However, it is limited by the fact that theoperating conditions encountered on an engine-dynamometer
Author:Gilligan-SID:13182-GUID:30506138-129.94.59.243Licensed to University of New South Wales
Licensed from the SAE Digital Library Copyright 2010 SAE International
E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM
including GT-Drive [1] and AVL ADVISOR. The latter is acommercialisation of the ADVISOR model developed at theNational Renewable Energy Laboratory (NREL) [7]; this hasbeen used extensively to evaluate various powertrainarchitectures for both light-duty [8,9] and heavy-dutyvehicles [10]. These previously-developed models all havebenefits and drawbacks. However, no model was identifiedthat met the key criteria for the current project:1. Accessible at low cost
2. Easily adaptable to the amount of information availablefor any given component
3. Flexible, to model various vehicle powertrains withminimum modification
4. Focused on defining engine test conditions rather thanevaluating performance.
As a result, it was decided to develop a new modelling tool tomeet these requirements. The first half of this paper providesan introduction to the modelling structure used and the keymodules developed. The second half of the paper describesthe application of the model to define engine test conditionsfor various powertrain architectures and vehicle drive cycles.
DRIVE CYCLE
The drive cycle used to evaluate a vehicle has a significantimpact on its fuel consumption and emissions. This work willfocus on two common drive-cycles: the first 505 seconds ofthe EPA urban dynamometer driving schedule (UDDS) and asingle repetition of both parts 1 and 2 of the new Europeandrive cycle (NEDC), with 195 seconds of the elementaryurban cycle and the 400 seconds of the extra-urban cycle(referred to as NEDC(reduced)). Both drive cycle profiles areshown in Figure 1. More information on these drive cycles,and the speed/time traces for both, can be found on the EPAwebsite [11]. The first 505 s of the UDDS drive-cycle wasused as it is the test-cycle from previous work that is used forone of the validation cases here. The second validation caseused the full UDDS, which involves the speed-time traceshown in Figure 1 followed by a further 900 s of lower-speedoperation; however, the general profile of this second phase isless demanding for the engine than the first phase. TheNEDC(reduced) was selected for comparison as it featuressignificantly different conditions, with slower and steadieracceleration events but higher peak speeds. It should be notedthat whether either of these cycles represent the drivingprofile of an in-use vehicle is questionable.
Author:Gilligan-SID:13Figure 1. Driving cycles used in this work. Top - velocityprofile for the NEDC(reduced) and first 505 seconds ofthe UDDS. Bottom - principal acceleration events in the
first 505 seconds of the UDDS.The velocity profiles in the various drive-cycles are based ona series of transient events. In developing engine testconditions representative of these events, it is important toevaluate the differences between the transients. The sixprincipal acceleration transients (velocity changes > 10 m/s)in the UDDS(505) are also shown in Figure 1. The generalslope of the transients are all the same; the main differencesare the maximum speed reached and (for the 187 - 207 sacceleration) the starting velocity. From an engine test pointof view, this suggests that conditions derived for one of thesetransients should be generally representative of the otherswithin the same test cycle. The NEDC (not shown) hassimilarly consistent acceleration profiles; however, there aresignificant differences in terms of acceleration rate andduration between the two drive cycles.
PART I - MODEL OVERVIEW ANDVALIDATION
The model developed for this project uses a modularstructure. This facilitates the creation of a model with a largescope, allowing for simple representations of individualcomponents to be used, or more complex processes to bemodelled if required. This also allows complex componentsto be split up into distinct modules if more detail is required.The model has been developed in Microsoft Visual Basic,with an MS ExcelTM interface.
The model uses a backward facing modelling approach. Theprocess begins with a required drive cycle (vehicle speed vs.time, as discussed above) for the modelled vehicle to follow.Environmental effects (wind speed and orientation, grade)
0506138-129.94.59.243182-GUID:3Licensed to University of New South Wales
Licensed from the SAE Digital Library Copyright 2010 SAE International
E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM
and vehicle parameters (mass, aerodynamic drag, rollingresistance) are used to calculate the total retarding force. Thisvalue, combined with the acceleration required to follow thedrive cycle, is then used to calculate the required wheeltorque. The angular velocity of the wheels is also calculatedbased on the drive cycle and wheel diameter. The calculatedwheel torque and angular velocity are then passed to the firstpowertrain component.
