8
Modeling Improvements in Vehicle Fuel Consumption

INTRODUCTION

The potential of technology to reduce fuel consumption can be estimated in three basic ways. One approach involves constructing an actual prototype vehicle with the technologies in question, performing the Environmental Protection Agency (EPA) city and highway dynamometer tests repeatedly, and then measuring the fuel consumption. Although there is some variability from test to test, this method is the most accurate but is also prohibitively expensive. A second approach is to construct a computer model that represents all of a vehicle’s components and their interactions, including representations of the technologies for reducing fuel consumption, and to simulate the behavior of the vehicle over the federal test procedures. This method, which the committee refers to as full system simulation (FSS), is now the state of the art throughout the automotive industry for modeling fuel consumption. Although it is less expensive, FSS still requires very large expenditures of time and money if it is used to calibrate models to the 1,000 or so different vehicle configurations offered for sale in the United States each model-year and to test all relevant combinations of technologies. The third alternative is to construct an algorithm that adds discrete technologies to the set of base-year vehicle configurations and that then calculates their cumulative impact while attempting to account for interactions between them by means of adjustment factors. The committee refers to this third method as partial discrete approximation (PDA). The simplicity of the third approach allows fuel consumption impacts to be calculated for thousands of vehicles and tens of thousands of technology combinations. The key question is whether the third method can be executed with sufficient accuracy to support fuel economy regulation. The Volpe Model (Van Schalkwyk et al., 2009), used by the National Highway Traffic Safety Administration (NHTSA) in its rulemaking analyses, and the EPA’s OMEGA model, used by the EPA in its rulemaking analysis (EPA and NHTSA, 2009), are PDA models that use data on technology costs and fuel consumption impacts from a variety of sources, including FSS models.

This chapter evaluates methods of estimating the potential to decrease automotive fuel consumption by changing vehicle design and technology. It begins with some general observations on modeling technologies’ potential for reducing fuel consumption. Because of the technological complexities of vehicle systems, predicting how combinations of technologies might perform in new vehicle designs involves uncertainty. The present committee summarizes and discusses the method used by the National Research Council (NRC) Committee on the Effectiveness and Impact of Corporate Average Fuel Economy (CAFE) Standards in its 2002 report (NRC, 2002). It then goes on to compare and evaluate the two most widely used approaches to estimating the technological potential for reducing fuel consumption—PDA and FSS. Both methods are described in detail, and applications of the two methods to various types and configurations of vehicles are compared. Although it was able to make useful comparisons between modeling methods, the committee found that information comparing the results of either the FSS or the PDA method to real-world vehicles is scarce. The committee also comments briefly on the methodology used by the NHTSA in its 2011 Final Rule.

Recognizing the limitations of all modeling approaches, the committee considers the FSS method to be the state of the art and therefore the preferred method for estimating the potential of technologies to reduce fuel consumption. However, given the cost of FSS modeling at present, the committee believes that the PDA method, properly executed and supplemented with estimates of technology interaction effects developed by FSS or lumped parameter modeling, can be a reasonably accurate method for assessing the potential for reducing light-duty vehicle fuel consumption over a time horizon on the order of 10 years.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 118
8 Modeling Improvements in Vehicle Fuel Consumption INTRODUCTION 2009), are PDA models that use data on technology costs and fuel consumption impacts from a variety of sources, The potential of technology to reduce fuel consumption including FSS models. can be estimated in three basic ways. One approach involves This chapter evaluates methods of estimating the poten- constructing an actual prototype vehicle with the technolo- tial to decrease automotive fuel consumption by changing gies in question, performing the Environmental Protection vehicle design and technology. It begins with some gen- Agency (EPA) city and highway dynamometer tests repeat- eral observations on modeling technologies’ potential for edly, and then measuring the fuel consumption. Although reducing fuel consumption. Because of the technological there is some variability from test to test, this method is the complexities of vehicle systems, predicting how combina- most accurate but is also prohibitively expensive. A second tions of technologies might perform in new vehicle designs approach is to construct a computer model that represents involves uncertainty. The present committee summarizes and all of a vehicle’s components and their interactions, includ- discusses the method used by the National Research Council ing representations of the technologies for reducing fuel (NRC) Committee on the Effectiveness and Impact of Cor- consumption, and to simulate the behavior of the vehicle porate Average Fuel Economy (CAFE) Standards in its 2002 over the federal test procedures. This method, which the report (NRC, 2002). It then goes on to compare and evaluate committee refers to as full system simulation (FSS), is now the two most widely used approaches to estimating the tech- the state of the art throughout the automotive industry for nological potential for reducing fuel consumption—PDA and modeling fuel consumption. Although it is less expensive, FSS. Both methods are described in detail, and applications FSS still requires very large expenditures of time and money of the two methods to various types and configurations of if it is used to calibrate models to the 1,000 or so different vehicles are compared. Although it was able to make useful vehicle configurations offered for sale in the United States comparisons between modeling methods, the committee each model-year and to test all relevant combinations of tech- found that information comparing the results of either the nologies. The third alternative is to construct an algorithm FSS or the PDA method to real-world vehicles is scarce. The that adds discrete technologies to the set of base-year vehicle committee also comments briefly on the methodology used configurations and that then calculates their cumulative im- by the NHTSA in its 2011 Final Rule. pact while attempting to account for interactions between Recognizing the limitations of all modeling approaches, them by means of adjustment factors. The committee refers the committee considers the FSS method to be the state of to this third method as partial discrete approximation (PDA). the art and therefore the preferred method for estimating The simplicity of the third approach allows fuel consumption the potential of technologies to reduce fuel consumption. impacts to be calculated for thousands of vehicles and tens However, given the cost of FSS modeling at present, the of thousands of technology combinations. The key question committee believes that the PDA method, properly executed is whether the third method can be executed with sufficient and supplemented with estimates of technology interaction accuracy to support fuel economy regulation. The Volpe effects developed by FSS or lumped parameter modeling, can Model (Van Schalkwyk et al., 2009), used by the National be a reasonably accurate method for assessing the potential Highway Traffic Safety Administration (NHTSA) in its for reducing light-duty vehicle fuel consumption over a time rulemaking analyses, and the EPA’s OMEGA model, used horizon on the order of 10 years. by the EPA in its rulemaking analysis (EPA and NHTSA, 118

OCR for page 118
119 MODELING IMPROVEMENTS IN VEHICLE FUEL CONSUMPTION CHALLENGES IN MODELING VEHICLE FUEL 1. Differences between the attributes of the representa- CONSUMPTION tive or typical vehicle being analyzed and the actual vehicles it represents; Along with the many potential benefits of using computer 2. Inaccuracies in the characterization of the base vehicle, models to understand vehicle systems come limitations as especially its energy flows; well. In addition to enabling insight into how an overall 3. Inaccurate assessment of technology impacts, includ- vehicle system might operate, vehicle system modeling can ing the inability to fully represent the physics of a new also help measure the interactions between vehicular sub- technology in FSS modeling; systems and how they affect overall vehicle performance. An 4. Differences in the implementation of a given technol- understanding of the physics underlying these interactions is ogy from vehicle to vehicle; important when trying to estimate how future vehicles might 5. Changes in the nature of a technology over time; and perform with different combinations of technologies. All 6. Inaccurate estimation of the synergies among tech- models are inherently simplifications of reality; the physics nologies and how they contribute to the overall end of real processes will always be considerably more compli- result of their combined application.1 cated than that reflected in a model. In the end, impacts can only be known with certainty when a technological concept In general, rigorous, quantitative assessments of these is realized in a real vehicle, and even then realizations of potential sources of error and their impacts on the potential the same technological concept can differ from one vehicle to reduce vehicle fuel consumption are scarce. to another. The meaningful question is whether any given In this chapter comparisons of the results of FSS and model or methodology has sufficient fidelity to competent PDA (with lumped parameter modeling) are presented. In executions of the technological concept to achieve the goals addition, the committee contracted with Ricardo, Inc., to for which the model has been developed. perform a statistical analysis of FSS modeling. The goal With even the most complex and comprehensive models, was to determine whether a limited number of FSS runs t here are challenges when modeling a k nown vehicle could be used to generate accurate data on the main effects configuration, and even greater challenges when trying to of technologies and their interactions, which could then be predict the behavior of future vehicles using new combina - used as basic input data for PDA modeling. The results of tions of technologies. When modeling a known or existing the analysis support the feasibility of this concept. Unfortu- vehicle the principal problems are in capturing the desired nately, scientific data about the accuracy of either modeling dynamics to a sufficient level of detail or fidelity, and in col- method in comparison to actual vehicles are very limited. lecting and inputting representative parameters or boundary conditions. The advantage of modeling a known vehicle is METHODOLOGY OF THE 2002 NATIONAL RESEARCH that data on the vehicle’s actual performance are usually COUNCIL REPORT available to the modeler, and the data can be used to tune or validate the model’s performance. Even for existing The 2002 NRC report Effectiveness and Impact of Corpo- vehicles, however, experimental data sets are frequently rate Average Fuel Economy Standards used a type of PDA sparse and may not include the precise performance situ - method to estimate the potential future reductions of fuel ations of interest. consumption by light-duty vehicles. The 2002 committee Detailed computer modeling of vehicle systems can be recognized the existence of synergies among technologies very expensive. Developing sufficient data on the perfor- applied to reduce fuel consumption but did not provide ex- mance of engines and other components, data that are not plicit estimates of the effects of such interactions. Technolo - generally available in the open literature, is a major source of gies were implemented in defined sequences called paths, the expense of FSS modeling. An automobile company might and the impacts of technologies on fuel consumption were spend many times the resources available to the committee to adjusted to account for interactions with other technologies develop dynamic models to help answer the kinds of ques- previously adopted. tions posed to the committee. On the order of 1,000 different vehicle configurations undergo fuel consumption testing each model year. FSS modeling of even the most promis- ing combinations of advanced technologies for such a large 1 In this report the committee chose to use the term synergies as defined in number of vehicles would be expensive for federal agencies. the joint EPA and NHSTA “Proposed Rulemaking to Establish Light-Duty PDA modeling, on the other hand, can be implemented in Vehicle Greenhouse Gas Emission Standards and Corporate Average Fuel Economy Standards” (EPA and NHTSA, 2009). Two or more technologies simplified algorithms that can estimate fuel consumption applied together might be negatively synergistic, meaning that the sum of potentials for thousands of vehicles or more, considering their effects is less than the impact of the individual technologies (contrib- virtually all logical combinations of technologies. utes less to reducing fuel consumption, in this case), or might be positively There are at least six sources of error in estimating the synergistic, meaning that the sum of the technologies’ effects is greater than potential to reduce vehicle fuel consumption: the impact of the individual technologies (in this case, contributes more to reducing fuel consumption).