The overall structure of the model is based on a series ofmodules, each representative of a single powertraincomponent. The modules are developed such that each isindependent but, they all have common inputs and outputs.As such, any desired architecture can be developed by simplyadjusting the order of the modules. A new component can bedeveloped by creating a new module and adding it to thelibrary of existing modules, while an existing module can bemodified without requiring any changes to any othercomponent.
The values passed between modules are torque and angularvelocity. These values are calculated for each module inseries and the required ‘input’ values are passed to the nextmodule. That is, for each physical powertrain component, therequired torque and speed inputs are calculated based on therequirements of the previous component (component closerto the wheels), combined with the performance of the specificcomponent of interest.
In a conventional powertrain the modules simulate the finaldrive, transmission, clutch and engine. Torques and speedsare calculated for each of these components; an outline of thestructure is shown in Figure 2. The final output from themodel in this case is engine torque and speed. Engine fuelconsumption maps allow the engine torque and speed to beconverted into fuel consumption values.
Author:Gilligan-SID:13Figure 2. Overall powertrain model structure.The torque and speed for each component of the entirepowertrain is calculated for each time step in the drive cycle.Once the performance at a specific time point has beencalculated, the model advances to the next time step andrepeats the process. Error-checking is used in individualmodules and for the overall model to ensure that nocomponents are exceeding any stated performance limits. Incases where performance limits are exceeded (e.g., the enginetorque exceeds full load), the time-step is repeated with theperformance of all other modules limited by the identifiedfactor. This may result in the model not being able to follow aprescribed drive-cycle. In such cases, each time-step istreated individually, such that the vehicle model follows thedrive-cycle as closely as possible. Such a divergence from thedesired drive-cycle is a limitation of the powertrain beingmodelled, rather than a feature of the model itself.
Individual ModulesThe core of the powertrain model is the series of individualmodules that represent the various physical components, asdemonstrated in Figure 2. The first two modules are the drivecycle, which defines the vehicle speed as a function of time,and the environment. These two modules combine todetermine the force that must be exerted at the wheels tofollow the prescribed drive-cycle. These modules are alwaysthe first two called, independent of what powertrain structureis being investigated. The first module in the sequence mustbe defined such that it converts the vehicle force input into anangular velocity and torque.
Each of the remaining modules is defined with common inputand output parameters: time, angular velocity and torque. Asthe model is backwards-facing, the ‘input’ to each module is
0506138-129.94.59.243182-GUID:3Licensed to University of New South Wales
Licensed from the SAE Digital Library Copyright 2010 SAE International
E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM
from the preceding module (e.g. the input to the gearboxwould be the torque and speed required at the driveshaft,while the output would be the torque and speed required to betransferred to the gearbox to provide that output). Thecomplexity of the individual modules is unlimited, as long asthe input and output parameters are consistent. For this work,the individual modules were constructed so that torquetransfer efficiency and moments of inertia were accountedfor. Angular velocity (ω) was calculated using:
(1)
where GR is gear ratio across the module and in and out referto the speed physically input to the component (i.e., theunknown output from the module) and the speed out of thecomponent (i.e., the speed calculated from the precedingmodule in the model), respectively. Similarly, the torque (T)at each component was calculated from:
(2)
where η is the efficiency of the component, Icomp is itsmoment of inertia, α the angular acceleration and in and outhave the same definitions as described for Eq. 1. Thisequation applies for positive torque transfer (i.e., torquetransfer from the physical input to the physical output, asidentified by the arrows in Figure 2). In the case of negativetorque, such as encountered during regenerative braking orwhen driving a generator, equation (2) was replaced with:(3)
where the various terms have the same meanings as in Eq. 2.In and out still refer to the unknown output from and knowninput to the module, respectively (i.e., they are now oppositeto the direction of torque transfer).
Most of the powertrain modules follow this same basicformat. However, added complexity occurs in cases wherecomponent parameters are a function of physical input (i.e.,unknown) conditions, such as in the case of the continuouslyvariable transmission (CVT) module used in Validation Case1. In this case, the efficiency depends on the torque and speedinput from the engine. To overcome this, an iterative processis used: an efficiency is assumed and the torque input to theCVT is then calculated using Eq. 2. From this calculatedvalue, the efficiency is re-estimated using a look-up tablebased on input conditions and the input torque is recalculated.This process repeats until the results converge.
The model presented in Figure 2 relates only to aconventional series-organized drive-train. Parallel drive-trains (such as the hybrid vehicle discussed in Validation
Author:Gilligan-SID:13Case 2) involve torque being divided between two separate‘arms’, which may be operating at different rotational speeds.This split can be managed by developing a new module thatdetermines how much torque should be being delivered to orfrom each arm of the driveline. A specific example of thetorque transfer between two drivetrain arms is presented inmore detail in Validation Case 2. Beyond this module, thetwo arms are treated independently.
For the model results reported here, the final output is fuelconsumption. This was determined from an engine torque-speed fuel consumption map. Once the engine speed andtorque were calculated for each time-step, a two-way linearinterpolation was used to determine fuel consumption at thoseconditions from the externally-prepared lookup table. Thefuel consumptions at each time point were then summed togive the drive-cycle fuel consumption. Similar methods couldbe used to evaluate other parameters, such as criteriapollutant emissions, if sufficient data was available.
VALIDATION
The model structure was validated using two publishedpowertrain architectures: (1) a conventional gasoline-enginevehicle equipped with a CVT transmission [8], and (2) aparallel hybrid vehicle equipped with a manual transmission[9]. These two architectures were selected as they providedenough test data and powertrain parameter information tovalidate the model. They also provide further insight into themodel and the design and performance of individual modules.The powertrain parameters used are given in the appendix.Validation Case 1 - CVTThe first validation case involves a 2001 Toyota Opaequipped with a direct injection gasoline engine and a CVT.Experimental results included fuel flow rate and enginetorque and speed; powertrain component efficiencies and anengine fuel consumption map were also provided [8]. Themodel structure for this case is as shown in Figure 2, with thefollowing powertrain components: driveline; CVT; clutch;engine. The results in [8] were acquired over the ‘hot-505’cycle, which is equivalent to the UDDS(505) discussedabove.
One feature of this vehicle is the CVT. It operates tomaximize the time the engine is at its peak efficiency point(low speed, moderate load). The module developed torepresent the CVT used an algorithm where the CVT held theengine speed at a low level, resulting in increased torque athigher power, until the line of best efficiency was reached. Aspower demand increased further, the gear ratio was adjustedso that both engine speed and torque followed the line of bestengine efficiency. The efficiency of the CVT itself dependedon engine power output; as discussed previously, an iterativeprocess was used to determine CVT efficiency for each time-step. The CVT gear ratio over the drive cycle is shown in
0506138-129.94.59.243182-GUID:3Licensed to University of New South Wales
Licensed from the SAE Digital Library Copyright 2010 SAE International
E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM
Figure 3. Model output for CVT equipped vehicle validation case.
Figure 3, along with engine torque and fuel flow rate at eachtime step. The speed-torque operating points for the engineare also plotted in Figure 3. These results indicate that, asdesired, the CVT control is maintaining the engine at itslower speed or peak-efficiency points. Qualitatively, themodel results shown in Figure 3 agree well with those fromthe reference case [8]. Table 1. CVT model result validation Honda Insight parallel-hybrid vehicle with a 1.0 L 50 kWgasoline engine [9]. Supplemental power was provided by a10 kW electric motor which also acted as a generator forregenerative braking; energy storage was in a 144V, 6.5 AhNiMh battery pack. The net power from the motor and enginewas connected to the powertrain through a torque coupler anda five speed manual gearbox. Validation data was extractedusing NREL's on-board data acquisition tool [9] when thevehicle was operated over the UDDS drive cycle. Vehicle andcomponent parameters are given in the appendix.This hybrid system is more complex to model, as it involvesparallel energy transfer to and from the electric motor/generator and the engine. A new module, the torque coupler,was developed to split the power transfers between the twomodules; however, the exact operation of the vehicle torquecoupler is uncertain [9]. As a result, the module assumes acontrol strategies based on four general operating conditions,as defined in [9]:1. Engine start/stop - if the battery state of charge (SOC) issufficiently high, the engine turns off when the vehicle isstationary. 