OCR for page 118
120 ASSESSMENT OF FUEL ECONOMY TECHNOLOGIES FOR LIGHT-DUTY VEHICLES and possible tradeoffs with other vehicle attributes. Ranges Technology changes modify the system and hence have complex effects that are difficult to capture and analyze. were given for costs in order to reflect manufacturer-specific It is usually possible, however, to estimate the impacts of conditions, market uncertainties, and the potential for evo- specific technologies in terms of a percentage savings in fuel lutionary costs reductions for new technologies. The 2002 consumption for a typical vehicle without a full examination committee did not specify a confidence interval for the of all the system-level effects. (NRC, 2002, p. 33) ranges, nor did it explicitly address interdependencies or synergies of performance or cost, except via the incremental For each technology assessed, the committee estimated not effects of sequential application in the technology paths. only the incremental percentage improvement in fuel con- The incremental fuel consumption improvement and retail sumption . . . but also the incremental cost that applying the price equivalent estimates in Table 8.1 are additive but only technology would add to the retail price of a vehicle. (NRC, for a particular technology path. The technologies included 2002, p. 35) in a path are indicated by an “X” in the columns labeled 1, 2, and 3. Technologies not contained in a path were not to The 2002 NRC committee grouped technologies into be added to that path. A range of estimates is provided for three categories: engine technologies, transmission technolo- both fuel consumption and cost impacts. However, only the gies, and vehicle technologies. Vehicles were grouped into 10 midpoints of those estimates can be directly accumulated (as classes. Table 8.1 is the 2002 committee’s list of technologies illustrated in Figure 8.1), since accumulation of all the high- for passenger cars, including ranges for the estimated incre- end or low-end estimates without adjustment would produce mental reductions in fuel consumption and for incremental misleading results. RPE impacts. The 2002 NRC committee’s method received some criti- For each vehicle class three different sequences of tech- cism for being overly simplistic. One notable critique (Patton nology implementation, called “production development et al., 2002) cited three major issues: paths,” were mapped out. Figure 8.1 shows impacts of the technologies included in the three paths for passenger cars, 1. Failure to examine system-level effects; as noted in Table 8.1, on the fuel consumption of a midsize 2. Inaccurate fuel consumption estimates for individual car. The paths were intended to provide a logical sequence technologies; and of implementation of the various technologies and to ensure 3. Overcounting of fuel consumption reductions. that the incremental fuel consumption reductions from a given technology could be estimated conditional on the The first point chiefly faulted the 2002 committee for technologies that had preceded it. Paths 1 and 2 comprised multiplying together the impacts of individual technologies proven technologies that could be introduced within the as if they were independent. It observed that when technolo- next 10 years (from 2002), with Path 2 including some more gies address different energy-loss mechanisms, their impacts costly technologies than Path 1. Path 3 included additional generally are independent, but when technologies address the emerging technologies the 2002 committee believed would same energy-loss mechanism (e.g., engine pumping losses), become available within the next 15 years. The list of emerg- the aggregate effect may be more complex. The committee ing technologies included several technologies that are now believed that it had addressed this issue by estimating the in use (intake valve throttling, automated manual transmis- incremental effects of technologies implemented in a speci- sion, advanced continuously variable transmissions (CVTs), fied order. However, that committee neglected to quantify integrated starter/generator, electric power steering) and the energy losses addressed by each technology and did not several that are still not available (camless valve actuation, separately quantify the interactions among technologies. variable compression ratio engine). In addition, the 2002 The second critique covered a variety of points including committee judged that the potential for diesels to meet tighter the degree of optimism in studies cited to support the com- emissions standards was highly uncertain and also excluded mittee’s estimates and inadequate attention to the depen- hybrids from its quantitative assessment due to uncertainty dence of fuel consumption impacts on the characteristics of about their future potential. However, both technologies are the vehicle to which they are applied. now available on mass market vehicles in the United States. In estimating the potential reduction in fuel consump- An example of this is cylinder deactivation. According to tion (gallons per 100 miles) of each technology, the 2002 the report, cylinder deactivation is “applied to rather large committee drew on a variety of sources of information, engines (>4.0 L) in V8 and V12 configurations.” Yet the re- from published reports to presentations to the committee by port applies the same fuel consumption reduction factor for experts and consultations with automotive manufacturers and cylinder deactivation to vehicles with six and four cylinder suppliers. Having studied the available information, the 2002 engines, where the actual benefit would be smaller. (Patton committee used its own expertise and judgment to decide on et al., 2002, p. 10) ranges of estimates for each technology. The ranges were in- tended to reflect uncertainties with respect to the technology However, the 2002 committee applied cylinder deactiva- of the baseline vehicle, effectiveness of the implementation, tion only to large passenger cars, midsize and larger sport

OCR for page 118
TABLE 8.1 Fuel Consumption Technology Matrix: Passenger Cars Retail Price Equivalent ($) Subcompact Compact Midsize Large Baseline: overhead cams, 4-valve, fixed timing, roller finger Fuel Consumption follower Improvement (%) Low High 1 2 3 1 2 3 1 2 3 1 2 3 Production-intent engine technology Engine friction reduction 1-5 35 140 X X X X X X X X X X X X Low-friction lubricants 1 8 11 X X X X X X X X X X X X Multivalve, overhead camshaft (2-V vs. 4-V) 2-5 105 140 Variable valve timing (VVT) 2-3 35 140 X X X X X X X X X X X X Variable valve lift and timing 1-2 70 210 X X X X X X X X Cylinder deactivation 3-6 112 252 X Engine accessory improvement 1-2 84 112 X X X X X X X X X X X X Engine supercharging and downsizing 5-7 350 560 X X Production-intent transmission technology Five-speed automatic transmission 2-3 70 154 X X X X X X Continuously variable transmission 4-8 140 350 X X X X X X Automatic transmission w/aggressive shift logic 1-3 70 X X X X — Six-speed automatic transmission 1-2 140 280 X X X MODELING IMPROVEMENTS IN VEHICLE FUEL CONSUMPTION Production-intent vehicle technology Aero drag reduction 1-2 140 X X X X X X — Improved rolling resistance 1-1.5 14 56 X X X X X X X X X X X X Safety technology Safety weight increase −3 to −4 0 0 X X X X X X X X X X X X Emerging engine technology Intake valve throttling 3-6 210 420 X X X X Camless valve actuation 5-10 280 560 X X X X Variable compression ratio 2-6 210 490 X X X X Emerging transmission technology Automatic shift/manual transmission (AST/AMT) 3-5 70 280 X X Advanced CVTs—allows high torque 0-2 350 840 X X Emerging vehicle technology 42-V electrical system 1-2 70 280 X X X X X X Integrated starter/generator (idle off-restart) 4-7 210 350 X X X X Electric power steering 1.5-2.5 105 150 X X X X X X Vehicle weight reduction (5%) 3-4 210 350 NOTE: An X means the technology is applicable to the particular vehicle. Safety weight added (EPA baseline + 3.5%) to initial average mileage/consumption values. SOURCE: NRC (2002), Table 3.1. 121