2. Acceleration - the electric motor provides supplementaltorque (‘motor assist’). The motor assist depends on thebattery SOC; the higher the SOC, the more supplementaltorque from the motor (up to a maximum of 50 N.m). Theelectric motor never operates independently from the engine.3. Cruising (small variations in speed) - engine torque is low,so the motor acts as a generator, increasing the engine torqueinto a more efficient range and storing the extra energy in thebattery. The torque absorption depends on SOC. Chargingstops at 70% SOC, to allow for energy storage fromregenerative braking. Validation Case 2 - HybridThe second validation case poses a greater challenge for themodel structure as it involved a parallel-hybrid powertrain.The validation data, generated by NREL, was for a 2000 Author:Gilligan-SID:13182-GUID:30506138-129.94.59.243Licensed to University of New South Wales Licensed from the SAE Digital Library Copyright 2010 SAE International E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM Figure 4. Model output for hybrid validation case over the full UDDS drive-cycle. 4. Deceleration - the electric motor runs as a generator toprovide regenerative braking. The rate is dependent on thecurrent supply limit to the battery. Any excess torqueextraction is accomplished with the brakes. This control strategy is implemented in the torque coupler. Assuch it includes two pairs of outputs: torque and angularvelocity from the engine and from the electric motor. Theelectric motor module then transfers energy to the batterymodule (both specially developed for this case). The torquecoupler module uses the battery SOC from the previous time-step when selecting its operating mode. The model predictions were in general agreement with theexperimental results reported elsewhere [9]. This is shown inFigure 4, where the battery current, motor and engine torques,and fuel consumption are plotted over the drive-cycle, alongwith a torque-speed plot. In general, the battery currentfollows the same pattern as shown in Figure 10 of reference[9]; the most notable discrepancy is that the model's predictedbattery draw over the first 100 seconds is less than thatreported from the experiments. Overall, however, the finalcurrent usage is within the range reported from theexperiments. The significant test-to-test variability shown in[9] demonstrates the complexity of the motor/battery control,and will have significantly influenced the final experimentalresults. The simplified model does not account for thesecomplexities; the fact that the same general trend and finaloutcome are obtained provides support that the model isadequately representing the overall vehicle performance. Table 2. Hybrid model result validation The two validation cases discussed above demonstrate thatthe simplified model is capable of providing a reasonableprediction of the total fuel flow rate for two very differentpowertrain configurations. While the accuracy of theindividual torque-speed points can not be conclusivelydemonstrated, they are reasonable and qualitatively similar tothose presented in the references. The fact that no tuningparameters were used to achieve these results indicates that Author:Gilligan-SID:13182-GUID:30506138-129.94.59.243Licensed to University of New South Wales Licensed from the SAE Digital Library Copyright 2010 SAE International E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM the overall modelling structure is sufficiently robust to allowfor evaluation of other powertrain configurations. CONVENTIONAL GEARBOX The validation cases provided two specific vehiclearchitectures. By combining various components from thetwo, it is possible to develop other powertrains. Replacing theCVT in the first validation case with the gearbox from thesecond permits the evaluation of a conventional series vehicleequipped with a manual transmission. The gearbox module requires further explanation. It has thesame inputs and outputs as the other powertrain modules, butdiffers in that a logical decision must be made as to what gearshould be selected. This is not always necessary: some drive-cycles include specified gear-shift points, and hence themodel would merely need to identify the gear based on thattime point. However, to provide greater flexibility, the modelincludes a more general routine where the engine speeds atwhich the gears should be changed are defined. It also definesa gear change time, which includes engaging and disengagingthe clutch and changing the gears. While the gear change isunder way, the vehicle is allowed to coast under its owninertia. Once the new gear is selected and the clutch re-engaged, the model attempts to regain the desired drive-cycle.For the work reported in the subsequent sections, the gearchanges were assumed to take 1 second to occur. Increasinggear changes were scheduled to occur at a speed of 320 rad/swhile decreases were scheduled for 100 rad/s. Combining theengine, powertrain and vehicle specifications from ValidationCase 1, with the conventional