OCR for page 118
122 ASSESSMENT OF FUEL ECONOMY TECHNOLOGIES FOR LIGHT-DUTY VEHICLES FIGURE 8.1 Estimated cost of fuel consumption reduction in model-year 1999 midsize cars. SOURCE: NRC (2002), Figure 3.6. Figure 8-1 new, still bitmapped utility vehicles (SUVs), minivans, and pickup trucks. Nearly et al. (2002) attributed essentially all of the 2002 committee’s all of these vehicles have engines with six or more cylin- 4 to 8 percent benefit to reduction in pumping loss (and even ders. Cylinder deactivation is applied today to six-cylinder added an additional 0.5 to 1.0 percent to pumping loss reduc- engines. Nonetheless, the 2002 committee’s characterization tion that compensated for reduced transmission efficiency). of baseline vehicles was based solely on the typical attributes Only a 0.0 to 0.5 percent benefit was assigned to increased of the 10 vehicle classes. Using the average characteristics of thermal efficiency, presumably due to operating the engine 10 classes of vehicles will lead to a certain degree of error if in a more efficient portion of the engine map more of the the resulting estimates are applied to the vehicles of specific time. Likewise, most of the benefits of 5-speed and 6-speed manufacturers. automatic transmissions (versus 4-speed) were attributed to The criticism of inadequate attention to individual ve- reducing pumping losses with no benefits for engine thermal hicle characteristics can also be leveled at the 2002 NRC efficiency. Similarly, 4.0 to 6.0 percent of the committee’s committee’s costs estimates. The costs of fuel consumption estimated 5.0 to 7.0 percent benefits of engine boosting and technologies in the 2002 NRC report were the same for all downsizing was attributed to reduced pumping losses. The vehicle classes. In fact, the costs of many technologies scale 2002 committee, on the other hand, judged that the tech- directly with measurable vehicle attributes such as weight nology derives much of its benefits from increased engine or cylinder count. efficiency at light load due to engine downsizing and, when The third critique is that the 2002 NRC committee’s possible, reduced friction due to reduced cylinder count at estimates overstated the potential benefits of technologies equivalent power. The 2002 committee asserted that the that primarily addressed pumping losses because the meth - energy efficiency benefits of multivalve, overhead camshaft odology did not take into account the theoretical limits of engines derived from four different sources: pumping loss reduction.2 The application of single and double overhead cam designs, Using their own judgments about the allocation of the ben- with two, three or four valves per cylinder, offers the poten- efits of technologies to reduction of pumping losses, Patton tial for reduced frictional losses (reduced mass and roller et al. (2002) divided the 2002 committee’s fuel consumption followers), higher specific power (hp/liter), engine down- benefit estimates into six categories of energy losses. Patton sizing, somewhat increased compression ratios, and reduced pumping losses. (NRC, 2002, p. 36) 2 Patton et al. (2002) estimated the theoretical limits at between a 13 percent and 17 percent reduction in fuel consumption, depending on the Patton et al. (2002) disagreed, assigning 2.0 to 5.0 vehicle in question. The U.S. EPA (2008b) estimated pumping plus friction percent of the committee’s estimated 2.0 to 5.0 percent losses at between 10 percent and 13 percent for actual vehicles, assuming a gross indicated engine efficiency of 37 percent.

OCR for page 118
123 MODELING IMPROVEMENTS IN VEHICLE FUEL CONSUMPTION MODELING USING PARTIAL DISCRETE improvement to reduced pumping losses, while adding a APPROXIMATION METHOD 0.5 to 1.0 percent benefit in thermal efficiency, offset by a −0.5 to −1.0 percent efficiency loss due to increased The PDA method incrementally adds discrete fuel- friction. consumption-reducing technologies to a baseline vehicle While the benefits of variable valve timing and lift until certain criteria are met. The method is sometimes (VVT + L) are largely reductions in pumping losses, they applied to individual vehicles but more often assumes that also include improved power, and the benefits of cylinder the fuel consumption impact and cost of a technology will deactivation include increased engine load (operation in a be approximately the same for all vehicles within at least a more efficient region of the engine map) as well as reduced subset (or class) of light-duty vehicles. In a presentation to pumping losses. Estimates of the benefits of the aforemen- the committee, K.G. Duleep of Energy and Environmental tioned technologies generated by FSS models have produced Analysis, Inc. (EEA) identified three important areas in results consistent with the 2002 NRC committee’s estimates. which the PDA method, and especially its application in the Recent estimates from the DOT/NHTSA (2009) and the 2002 NRC study, had come under criticism (Duleep, 2008). EPA (2008a) are compared with the 2002 NRC committee’s estimates in Table 8.2. The chief area of disagreement is the 1. Adequate definition of baseline vehicles; benefit of cylinder deactivation applied to multivalve, over- 2. Order of implementation of fuel consumption tech- head camshaft engines with VVT and discrete or continuous nologies; and lift control. The NHTSA estimated a benefit of 0.0 to 0.5 3. Accounting for synergies among fuel consumption percent, whereas the NRC and the EPA estimated benefits technologies. of 3 to 6 percent. The critics of the 2002 NRC report’s methodology make The chief disadvantage of the PDA method is that it is an important and valid point in calling attention to the lack entirely empirically based and therefore does not explicitly of a rigorous relation between the estimates of fuel consump- represent the interactions among any set of technologies. tion reduction and the physical energy flows in a vehicle. As Synergies among technologies are estimated by engineering a consequence, the plausibility of the 2002 NRC estimates judgment or by means of simplified analytical tools, such relied heavily on the expert judgment of the committee mem- as lumped parameter models of vehicle energy use (Duleep, bers. The 2002 NRC study’s method also did not explicitly 2008; Sovran and Blaser, 2003, 2006). Computational sim- account for the current use of the identified fuel economy plicity and the ability to quickly and economically process technologies in existing vehicles. Practitioners of the PDA information on thousands of individual vehicles and dozens method can and often do account for energy constraints us- of alternative combinations of technologies are the method’s ing simplified modeling methods called “lumped parameter” chief advantages. models, based on methods developed by Sovran and Bohn The main steps in the PDA process are the following: (1981) and extended by Sovran and Blaser (2003, 2006) and reviewed in Chapter 2 of this report. FSS models inherently 1. Identify discrete technologies with fuel consumption account for energy flows and ensure that physical limits will reduction potential. not be violated. TABLE 8.2 Comparison of Benefits of Valve Train Technologies as Estimated by NRC (2002), NHTSA’s Final Rule for 2011, and the EPA NHTSAa (%) Technology NRC (2002) (%) Midpoint (%) Midpoint (%) EPA (%) Midpoint (%) Multivalve OHC 2-5 3.5 1-2.6 1.8 NA NA Variable valve timing 2-3 2.5 3-5 4 2-4 3 Variable valve lift and timing 1-2 1.5 1.5-3.5 2.5 3-4 3.5 Cylinder deactivation 3-6 4.5 0.0-0.5 0.25 6 6 Subtotal 12 8.5 12.5 3-6 4.5 1.5-3.5 2.5 1-2 1.5 Intake valve throttlingb Total 16.5 11 14 5-10 7.5 NA NA 5-15 10 Camless valvesc aNHTSA’s fuel consumption benefits are path dependent. The path shown here is for dual overhead camshaft engines. bIn NHTSA’s terminology IVT is continuously variable valve lift (CVVL) and is a substitute for discrete variable valve lift (DVVL). NHTSA argues that cylinder deactivation applied to CVVL has little benefit since pumping losses have already been greatly reduced. Others argue that this misses the benefit of increased engine efficiency at higher load when a six-cylinder engine is operated on only three cylinders. cEffect of camless valve actuation is incremental to variable valve lift and timing not to intake valve throttling. The two are mutually exclusive. SOURCE: Based on data in NRC (2002), DOT/NHTSA (2009), and EPA (2008a).

OCR for page 118
124 ASSESSMENT OF FUEL ECONOMY TECHNOLOGIES FOR LIGHT-DUTY VEHICLES 2. Determine the applicability of each technology. select technologies that meet all of the following three 3. Estimate each technology’s impact on fuel consump- criteria: tion and cost. 4. Determine implementation sequences based on 1. Technologies already incorporated in at least one mass- a. Cost-effectiveness and produced vehicle somewhere in the world or prepro- b. Engineering and manufacturing considerations. duction technologies judged to have a strong likelihood 5. Identify and estimate synergistic effects of widespread adoption within the next decade; a. Based on empirical data and expert judgment, 2. Technologies having no significant negative impact, b. Using a simplified model of vehicle energy flows or a beneficial impact on attributes that are valued by (e.g., lumped-parameter model), or consumers or that are necessary to meet safety and c. Using estimates from FSS models. emissions regulations; and 6. Determine the “optimal” fuel consumption level by 3. Technologies whose cost does not far exceed the poten- a. Using a computer algorithm that sequentially applies tial value of fuel savings and other private and social technologies, benefits. b. Using fuel consumption cost curves. For example, all but a few of the technologies considered by the 2002 NRC study were already in mass production. In Identifying Technologies That Reduce Fuel Consumption general, PDA studies are most reliable when they are limited The PDA method, like the FSS method, begins with the to technologies already in production. However, the farther identification of distinct technologies that have the potential one must look into the future the less tenable this constraint to reduce vehicle fuel consumption at a realistic cost.3 The becomes. list of all possible technologies with some potential to reduce fuel consumption could range from lower-rolling-resistance Determining Applicability tires and improved engine lubricants to human-powered vehicles and the compressed air engine. When the purpose Not every technology will be applicable to every vehicle. is regulatory rulemaking, not all possible fuel consumption Torque limitations, for example, have so far prevented the technologies should be included. The world record for auto- use of CVTs in the largest, most powerful light-duty vehicles. motive fuel economy is held by the Pac Car II, a fuel-cell- Engine downsizing by reducing the number of cylinders with powered vehicle that won the 2005 Shell Ecomarathon in turbo-charging may be considered applicable to six-cylinder Ladoux, France, with a gasoline equivalent fuel economy of engines but less so to four-cylinder engines due to vibration 12,666 miles per gallon.4 The three-wheel vehicle accommo- and harshness considerations. Applicability appears to be dates one small passenger, who must drive lying down. The largely a matter of expert judgment, determined on a case- 0.57-m wide, 0.61-m high, 2.78-m long carbon-fiber body by-case basis. The applicability step reduces the full set of has no room for cup holders, not to mention air conditioning. technologies to only those that can be used on the baseline It is a zero-emission vehicle, but meeting safety standards vehicle being considered. was not a design consideration. Clearly much of the PAC Car II’s fuel economy was achieved by making unaccept- Estimating Fuel Economy and Cost Impacts able tradeoffs with other vehicle attributes. The CAFE law requires that fuel economy standards must be technologically Fuel consumption impacts are estimated for each technol- feasible and economically practicable. This is ultimately a ogy and each class of vehicles (or each individual vehicle) matter of expert judgment, yet there is remarkable agreement to which it is applicable. Practitioners of the PDA method among diverse studies on the list of relevant technologies. derive their estimates from a variety of sources. Unlike Most assessments assume no reduction in size or power-to- FSS, the PDA method, by itself, is not able to produce fuel weight ratios as a premise. consumption impact estimates for individual technologies. It In general, studies of fuel consumption potential intended is a method of aggregating the fuel consumption impacts of to inform the regulatory process and using the PDA method various technologies and must obtain the individual technol- ogy benefit estimates from other sources. In its report to the committee, EEA cited three principal sources of information on fuel economy benefit. 3 The CAFE guidance states that fuel economy standards should be set at the maximum feasible level, taking into consideration technological First, the trade press, engineering journals and technical feasibility, economic practicability, the effect of other federal motor vehicle papers presented at engineering society meetings provide standards on fuel economy, and the need of the nation to conserve energy detailed information on the types of technologies available (Motor Vehicle Information and Cost Saving Act, Title V, Chapter 329, to improve fuel economy and the performance, when ap- Section 32902[a]). plied to current vehicles. Second, most of the technologies 4 Details about the competition, the car, and its design can be found at http://www.paccar.ethz.ch/.

OCR for page 118
125 MODELING IMPROVEMENTS IN VEHICLE FUEL CONSUMPTION automatic transmission and a CVT at the same time) must considered in this report have been introduced in at least a few vehicles sold in the marketplace, and actual test data on also be taken into account. fuel economy can be used. Third, the world’s largest auto- manufacturers have research and development staff with Accounting for Synergies detailed knowledge of the attributes of each technology, and their inputs in an unconstrained situation can be used Undoubtedly the most serious criticism of the PDA method to estimate the benefits of technologies. (EEA, 2007, p. 9) is that it does not adequately account for synergies among fuel economy technologies. Whether or not the PDA approach is The EPA has provided a similar list of sources of information. capable of appropriately accounting for synergies is one of the key issues addressed by the present committee. These data sources included: vehicle fuel economy certifi- Fuel economy technologies can have both positive and cation data; peer reviewed or publicly commented reports; negative synergies (see footnote 1). In addition, the impacts peer reviewed technical journal articles and technical papers of technologies applied to vehicle subsystems could poten- available in the literature; and confidential data submissions tially be significantly nonlinear, and therefore the effects of from vehicle manufacturers and automotive industry compo- multiple technologies might not be accurately estimated by nent suppliers. (EPA, 2008a, p. 2) summing the effects of the individual technologies. Practitio- The EPA considers the vehicle certification test data to ners of the PDA method draw on three sources of information be an especially reliable source when a directly comparable to estimate such synergistic effects (EEA, 2007). Because vehicle is offered with and without a specific technology. In most of the technologies under consideration are in use in addition, the NHTSA’s staff has access to proprietary data some mass-produced vehicle, it is occasionally possible to provided by vehicle manufacturers to directly support the find models using a combination of several technologies. rulemaking process. Comparing the actual fuel consumption performance of these Recently, FSS models have been extensively used to vehicles to an estimate based on the sum of their individual estimate the fuel economy impacts of individual technolo- effects can provide an estimate of the degree of synergy. gies and combinations of technologies (e.g., Ricardo, Inc., Second, simplified lumped parameter models of vehicle 2008a,b; Sierra Research, Inc., 2008). A study done by energy use (e.g., Sovran and Bohn, 1981) provide a means Ricardo, Inc., for the committee and described below indi- of avoiding the double counting of energy savings. Given cates that data on technologies’ main and synergistic effects a few key parameters, lumped parameter models allow the generated by FSS models can be used effectively in PDA quantification of sources of energy loss and the components analyses (Ricardo, Inc., 2009). of tractive force requirements for a vehicle. By attributing the impacts of technologies to specific energy losses and tractive force requirements, analysts can check that the sequential ap- Sequencing Implementation plication of technologies has plausible impacts on the factors Sequences for implementing fuel economy technologies determining energy use. A key question is whether the use are usually determined by a combination of cost-effectiveness of a lumped parameter model can sufficiently accurately ac- and engineering considerations. All else equal, it would be count for synergistic effects or whether the FSS method must economically efficient to implement first the technology that be used in all cases (Hancock, 2007). An analysis of this sub- offered the greatest reduction in fuel consumption per dollar ject by Ricardo, Inc. (2009) commissioned by the committee, of cost, followed by the technology with the second largest together with an assessment by the EPA considered below, ratio, and so on. Engineering considerations may dictate a indicates that a reasonably accurate accounting is possible. different sequence, however. For example, VVT for both The ability of lumped parameter models to accurately intake and exhaust must come after VVT for intake only, predict vehicle fuel use was first demonstrated by Sovran and regardless of cost-effectiveness. Bohn (1981). In an updated version of the same methodology, Fuel consumption benefits must then be converted to Sovran and Blaser (2003) showed that despite major changes incremental benefits, given the implementation sequence. in automotive technology, lumped parameter models still For example, the benefit of a 6-speed transmission must be predicted tractive energy requirements with a high degree of defined as incremental to that of a 5-speed transmission, accuracy. Development of a lumped parameter model begins even if the base vehicle has a 4-speed, assuming that the with the fundamental physics equations that determine the 5-speed will be implemented before the 6-speed.5 Obvious energy requirements of vehicles over fixed driving cycles, incompatibilities (e.g., a vehicle cannot have both a 6-speed in particular the EPA urban and highway cycles (equations of the lumped parameter model are presented in Chapter 2). Any cycle can be divided into three regimes: 5 In the PDA method a leap from a 4-speed transmission directly to a 6-speed transmission would be calculated by combining the incremental 1. Times when tractive force (FTR) is required from the costs and fuel consumption effects of the 4- to 5-speed transition and the engine; 5- to 6-speed transition.

OCR for page 118
126 ASSESSMENT OF FUEL ECONOMY TECHNOLOGIES FOR LIGHT-DUTY VEHICLES 2. Times when deceleration force is greater than rolling flows and tractive requirements be maintained. As such, it is a resistance (R) and aerodynamic drag (D); and powerful tool for quantifying synergistic effects for use in the 3. Times when no tractive force is required (vehicle sta- PDA method. The lumped parameter method cannot, how- tionary or undergoing deceleration provided by R + D). ever, predict the kind of synergistic effects that occur when two or more technologies alter each other’s performance. When tractive force is required on either cycle, it must This topic is taken up in detail in the following section. equal the sum of forces required to overcome rolling resis- FSS modeling more completely represents such synergistic tance, aerodynamic drag, and inertia. The lumped parameter effects and so it is useful to compare lumped parameter and method simplifies the equation for tractive force and other FSS estimates to test the adequacy of PDA synergy estimates. equations for braking and idling modes by integrating over The U.S. EPA (2008a) used both methods to estimate the fuel the drive cycles, as explained in detail in Chapter 2 of this economy benefits of 26 technology packages applied to five report. Sovran and Blaser (2003) found that the lumped vehicle types. For most packages they found close agreement parameter model defined by Equations 2.2 and 2.3 could between the two types of estimates (Figure 8.2). The EPA’s explain the tractive energy required at the wheels and hence general conclusion was that both methods were valuable and indirectly the engine output of vehicles over either EPA test that the use of lumped parameter modeling in PDA estimation cycle with an R2 = 0.9999. gave reasonable estimates of synergies. The lumped parameter method allows changes in pump- Based on this, EPA concludes that the synergies derived from ing losses, engine friction, accessory loads, and other factors the lumped parameter approach are generally plausible (with to be related in a manner that can prevent double counting if a few packages that garner additional investigation). (EPA, done properly. It reduces the likelihood of overestimating the 2008b, p. 44) combined fuel consumption impacts of multiple technolo- gies by requiring that the laws of physics controlling energy FIGURE 8.2 EPA’s comparison of full vehicle simulation model (Ricardo, Inc.) and lumped parameter (L-P) PDA model results. SOURCE: Figure 8-2.eps EPA (2008a), Figure 3.3-1. low-resolution bitmap