gearbox from Validation Case2 results in the engine torque-speed map shown in Figure 5.Figure 5. Engine torque/speed map over the UDDS(505)cycle for the vehicle described in validation case 1 but using a manual gearbox rather than the CVT. Author:Gilligan-SID:13PART II - APPLICATION OF MODELTO DEVELOP ENGINE TESTCONDITIONS One of the principle purposes for developing this model wasto provide a simple tool that links vehicle drive-cycleperformance to engine operating conditions. The requiredengine speed and torque for each time-step have been shownin Figures 3,4,5 for three powertrain architectures. The resultsdemonstrate the importance of powertrain and vehicleparameters on the engine operating mode. For example, theCVT-equipped vehicle's engine operation is biased towardslower speeds and higher torques, as its control strategyattempts to reach the optimum engine operating efficiency asquickly as possible. Conversely, the manual gearbox and thehybrid are biased towards higher engine speeds at lowerloads. The results presented above treat each time-step as being atsteady-state conditions, neglecting transients within the time-step. For some situations, this may be appropriate; as thevalidation tests demonstrated, fuel consumption and energystorage can be reasonably well predicted from such a series ofquasi-steady test points. Pollutant emissions, however, aremore sensitive to transient effects. As a result, it is importantto consider transient effects when selecting engine testconditions that best represent the operating modesencountered by a vehicle following a real drive-cycle. Detailed evaluation of one transient eventThe various drive cycles, as discussed previously and shownin Figure 1, involve a series of relatively smoothaccelerations and decelerations. The changes in speed andtorque requirements at the engine do not follow such smoothtraces, as they depend on both the drive-cycle and the vehiclepowertrain. As an example, a single acceleration from 346 sto 370 s of the UDDS(505) cycle will be investigated in moredetail. As demonstrated in Figure 1, this acceleration is verysimilar in most respects to the other accelerations in thiscycle. In this phase, the vehicle accelerates from stationary toa velocity of 16.32 m/s (58.7 km/hr) in 24 seconds. Theengine speed and torque required to meet this drive cycle forthe three simulated powertrain architectures (manual gearbox;CVT; hybrid) are shown in Figure 6.0506138-129.94.59.243182-GUID:3Licensed to University of New South Wales Licensed from the SAE Digital Library Copyright 2010 SAE International E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM Figure 6. Engine Speed (top) and Torque (bottom)during a selected transient for three vehicle powertrain architecturesThe three powertrains modeled result in very different enginespeed/load requirements over this specific acceleration, asshown in Figure 6. The CVT-equipped vehicle had thehighest torque demands (as suggested in Figure 4) and thegreatest variations in torque, but relatively small variations inengine speed. This is a result of its control strategy, withlower engine speeds and higher torques to maximize engineefficiency. As a result, the engine is more likely to see greaterchanges in torque at relatively constant speed. The samevehicle transient in the gearbox-equipped vehicle results inlower absolute magnitudes and small changes in torque butlarger changes in speed. Specifically, gear changes result inlarge steps in engine speed and torque. For the hybridpowertrain, the ‘torque assist’ electric motor helps to dampout the largest transient effects, although there are stillsignificant changes in engine speed, especially during theearly acceleration. The most severe engine transient for eachof the three powertrain architectures is shown in Table 3.Table 3. Most significant engine transient for each ofthree powertrain architectures on the UDDS(505) cycle during the period 345 s to 370 s. Author:Gilligan-SID:13182The detailed analysis of one transient event provides ademonstration of the model's ability to predict engineperformance. However, the result is still a transient event. Togenerate representative steady-state test conditions, thetransients identified in Table 3 could be entered into anengine simulation program. This would allow evaluation ofhow the engine operating parameters (e.g., fuel flow, chargeair pressure, EGR level) varied during such a transient. Fromthese results, ‘quasi-steady’ engine operating conditionscould be defined, which are representative of some of theconditions encountered by individual engine cycles duringthis transient. This is the focus of ongoing work; for thecurrent paper, it is desirable to evaluate the variability inengine operating conditions over the entire drive-cycle. Speed-Torque transientsThe preceding section focused on a single accelerationprocess. To more clearly