OCR for page 118
127 MODELING IMPROVEMENTS IN VEHICLE FUEL CONSUMPTION Determining the “Optimal” Level of Fuel Economy In 10 cases, significant differences were found (EPA, 2008b). For Standard Car Package 1 and Small MPV Pack- Calculation of fuel economy potential and its cost can be age 1, the lumped parameter method estimated a larger fuel accomplished by algorithms that decide which technologies economy improvement. The difference was traced to the to apply and in what order, or by the use of fuel economy CVT component. The Ricardo, Inc., FSS CVT representation cost curves. The algorithmic approach relies on predefined had a lower efficiency than assumed in the lumped parameter technology implementation sequences (decision trees or model. Two other cases involved turbo-charging with engine pathways) and is the basis of the Department of Trans- downsizing. The lumped parameter model estimate was also portation’s Volpe Model (Van Schalkwyk et al., 2009) and much higher in the case of Large Car Package 6a, involv- the Energy Information Administration’s NEMS model’s ing continuously variable valve lift. In the case of Large Manufacturers’ Technology Choice Submodule (DOE/EIA, Car Package 4, the lumped parameter model estimated a 2007). The decision tree methodology is described below. large benefit, but in the case of Truck Package 10, the FSS Cost curves developed by the NRC (2002) CAFE study and model produced the higher benefit estimate. For the pack- in a number of other studies have been reviewed in Greene ages including cylinder deactivation and coordinated cam and DeCicco (2000). phasing (Large Car 16, Large MPV, and Truck 12), the FSS modeling results were consistently higher. FSS estimates A PDA Algorithm: The NHTSA’s Volpe Model were also higher for the cases involving camless valve trains (Large Car Y1, Truck X1). The EPA staff is still investigating The NHTSA’s Volpe model contains a compliance simula- reasons for the differences but had identified at least some tion algorithm that simulates the response of manufacturers cases in which the comparison between the two methods led to various forms of fuel economy standards. Data are put into to the discovery of inadvertent errors in the FSS modeling. the model describing a “CAFE scenario,” a combination of For example, EPA judged that Ricardo’s modeling of cyl- definitions of vehicles included in the program, definitions inder deactivation and coupled cam phasing was incorrect of vehicle classes, levels of fuel economy standards that because it did not account for cylinder deactivation’s effect must be met each year, and the structure of the standards. of approximately doubling brake mean effective pressure The structure comprises several elements, the mathematical (BMEP) in the firing cylinders. The EPA staff suggested that f ormulation (e.g., sales-weighted harmonic mean), the conducting both FSS and lumped parameter analysis was a functional form (e.g., footprint metric function), the classes wise strategy since the discrepancies between the two meth- of vehicles to which it applies (e.g., foreign or domestic ods had led to the discovery of correctable errors. manufacture), and provisions for trading credits over time Twenty-three of the 26 packages evaluated by Ricardo and among firms. In the description below, the focus is the were also estimated by EEA, Inc. (Duleep, 2008) for com- determination of a manufacturer’s “optimal” fuel economy parison. EEA was not able to estimate the packages includ- level for a given CAFE scenario. ing homogeneous charge compression ignition (HCCI) due The algorithm begins with a list of vehicles expected to to the novelty of the technology. The FSS method requires be available during the future period being evaluated. This an externally provided representation of the physics of a is typically a narrow window of three to five model years, device in order to estimate its impact on fuel consumption. beginning 2 years in the future. Vehicles are distinguished While the FSS method itself cannot characterize the physics by make, model, engine, and transmission, as in the EPA test of technologies, it can produce impact estimates given such car list. Many other vehicle attributes are in the vehicles data characterizations. The PDA method, on the other hand, must base, including sales volumes, prices, and specifications. The be given estimates of impacts for novel technologies. In 16 compliance algorithm applies technologies to each vehicle of the 23 comparisons the two methods produced estimates in the database individually. In the past, the technologies with relative differences of less than 5 percent. In two cases were largely taken from the NRC 2002 report’s three tech- involving CVT transmissions the Ricardo estimate was much nology path lists, but for the 2011 Fuel Economy Rule, the lower. In the committee’s discussions with Ricardo and EEA, NHTSA developed a new technology list with the assistance it was determined that this was due to Ricardo’s estimated of Ricardo, Inc. The new list adds diesel and hybrid power efficiency of the CVT being much lower than EEA’s. This in- trains (including plug-in hybrids) and materials substitution stance illustrates how both methods depend on assumptions to reduce vehicle weight. It represents other technologies at about the performance of key technologies. In the remaining a greater level of detail. It also provides a table of estimated five cases, Ricardo’s FSS estimates were higher but there ap- pair-wise synergies between technologies. However, the peared to be no common technology that could explain the synergies used in the final rule appear to be the same for differences. One of these cases was again the Truck Pack- all vehicles classes. The analysis done for the committee by age 10 involving a turbo-charged gasoline direct injection Ricardo, Inc., described below, indicates that synergy effects engine: EEA’s lumped parameter PDA method estimated a can vary across applications to different classes of vehicles fuel economy benefit of 26.4 percent, whereas the Ricardo (Ricardo, Inc., 2009). estimate was 42 percent.

OCR for page 118
128 ASSESSMENT OF FUEL ECONOMY TECHNOLOGIES FOR LIGHT-DUTY VEHICLES The algorithm evaluates the applicability of each tech- The algorithm then begins an iterative process of deter- nology to each individual vehicle based on timing of avail- mining a manufacturer’s compliance with the CAFE stan- ability and whether or not it is included in decision trees dards. If a manufacture is not in compliance, the algorithm selects the next-best technology to add to the vehicle.6 A for that vehicle class. The Volpe model’s decision trees are analogous to the 2002 NRC study’s “paths” except that there technology is selected from the next steps on each of the are separate decision trees for internal combustion engines, applicable decision trees. The single technology that has transmissions, electrical accessories, material substitution, the lowest “effective cost” is chosen for implementation. dynamic load reduction, aerodynamic drag reduction, and Effective cost is defined as the total retail price equivalent hybrid electric technology. The engine technology decision (RPE) cost of implementing the technology (the change in tree is shown in Figure 8.3. After low-friction lubricants and RPE times the number of vehicles affected), plus any change engine friction reduction are accomplished, the tree splits in the manufacturer’s potential CAFE fine, minus the total into three paths depending on camshaft configuration. This discounted value of fuel saved by the increase in fuel econ- allows the NHTSA to tailor the technology sequencing to the omy, all divided by the number of vehicles affected. Fines base vehicle’s engine attributes. If fuel economy is pushed to are calculated so as to take account of credits for exceeding higher levels the three paths then converge on the stoichio- standards on some vehicles that can be transferred to other metric, gasoline direct-injection engine. A table of notes can vehicles. Some manufacturers are assumed not to be willing be used to “override” the algorithm’s logic and determine to pay fines and so for them that option is removed. The cur- applicability in special cases (e.g., as in Table 4, DOT, 2005). rent version of the model calculates credits or deficits (nega- In the committee’s judgment, it is not necessary to have tive credits) generated by exceeding or failing to meet the separate decision trees for engines and transmissions. This standard in any given year. It does not, however, attempt to view is supported by the Ricardo, Inc. (2009) analysis, model credit trading either within a manufacturer over time which demonstrates that the important across, or inter- or among manufacturers. The algorithm continues consider- decision-tree, synergies are between engines and transmis- ing and implementing next-best technologies for all vehicle sions (Ricardo, Inc., 2009). These inter-tree synergies can classes until a manufacturer either achieves compliance with be transformed to incremental improvements by combining the standard, exhausts all available technologies, or finds engines and transmissions into a single power train decision that paying fines is more cost-effective than increasing fuel tree. Once this has been done, nearly all important synergies economy (Van Schalkwyk et al., 2009, p. 2). can be addressed by adjusting technology impacts to account In a joint EPA and NHTSA (2009) notice of proposed for interactions with technologies previously implemented in rulemaking (NPRM) the EPA introduced its optimization the decision tree, or pathway. model for reducing emissions of greenhouse gases from In the Volpe model, the cost and fuel economy impact a uto mobiles (OMEGA) model. Like the Volpe model, of each technology vary by vehicle class. Previously the OMEGA is based on the PDA method and although the logic 10 vehicle classes of the 2002 NRC report were used, but of the two models is fundamentally the same, there are some the 2011 rule is based on 12 vehicle classes that include 4 notable differences. The Volpe model operates on individual performance-based classes: vehicle configurations (on the order of 1,000 make, model, engine, and transmission combinations), taking into account 1. Small light truck (including SUVs and pickups), the existing or planned use of fuel economy technologies 2. Midsize light truck (including SUVs and pickups), on each one. The OMEGA model deals with approximately 3. Large light truck (including SUVs and pickups and 200 vehicle platforms broken down by engine size (EPA and full-size vans), NHTSA, 2009). For the purpose of estimating technology 4. Minivans, impacts the 200+ platforms are divided into 19 vehicle types 5. Subcompact cars, that attempt to distinguish among power trains and market 6. Subcompact performance cars, intent. Each of the 19 vehicle types are grouped into five 7. Compact cars, vehicle classes (small car, large car, minivan, small truck, 8. Compact performance cars, and large truck) for the purpose of scaling cost estimates. In 9. Midsize cars, general, the EPA’s baseline vehicle is defined as one with a 10. Midsize performance cars, port-fuel-injected, naturally aspirated gasoline engine with 11. Large cars, and two intake and two exhaust valves and fixed valve timing and 12. Large performance cars. lift, and a 4-speed automatic transmission. For the NHTSA’s Volpe model the baseline is the actual configuration of each The sequence in which the technologies are applied to any given vehicle is determined by an optimization algo- 6 The rithm. Technologies already in use in a given vehicle are Volpe model allows manufacturers to opt for non-compliance if paying a fine is less costly than missing the standards, and if a switch set in “carried over” from the previous year so that they are not input data files allows such non-compliance. This option is not discussed duplicated. here for the sake of brevity.

OCR for page 118
129 MODELING IMPROVEMENTS IN VEHICLE FUEL CONSUMPTION FIGURE 8.3 Volpe model engine technology decision tree. Figure 8-3.eps bitmap

OCR for page 118
130 ASSESSMENT OF FUEL ECONOMY TECHNOLOGIES FOR LIGHT-DUTY VEHICLES vehicle configuration as it exists or is predicted to exist in −1   the baseline fleet. −1  Nk s     nj The Volpe model applies individual technologies one at a ∑ j  MPG j  1 + ∑ ∆ ij    Equation 1  j =1 N k   time in a sequential algorithm, whereas the OMEGA model    ∑ sj   i =1 applies predefined packages of technologies that have been   j =1 ranked by cost-effectiveness for each vehicle type. However, the packages are assembled from individual technology im- If the calculation is done in terms of fuel consumption, pact estimates, with synergies between technologies within a or gallons per mile (GPM), the corresponding equation for package incorporated in the technology package impact esti- the manufacturer’s fuel consumption target is the following: mates. The EPA used the lumped parameter method to deter- mine the adjustment factors (EPA and DOT, 2009, p. 171). Because neither the Volpe CAFE Compliance and Effects sj   nj ) ( Nk ∑ Nk GPM j ∏ 1 − δ ij  Equation 2 Modeling System nor the EPA’s OMEGA model make use of   ∑ sj   j =1 i −1 cost curves but rather employ computer algorithms, neither NHTSA nor EPA require cost curves but rather a list of fuel j=a economy technologies including cost, applicability, and Equations 1 and 2 make two strong assumptions. First, synergy estimates. This committee’s method is based on they assume that the relative fuel consumption impact of a implementation pathways that are analogous to the Volpe technology will not vary from vehicle to vehicle. Because model’s decision trees and the OMEGA model’s packages. impacts will vary depending on the initial design of each Therefore, this committee determined that it was not neces- vehicle, some error will be introduced for each vehicle. sary for this study to produce cost curves as such. In addition, it is assumed that, for a given implementation sequence, any interactions (synergies) among technolo- Aggregating to Estimate Manufacturers’ Fleet Average gies have already been accounted for in the ∆ or δ terms. Fuel Economy Given information on technology synergies generated by FSS models, equations 1 or 2 could be modified to include Because fuel economy standards are enforced on auto- synergistic effects as each technology is added. Summing mobile manufacturers, both the FSS and PDA methods relative fuel economy increases as in equation 1 produces require a means of inferring the fuel economy potential a smaller estimate than sequentially multiplying one plus of an OEM from the fuel economy potential of individual the relative fuel economy increases. Most fuel economy vehicles or vehicle classes. The FSS method is sufficiently impact estimates have been determined with the expectation computationally intensive that it has not been practical that they will be added to obtain the overall fuel economy to carry out simulations for all thousand or so vehicles in benefit. Likewise, multiplying the terms in equation 2 will the EPA test car database for all relevant combinations of produce a smaller estimated change in fuel consumption than technologies. Using the PDA method, a manufacturer’s fuel adding the δi, which could erroneously lead to negative fuel economy potential can be calculated using data on individual consumption. In either case, adding fuel economy impacts or configurations (make, model, engine, transmission, i.e., a multiplying fuel consumption impacts is intended to produce single entry in the test car database) or using data on classes an approximation to the true impact in a way that reduces the of vehicles. The NHTSA’s Volpe model, for example, calcu- chances of overestimating fuel consumption benefits. lates a manufacturer’s fuel economy target using individual vehicle data since each vehicle has its own fuel economy target as a function of its footprint. The model also calculates Aggregation over Vehicles in a Class each manufacturer’s fuel economy potential at the test car list The PDA method can be applied to an individual vehicle level of detail. Estimates based on vehicle classes can also or to a representative vehicle (e.g., a midsize passenger car). be computed but they will only be approximately equal to For an individual vehicle, it is necessary to know the existing estimates based on individual configurations. technology makeup of the vehicle so that incompatibilities Assume that the optimal level of fuel economy for a single are avoided and technologies are not applied twice. In the vehicle configuration j has been determined to include tech- case of a representative vehicle, it is necessary to know the nologies k = 1 to nj (given a technology implementation se- market shares of fuel economy technologies for vehicles in quence and fuel economy impacts adjusted for implementation its class. In general, the exact distribution of all combina- order and synergies). The cumulative fuel economy impact tions of technologies within the vehicle class is not known. is calculated by summing the fractional fuel economy (miles Instead, the total market shares of each technology are used, per gallon) improvements, adding one, and multiplying by the in effect assuming that their distributions are independent. base fuel economy MPG0j. If the sales of vehicle configura- This introduces a further element of approximation into the tion j are sj, then the fuel economy for manufacturer k selling estimation. configurations j = 1 to Nk is the following:

OCR for page 118
131 MODELING IMPROVEMENTS IN VEHICLE FUEL CONSUMPTION the vehicle (mass, frontal area, drag coefficient, etc.), the Let sij,0 be the initial market share of technology i in the components that compose the driveline (engine and trans- vehicle class j, and let sij,max be the maximum market share mission, etc.), accessories (pumps, fans, generators, etc.), for technology i. The estimated change in fuel economy and a specification of the drive cycle, or vehicle speed trace, (MPG) by application of the full set of technologies is given the vehicle is to perform. Components are represented by by equation 3: computer modules and may be described by performance nj ) ( D j max = ∑ sij ,max − sij ,0 ∆ ij maps represented by tables or equations. All energy flows Equation 3 among components are accounted for by equations link- i =1 ing the modules. FSS models may be backward-looking or The estimated change in fuel consumption by application forward-looking. Backward-looking models assume that of the full set of technologies is given by equation 4: the drive cycle’s velocity and acceleration trajectory will be met, calculate the force required at the wheels, and then  nj  )) (( d j max = GPM  ∏ 1 − sij ,max − sij ,0 δ ij  work backward to the resulting engine speed, and the neces- Equation 4  i =1  sary throttle and brake commands. Forward-looking models choose throttle and brake commands in order to achieve the The cost of the above fuel economy increase is calculated specified trace. Some models use a combination of both similarly, where Ci is the cost of technology i in retail price strategies (see, e.g., Markel et al., 2002). equivalent: Modeling can have the potential benefit of helping one to nj ) ( understand these synergies and better predict future perfor- C j max = ∏ sij ,max − sij ,0 Cij Equation 5 mance, either through the careful analysis of available vehi- i =1 cle data, or through creating dynamic models of the vehicles and analyzing the performance of these virtual vehicles. In Although equation 3 approximates the share-weighted addition to the synergies within various subsystems of the harmonic mean change in fuel economy for a class of ve- vehicle, many subsystems within the vehicle exhibit non- hicles with a mixture of technologies it does not precisely linear behavior. Considering the performance of individual equal it. Even performing the calculations in terms of fuel subsystems independently, even if this performance is well consumption, as in equation 4, will not produce the exact known and understood, can therefore result in misleading harmonic mean fuel economy, in general. conclusions for the overall system. When an understanding of each subsystem can be represented by a computer model MODELING USING FULL SYSTEM SIMULATION to an appropriate level of detail, and the interconnectivity or physical communication between each of these subsystems The FSS approach to modeling vehicle fuel consump- can also be adequately represented, the synergistic and non- tion involves capturing the physics or characteristics of linear effects can be included and analyzed in the behavior subsystems of the vehicle in software, assembling these sub- of the entire system. Computer modeling of vehicle systems systems by passing relevant operational variables between is widely used in the industry for this purpose, as well as these subsystems, and choosing preferred input variables and to help predict future performance or performance under trajectories to simulate desired vehicle operation. The overall various conditions. Manufacturers use FSS in the product goal is to have the subsystem models work in a synergistic development process to optimize factors such as shift logic way to reflect the actual performance of the vehicle in vari- and final drive ratio. ous maneuvers. Because of the complexity and nonlinearity For new technologies not implemented in any mass- of these vehicle subsystems, it is often difficult to anticipate produced vehicle, FSS model results are probably the most the synergistic effects, especially during transients, and this reliable source of estimates of synergistic effects. His- approach usually provides this useful information to some torically, the PDA approach has generally not been used for degree of accuracy. FSS modeling has been used by the estimating the fuel consumption impacts of novel vehicle automotive industry since the 1970s, and is a proven method systems for which there are no actual test data (Greene and of estimating the impacts of existing and new technologies DeCicco, 2000). Today FSS modeling is more widely used on vehicle systems (Waters, 1972; Blumberg, 1976). More to estimate the potential for reducing fuel consumption than recently, regulatory agencies and other groups outside the even 5 years ago. A number of studies are available that automotive industry are undertaking efforts to develop and have used FSS to estimate the fuel consumption impacts of utilize FSS in their analysis (NESCCAF, 2004; Rousseau, advanced technologies (e.g., Ricardo, Inc., 2008a,b, 2009; 2007; EPA, 2008a). Kasseris and Heywood, 2007; Kromer and Heywood, 2007; Although modeling approaches differ, all FSS models are Sierra Research, 2008). It should be noted, however, that based on the time integration of Newton’s second law (i.e., F = m⋅a) over some driving maneuver, in this case over the sufficient knowledge of the technology package being in- vestigated is necessary to allow its representation within the FTP and highway driving cycles. The boundary and initial model to have an acceptable degree of accuracy. For an ag- conditions for this integration are based on a description of