identify the transient operatingconditions to be encountered with the varying powertrainarchitectures, Figure 7 shows the ‘transient magnitude’ - thatis, the change in engine torque and speed from one time stepto the next - for all the test points for each powertrain overthe UDDS(505) test cycle. All three architectures show thehighest density of points around the origin, which isrepresentative of near-steady state operation. In general, thevariations tend to be clustered around either the vertical or thehorizontal axes, indicating that larger changes in either speedor load are accompanied by only small changes in the otherparameter; this agrees with the results shown in Table 3.There are also significant differences between powertrains; asFigure 3 suggests, the CVT shows larger changes in torque atconstant speed than either of the other two architectures. 0506138-129.94.59.243-GUID:3Licensed to University of New South Wales Licensed from the SAE Digital Library Copyright 2010 SAE International E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM Figure 7. Changes in speed and torque between time-points for the three powertrains at all points of the UDDS(505) cycle. Effects of drive cycleAs discussed earlier, the drive cycle being used cansignificantly influence the conditions that the engineencounters. To demonstrate this, the engine speed and torquefor the conventional gearbox equipped vehicle over theUDDS(505) and NEDC(reduced) cycles are shown in Figure8. The figure also shows the changes in speed and torquebetween subsequent time-points. The results show that theNEDC(reduced) drive-cycle involves higher speed operation,but at lower torques. This is consistent with theNEDC(reduced)'s slower transients but higher peak speedsrelative to the UDDS(505). The changes in speed and torqueat one second intervals indicate that with theNEDC(reduced), changes in speed are larger and changes intorque are smaller than in the UDDS(505). It is particularlynotable that the NEDC(reduced) features very few test points Author:Gilligan-SID:13which see simultaneous significant changes in speed andtorque. Figure 8. effect of drive cycle on steady state (top) and transient (bottom) engine conditions, for the conventional gearbox situation. DEVELOPMENT OF ENGINE TESTCONDITIONS The preceding results demonstrate the sensitivity of theengine operating condition to powertrain architecture anddrive-cycle. One way to use these results to develop a set ofsteady-state engine test conditions is to evaluate thefrequency of occurrence of each engine speed/load condition,as shown in Figure 9 for the three powertrain architectures ofinterest. For these plots, the engine operating range wassubdivided into a number of bins (bin sizes were 20 N.mtorque by 30 rad/s speed); the number of points within eachspeed / torque bin were then determined. Figure 9 providesmore information on torque-speed point densities than didthose in Figures 3,4,5.0506138-129.94.59.243182-GUID:3Licensed to University of New South Wales Licensed from the SAE Digital Library Copyright 2010 SAE International E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM Table 3. Five most frequent engine speed / torque conditions: Figure 9. Frequency of occurrence of specific torque-speed points on the basis of time spent at each condition in the UDDS(505) cycle.The speed / torque distributions shown in Figure 9 clearlydemonstrate the very different engine operating conditions Author:Gilligan-SID:13encountered with different powertrain architectures. The fivemost frequently occurring speed-torque points based on thefrequency of occurrence are listed in Table 5 for each of thethree architectures. The equivalent results for theconventional powertrain operating on the NEDC(reduced)cycle are also shown. The five points identified representapproximately ½ of all the test points encountered during thedrive cycle. The most frequently occurring condition for bothcases was idle; the remaining test conditions varied withpowertrain architecture. Licensed from the SAE Digital Library Copyright 2010 SAE International E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM Table 4. Five highest fuel consumption engine speed / torque conditions: Figure 10. Frequency of occurrence of specific torque-speed points on the basis of fuel consumed at each condition in the UDDS(505) cycle. Transient magnitudeThe magnitude of change in engine speed and load for eachtest point is presented in Figures 7 and 8. The figures suggestthat there are relatively few test conditions with large speedand torque transients. These findings can be supported byevaluating frequency plots of the changes in speed and loadfor each test point, as shown in Figure 11 for the conventionaland hybrid powertrains. Similar to the preceding discussion,the results were put into speed/torque bins; bin sizes werechanges of 100 RPM speed and 20 N.m torque. The figuredemonstrates that the vast majority of test points see smallchanges in speed and torque (< 100 RPM or 20 N.m) fromthe preceding test point. The hybrid case shows a slightlyhigher frequency of points with larger changes in speed, butthe changes are in general small (at most 500 RPM) forfrequencies in excess of 2%. While individual test points aremore widely scattered, as shown in Figures 7 and 8, the vastmajority of test points on the UDDS(505) cycle undergo onlysmall changes in speed and load between test points. Thissuggests that, steady-state engine conditions should berepresentative of the majority of actual engine operatingconditions. 