OCR for page 118
132 ASSESSMENT OF FUEL ECONOMY TECHNOLOGIES FOR LIGHT-DUTY VEHICLES gregate technology, this may take the form of a performance Unfortunately, data on the predictive accuracy of FSS map describing its efficiency over a range of operating condi- models are scarce. In part this is because some models tions. For a technology described by unique operation of an and more often the representation of their components are existing subcomponent, relevant performance insight in the proprietary to firms that use them in their own research or corresponding new regime of operation would be necessary. consulting. The committee is not aware of any rigorous study It is important to note that, although FSS models have evaluating the accuracy of models for various applications. the ability to estimate the absolute impacts of vehicle The few comparisons the committee has seen indicate that technologies due to their ability to model the physics of for known vehicles, simulation models can reproduce fuel system components, they have limited ability to model the consumption and performance with a high degree of accu- dynamic working of individual fuel efficiency technologies racy. Data provided by Ricardo, Inc., based on its research and generally rely on a limited set of input data. For novel for the EPA indicated a range of error in predicting fuel technologies, many of the input parameters are assumptions consumption of 1 to 3 percent for five vehicles (Figure 8.4). based on engineering judgment and experience with related For this modeling, the EPA chose a specific representative technologies. This emphasizes the fact stated at the outset vehicle for each of the five classes: the Toyota Camry for the of this chapter, that one cannot know with absolutely ac- standard car, the Saturn Vue for the small MPV, the Chrysler curacy the impact of technologies until an actual vehicle is 300 for the full-size car, the Dodge Grand Caravan for the constructed and repeatedly tested. large MPV, and the Ford F150 for the truck. Ricardo, Inc., (2008a) attributed any discrepancies between the simula- tion results and the actual vehicle data to the use of generic Model Fidelity input data for that vehicle class instead of actual data for a An important consideration for FSS modeling is deciding specific vehicle. Of course, these are known vehicles so that what level of fidelity of the equations or look-up tables is component representations and the overall model can be required for the problem being addressed. No set of equa- calibrated. Prediction errors for truly novel technologies for tions completely reflects the detailed physics of the actual which no vehicle exists to calibrate to would presumably be process, so the choice of fidelity should be a conscious choice larger. In any event, it is the change in fuel consumption from from a continuum of models of varying fidelity, all of which the implementation of a technology that is of most interest. represent simplifications of the actual process. The objective The absolute error of a predicted change can be smaller when is to achieve an appropriate balance of fidelity with modeling prediction errors similarly exist in both the “before” and goals, modeling effort and resources, simulation speed, and “after” simulations (i.e., the modeling errors of the before available data that specifically characterizes the system being and after cases are strongly correlated). Still, relative errors modeled. There is always a difference between the simula- for a predicted change are likely to be greater. The accuracy tion and actual subsystem operation, known as the modeling of FSS models in predicting fuel consumption changes in error. The tolerable level of error depends upon the goals of actual vehicles deserves additional study. Note that such an the simulation. accuracy study is made more difficult by the fact that the FIGURE 8.4 Comparison of actual vehicle combined fuel economies and Ricardo simulated fuel economies for five vehicles. SOURCE: Ricardo, Inc. (2008a). Figure 8-5.eps bitmap

OCR for page 118
133 MODELING IMPROVEMENTS IN VEHICLE FUEL CONSUMPTION accuracy of FSS estimations depends significantly on the tions. One set of data can be used to determine parameters experience and skill of the FSS practitioner. or tune the subsystem models, and a separate and distinct The flexibility, rigor, and comprehensiveness of the FSS set of data can be used to test the predictive capabilities of approach to vehicle modeling are significant advantages. the model in different situations after it has been tuned or Subsystem models may be as simple as a single parameter calibrated. The model should not be tested using the same or table based on steady-state operation, or a detailed, non- set of data that was used to calibrate the model. linear, multivariable representation of the dynamics of the subsystem, including transient operation. The choice of FSS Model Example how to represent each subsystem model is not only based on modeling error considerations discussed above, but also An example of an FSS compression-engine model is on balancing fidelity between subsystem models, in order to illustrated in Figure 8.5 in order to give the reader a better use computational resources as effectively as possible. One visual idea of a possible subdivision of subsystems within way of looking at balancing fidelity between subsystems the overall system model, as well as possible choices of is to consider the filtering properties or bandwidth of these fidelity within each subsystem. The overall goal of this model subsystems. If one subsystem model has a level of fidelity is to represent engine transient performance within the ve- that generates details in an output variable that are filtered hicle power train, including cylinder-by-cylinder rotational out by a subsequent system, then the effort in generating dynamic effects as well as first order intake and exhaust those details is mostly wasted if the intermediate variables dynamics that affect turbocharger transient effects on the between the subsystems are not of interest. This balance of engine. Some simple emission transient predictive capabil- fidelity within an overall FSS model is a judgment call that ity is included but is not comprehensive for all constituents. is typically developed through experience or trial and error, This model was developed using the MATLAB/Simulink although the effects can be clearly seen by looking care- modeling software, and its overall structure is presented by fully at the content of the variables that are passed between the block diagram structure of MATLAB/Simulink in hier- subsystems to see what effects are preserved or eliminated. archical form. Most of the subsystem models are identified An example of these considerations can be seen by ex- for the reader. The core of the model is the engine map that amining a typical system model of a turbocharger. In many provides brake-specific fuel consumption as a function of en- dynamic system models, the characteristics of both the gine speed and load. Numerous other modules are necessary turbocharger compressor and turbine are simulated based to represent the many interacting components of the engine on steady-state maps. However, the rotational dynamics system. Most of these components must be calibrated to the of the rotor is simulated based on Newton’s second law specific engine system of interest. (i.e., a differential equation reflecting dynamic or transient operation). The rationale for choosing and combining these AN ANALYSIS OF SYNERGISTIC EFFECTS AMONG two different types of models is based on the idea that the TECHNOLOGIES USING FULL SYSTEM SIMULATION time constants for the gas dynamics in the compressor and turbine are considerably shorter (i.e., faster) than the time At the request of the committee, Ricardo, Inc. (2009) constant of the rotor. If much more detailed dynamic models undertook a study to quantify the synergistic effects captured of the gas dynamics were included in the model when the by FSS models. It is important to note that the study is based rotational speed of the rotor is the desired output variable, solely on the predictions of Ricardo, Inc.’s FSS models and almost all of the gas dynamic effects would be filtered out therefore can quantify only the synergies those models can by the rotor inertia or rotational bandwidth. This combina- represent. In its report, Ricardo estimated the accuracy of its tion of steady-state and dynamic models to represent the models for predicting fuel economy at 1 percent for well- turbocharger usually provides an effective dynamic model characterized vehicle systems (systems for which nearly all of its rotational dynamics and transient operation in relation model subsystems have been calibrated to the actual com- to the rest of the engine. However, if the goal is to capture ponents) and 3 percent for novel vehicle systems. However, the pulsed gas dynamics in the turbine or compressor, this each estimate of accuracy was based on a single data point choice of subsystem models may not be appropriate (for and so cannot be considered definitive. that specific goal). The important point is that more detail Ricardo’s approach was to simulate the technologies is not necessarily better, but fidelity and balance should be contained in five different packages of technologies it had conscious decisions reflecting modeling goals. previously modeled for the EPA (2008a) as applied to five different types of vehicles. The technologies were applied one at a time and in combinations according to a rigorously Model Validation defined design of experiments. The results were then fitted by An effective way of carrying out model validation, given a response surface model using a neural network method. The available data on the system operation, is to subdivide the response surface model fit the data with maximum errors of 1 data into at least two sets covering different operating condi- percent using terms no higher than second order (Figure 8.6).

OCR for page 118
134 Powertrain Control Research Laboratory Engine module The University of Wisconsin -Madison Powertrain System Simulation with Cylinder -by -Cylinder Model Copyright (c) 1999, All R ights Reserved After treatment system module rpm_avg Crankshaft speed Ambient Dynamic driveline model Conditions pressures 0 Backpressure Reservoir mdot_fuel Conditions CA_inj Turbocharger speed calculations feedgas Engine Product eng power After treatment system Engine 0 time Load Torque Clock Compressor and turbine data cont xdot_veh Selector airflow mair To Workspace Controller flow_turb_in Rotor torque and flow calculations Goto flow_turb_out fresh air mass Driveline and Ambient flow_comp_in vehicle dynamics 1 Atmospheric Load Vehicle Conditions Specific Data Turbocharger mdot_table Cushioning double -click before Manifold Crankshaft dynamic calculations starting simulation flow_filt_out state_man Backpressure Air Filter CAs 2 Atmospheric Conditions Crank angle-varying inertia calculations Phased CAs RPMs torq_inds landscape Air filter flow 2 flow_comp_out 3 P_cyls 4 1 fuel Figure 8-6.eps and pressure drop torq_load N_eng kg/cycle Outlet manifold state calculations Rigid Crankshaft calculations flow_im_in Intercooler state_exhaust_man flow_em_out exh temp q flow_intake_manifold_out EGR Mux 3 T_wall T_exh exhaust flow_im_out 6 Cylinder Engine Manifolds co,nox state_cylinders state_intake_man m_exh 1 flow_im_in Manifold pressure CAs RPMs flow_em_in state_im 3 state_em rpmeng calculations P_im T_im Goto1 Single 3 4 phi_im Mux Inlet mdot_im_in flow_em_out Manifold Hdot_im_in Mux1 phi_im_in Intake Manifold 4 flow_exhaust_manifold_in state_exhaust_manifold 0 Intake Manifold 2 6 Heat flow P_cyls exhaust composition Exhaust manifold Intercooler flow and 1 torq_cyls efficiency calculations 2 5 fuel kg/cycle torq_cyls1 CA_inj Q_cyls 5 T_wall EGR calculations 6 Cylinders Intake manifold state calculations FIGURE 8.5 An example structure for a full system simulation diesel engine dynamic model. SOURCE: Reprinted with permission from John J. Moskwa, Powertrain Control Research Laboratory, University of Wisconsin, Madison, Wisc. ASSESSMENT OF FUEL ECONOMY TECHNOLOGIES FOR LIGHT-DUTY VEHICLES