0506138-129.94.59.243:13182-GUID:3Licensed to University of New South Wales Licensed from the SAE Digital Library Copyright 2010 SAE International E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM Figure 11. Frequency plots for change in speed andtorque, per second, for conventional and hybrid powertrains on the UDDS(505) cycle. DISCUSSION A series of steady-state engine test conditions have beendefined for various vehicle architectures and different drive-cycles based on the results from a simple, modular powertrainmodel. The results from this study indicate that the vehicletransients encountered during the investigated drive-cyclesare relatively repeatable. As a result, it should be possible todefine a few ‘quasi-transient’ test conditions. These testconditions would be defined based on the engine speed andtorque values reported here for the specific transient ofinterest; however, it would also be necessary to define engineparameters that may vary during the transient, such as fuelquantity, intake manifold pressure, or EGR level. Definingthese values would require either a detailed engine model ortransient test-bed engine tests. The transient magnitude evaluation may be of use to expandthe capabilities of the powertrain model presented here. Theability to accurately predict engine emissions over transienttest-cycles remains a challenge for powertrain models. Byincorporating ‘transient magnitude’, it may be possible toquantify not only the engine speed and torque, but also therate of change of these conditions at that test point. Thiscould be combined with basic modelling or experimentalinformation on the effect of the magnitude of change onengine-out emissions to provide a more accuraterepresentation of emissions over the entire drive cycle. Author:Gilligan-SID:13However, further model development and experimentalresearch is required before this can be achieved. CONCLUSIONS The overall objective of this work was to describe andvalidate a simple powertrain model, and then to use it todefine representative operating conditions for steady-stateengine testing. This work led to the following specificconclusions: 1. A simple model with a modular structure has beendeveloped that can be used to model a wide array ofpowertrain architectures. Validation of the results on twodifferent powertrain architectures demonstrated that total fuelconsumption could be estimated to within 7% ofexperimental results, without the use of manual tuningparameters. 2. Various aspects of the powertrains used in the validationcases were simplified either to ease the modelling process ordue to a lack of information. The fact that the model stillprovided a reasonable prediction of the experimentalperformance indicates that it can be used to define enginedynamometer test conditions that are representative of agiven powertrain configuration and drive cycle. 3. Using the model, it has been shown that vehiclearchitecture and test cycle significantly influence the engineoperating conditions and hence the experimental testconditions that a test engine would need to replicate.4. Engine test conditions can be defined based on steady-state results that are representative of the vast majority of testconditions. These test points can be based on either timespent at a specific test point or on the total fuel quantityconsumed at a certain condition. 5. The model results show that when following theprescribed drive-cycles, relatively few transient testconditions with large variations in engine speed or load areencountered over short time-scales. REFERENCES 1. Samuel, S., Morrey, D., Fowkes, M., Taylor, D., Garner,C., Austin, L., “Numerical Investigation of Real-WorldGasoline Car Drive-Cycle Fuel Economy and Emissions,”SAE Technical Paper 2004-01-0635, 2004.2. Heath, R., Mo, C., “A Modular Approach to PowertrainModelling for the Prediction of Vehicle Performance,Economy and Emissions,” SAE Technical Paper 960427,1996. 3. Ali, M., Moskwa, J., “Developing a Generalized ModularModeling Structure for Dynamic Engine Simulation,” SAETechnical Paper 2002-01-0202, 2002.0506138-129.94.59.243182-GUID:3Licensed to University of New South Wales Licensed from the SAE Digital Library Copyright 2010 SAE International E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM 4. Sun, G., Wei, M., Shao, J., Pei, M., “Automotive Powertrain Modelling and Simulation Based on AMESim,”SAE Technical Paper 2007-01-3464, 2007.5. Botelle, E., Hayat, O., Lebrun, M., Domingues, E.,“Powertrain Driveability Evaluation: Analysis and Simplification of Dynamic Models,” SAE Technical Paper2003-01-1328, 2003.6. Simpson, A., “Validation of a Parametric VehicleModelling Tool Using Published Data for Prototype andProduction Vehicles with Advanced Powertrain Technologies,” SAE Technical Paper 2005-01-3481, 2005.7. Markel, T, Brooker, A, Hendricks, T. et al. ADVISOR: Asystem analysis tool for advanced vehicle modelling. Journalof Power Sources. 4801. 2002. pp. 1-12. 8. Min, B., Matthews, R., Duoba, M., Larsen, B., Ng, H.,“Direct Measurement of Powertrain Component Efficienciesfor a Light-duty Vehicle with a CVT Operating Over aDriving Cycle,” SAE Technical Paper 2003-01-3202, 2003.9. Kelly, KJ, Zolot, M, Glinsky, G and Hieronymus, A. TestResults and Modeling of the Honda Insight using ADVISOR.SAE Technical Paper 2001-01-2537, 2001.10. Montazeri-Gh, M., Varasteh, H., Naghizdeh, M., “Driving Cycle Simulation for Heavy Duty Engine EmissionEvaluation and Testing,” SAE Technical Paper2005-01-3796, 2005.11. available from: http://www.epa.gov/otaq/emisslab/testing/dynamometer.htm#vehcycles. Last accessed 17-09-09.Author:Gilligan-SID:13182-GUID:30506138-129.94.59.243Licensed to University of New South Wales Licensed from the SAE Digital Library Copyright 2010 SAE International E-mailing, copying and internet posting are prohibitedDownloaded Sunday, April 04, 2010 4:08:12 AM APPENDIX Table A.1. Vehicle and Powertrain Parameters for Validation Cases *All values from [8], unless otherwise stated**All values from [9], unless otherwise statedThe Engineering Meetings Board has approved this paper for publication. It has successfully completed SAE's peer review process under the supervision of the sessionorganizer. This process requires a minimum of three (3) reviews by industry experts.All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical,photocopying, recording, or otherwise, without the prior written permission of SAE.ISSN 0148-7191doi:10.4271/2010-01-0002 Positions and opinions advanced in this paper are those of the author(s) and not necessarily those of SAE. The author is solely responsible for the content of the paper.SAE Customer Service: Tel: 877-606-7323 (inside USA and Canada)Tel: 724-776-4970 (outside USA)Fax: 724-776-0790 Email: CustomerService@sae.org SAE Web Address: http://www.sae.orgPrinted in USA Author:Gilligan-SID:13182-GUID:30506138-129.94.59.243 因篇幅问题不能全部显示，请点此查看更多更全内容The total fuel consumption over the drive cycle can be usedto provide a quantitative validation of the model, as shown inTable 1. The results indicate that the model over-predictedfuel consumption by 3.6%; no tuning parameters were used toachieve these results. The difference shown in Table 1 isconsidered acceptable given the uncertainty in having derivedthe engine fuel consumption and CVT efficiency results fromgraphical sources in [8]. Other sources of uncertainty includethe CVT control algorithm; this is likely more complex thanwas used in the model. The good agreement shown in Table 1indicates that such a simplified model can give a reasonableprediction of vehicle performance.Quantitative validation of the model results are provided bythe full-cycle fuel consumption and battery current usagereported in [9], as shown in Table 2. The results demonstrategood agreement for the simplified model; as in the CVT case,this was achieved without using any arbitrary tuningparameters. The discrepancy between the test and modelconditions is most likely related to the complexities of thepower control strategies not being fully incorporated into thesimplified model.The frequency of occurrence shown in Figure 9 is based onthe number of time points occurring within a given enginespeed and torque range. However, it may be more interestingto select the points which represent the greatest level of fuelconsumption, as shown by ‘fuel-density’ plots in Figure 10.These figures represent the fraction of the total fuelconsumption that occurs in each of the speed/torque bins overthe entire UDDS(505) test cycle. The bins were defined in thesame manner as those presented in Figure 9. The five testconditions with the greatest fuel consumption are shown inTable 4.0506138-129.94.59.243182-GUID:3Licensed to University of New South Wales
Author:Gilligan-SIDComparing the results shown in Tables 3 and 4, it is apparentthat basing the selection on time or fuel consumption has asubstantial impact on the most important operatingconditions. Other parameters, for example emissions, couldalso be used: however, the appropriateness of the selected testpoints would depend on the accuracy of the emissionsprediction included in the model.