OCR for page 118
135 MODELING IMPROVEMENTS IN VEHICLE FUEL CONSUMPTION fact that when technologies are applied in sequence the fuel consumption impact of a technology depends on which tech- nologies have been previously applied. For example, given a base vehicle with a 4-speed transmission, the impact of a 6-speed transmission will be smaller if a 5-speed transmis- sion has been previously applied. The PDA method explicitly recognizes this kind of interaction by ordering technologies for interaction and using only incremental impacts, given that ordering, to estimate the total fuel consumption impact. But incremental effects, as defined in the Ricardo ANOVA, also include true synergistic effects, such as when a 42-volt electrical system is implemented together with electric ac- cessories (e.g., electric power steering). Most PDA model- ers attempt to take such interactions into account, but the accuracy with which they do so will depend on the available FIGURE 8.6 Ricardo, Inc., statistical (response surface model Figure 8-7.eps [RSM]) predictions versus full system simulation model predic- data sources and the engineering judgment of the analyst. bitmap tions. SOURCE: Ricardo, Inc. (2009), Figure 3-2. There are additional synergies of interest that Ricardo terms “inter-tree” or “true” synergies. These are the interac- tions among technologies that are neither second-order main effects nor incremental effects. PDA modeling cannot, in This shows that a relatively simple 2nd order regression general, account for this type of synergy. According to the model provides a very suitable representation of the more results of Ricardo’s study, these effects are quite small. For complex vehicle simulation output with maximum RSM example, adding up the synergy (inter-tree) effects for Small (Response Surface Model, ed.) residual errors of about MPV Package 5 (allowing positive and negative effects to 1 percent, or that higher order effects (3rd order and above) cancel) results in a total synergy effect of −1.3 percent of account for less than 1 percent of the vehicle simulation the total fuel economy impact of the technology package. output characteristics. (Ricardo, Inc., 2009, p. 13) Adding up the inter-tree synergies produces a positive syn- ergy of 4.6 percent for Small MPV Package 15, a positive This finding is significant in that it indicates that important synergy of 2.8 percent for Large MPV Package 16, and a synergistic effects (as represented by the FSS models) are of positive synergy of 10.3 percent of the total fuel economy no higher order than two-way interactions. It is also gener- impact for Truck Package 11. These are percentages of the ally consistent with the ability of a much simpler lumped total fuel economy change and so suggest that errors due to parameter model to accurately estimate fuel economy over completely ignoring inter-tree synergies are on the order of the federal test cycles with Sovran and Blaser (2006). 10 percent or less for the total fuel economy impact. The size The next step was to carry out an analysis of variance of these effects is roughly consistent with the discrepancies (ANOVA) to quantify the first-order (main) and second-order EPA (2008b) found in its comparison of lumped parameter effects. The ANOVA estimated that main effects of tech- and FSS modeling. nologies accounted for 80 to 86 percent of the fuel economy Ricardo, Inc. (2009) concluded that PDA modeling, such increase. Interaction effects, taken together, accounted for as that used in the NHTSA’s Volpe model, if informed by 14 to 20 percent. Ricardo, Inc., concluded that simplified rigorously designed FSS modeling of the kind represented in models that did not properly account for interaction effects its study, can produce accurate estimates of fuel consumption could have estimation errors of up to 20 percent. However, 20 reduction potential. This conclusion, however, is conditional percent not only is the upper bound on estimation error but on the accuracy of FSS models for predicting EPA test cycle also assumes that the error in estimating interaction effects fuel economy. Given the scarcity of evidence on this subject is 100 percent (for example, if they were all estimated to and its importance to validating Ricardo’s conclusion, it be 0). Interaction effects estimated using lumped parameter merits further investigation. models, for example, are likely to be much smaller. Even more importantly, the interaction effects include second-order main effects and incremental effects. Second- FINDINGS order main effects represent the interaction of a technology Finding 8.1: The state of the art in estimating the impacts of with itself and are introduced to account for nonlinear effects fuel economy technologies on vehicle fuel consumption is in the linear ANOVA model. Thus, they do not depend on full system simulation (FSS) because it is based on integra- the presence or absence of other technologies and so are tion of the equations of motion for the vehicle carried out not synergies in the sense that is of interest. Incremental ef- over the speed-time representation of the appropriate driv- fects include some true synergistic effects and some purely ing or test cycle. Done well, FSS can provide an accurate incremental effects. Purely incremental effects reflect the

OCR for page 118
136 ASSESSMENT OF FUEL ECONOMY TECHNOLOGIES FOR LIGHT-DUTY VEHICLES assessment (within +/-5 percent or less) of the impacts on demonstrated a practical method for using data generated by fuel consumption of implementing one or more technologies. FSS models to accurately assess the fuel consumption poten- The validity of FSS modeling depends on the accuracy of tials of combinations of dozens of technologies on thousands representations of system components (e.g., engine maps). of vehicle configurations. A design-of-experiments statistical Expert judgment is also required at many points (e.g., deter- analysis of FSS model runs demonstrated that main effects and mining engine warm-up rates or engine control strategies) first-order interaction effects alone could predict FSS model outputs with an R2 of better than 0.99. Using such an approach and is critical to obtaining accurate results. could appropriately combine the strengths of both the FSS and Finding 8.2: The partial discrete approximation (PDA) the PDA modeling methods. However, in Chapter 9 the com- method relies on other sources of data for estimates of the mittee recommends an alternate approach that would use FSS impacts of fuel economy technologies. Unlike FSS, the PDA to better assess the contributory effects of technologies applied method cannot be used to generate estimates of the impacts for the reduction of vehicle energy losses and to better couple of individual technologies on vehicle fuel consumption. the modeling of fuel economy technologies to the testing of Thus, the PDA method by itself, unlike FSS, is not suitable such technologies on production vehicles. for estimating the impacts on fuel consumption of technolo- gies that have not already been tested in actual vehicles or REFERENCES whose fuel consumption benefits have not been estimated by Blumberg, P.N. 1976. Powertrain simulation: A tool for the design and means of FSS. Likewise, the effects of technology interac- evaluation of engine control strategies in vehicles, SAE Technical Paper tions must be determined from external estimates or approxi- Series 760158. SAE International, Warrendale, Pa. February 23. mated by a method such as lumped parameter modeling. DOE/EIA (U.S. Department of Energy/Energy Information Administration). Even FSS, however, depends directly on externally generated 2007. Transportation sector module of the National Energy Modeling information on the performance of individual technology System: Model documentation 2007, DOE/EIA-M070(2007). Office of Integrated Analysis and Forecasting, Washington, D.C. components. DOT (U.S. Department of Transportation). 2005. CAFE compliance and effects modeling system. Volpe Systems Center, Cambridge, Mass., Finding 8.3: Comparisons of FSS modeling and PDA es- July 19. timation (within the range of cases where the PDA method DOT/NHTSA (U.S. Department of Transportation/National Highway is applicable) supported by lumped parameter modeling to Traffic Safety Administration). 2009. Average fuel economy standards, passenger cars and light trucks, model year 2011: Final rule, RIN 2127 eliminate double counting of energy efficiency improve - AK-29, Docket No. NHTSA 2009-0062, Washington, D.C., March 23. ments have shown that the two methods produce similar re- Duleep, K.G. 2008. EEA-ICF analysis update, Presentation to the Commit - sults when similar assumptions are used. In some instances, tee on Technologies for Improving Light-Duty Vehicle Fuel Economy, comparing the estimates made by the two methods has April 1, Washington, D.C. enhanced the overall validity of estimated fuel consumption EEA (Energy and Environmental Analysis, Inc.). 2007. Technologies to improve light-duty vehicle fuel economy, Draft report to the National impacts by uncovering inadvertent errors in one or the other Research Council Committee on Fuel Economy of Light-Duty Vehicles, method. In the committee’s judgment both methods are valu- Arlington, Va., September. able, especially when used together, one providing a check EPA (U.S. Environmental Protection Agency). 2008a. EPA Staff Techni- on the other. However, more work needs to be done to es- cal Report: Cost and Effectiveness Estimates of Technologies Used tablish the accuracy of both methods relative to actual motor to Reduce Light-Duty Vehicle Carbon Dioxide Emissions. EPA420- R-08-008, Ann Arbor, Mich. vehicles. In particular, the accuracy of applying class-specific EPA. 2008b. EPA’s technical review of Ricardo simulations, Presentation to estimates of fuel consumption impacts to individual vehicle the Committee on Technologies for Improving Light-Duty Vehicle Fuel configurations needs to be investigated. The magnitude of the Economy, March 31, 2008, Detroit, Mich. errors produced when such estimates are aggregated to cal- EPA and DOT (U.S. Environmental Protection Agency and U.S. Department culate the potential of individual automobile manufacturers of Transportation). 2009. Proposed Rulemaking to Establish Light-Duty Vehicle Greenhouse Gas Emission Standards and Corporate Average to reduce fuel consumption should also be analyzed. Fuel Economy Standards. August 24. Washington, D.C. EPA and NHTSA (U.S. Environmental Protection Agency and National Finding 8.4: The U.S. Department of Transportation’s Volpe Highway Traffic Safety Administration). 2009. Draft Joint Technical National Transportation Systems Center has developed a Support Document: Proposed Rulemaking to Establish Light-Duty Ve- model for the NHTSA to estimate how manufacturers can hicle Greenhouse Gas Emission Standards and Corporate Average Fuel Economy Standards, EPA-420-D-09-901, September. comply with fuel economy regulations by applying additional Greene, D.L., and J. DeCicco. 2000. Engineering-economic analysis of fuel savings technologies to the vehicles they plan to produce. automotive fuel economy potential in the United States. Annual Review The model employs a PDA algorithm that includes estimates of Energy and the Environment 25:477-536. of the effects of technology synergies. The validity of the Hancock, D. 2007. Assessing fuel economy. Presentation to the Committee Volpe model, and probably also the OMEGA model, could on Fuel Economy of Light-Duty Vehicles, National Research Council, September 10, Washington, D.C. be improved by making use of main effects and interaction Kasseris, E.P., and J.B. Heywood. 2007. Comparative analysis of automotive effects produced by the FSS methodology described in this powertrain choices for the next 25 years. SAE Technical Paper Series chapter. In particular, research done for the committee has No. 2007-01-1605. SAE International, Warrendale, Pa.

OCR for page 118
137 MODELING IMPROVEMENTS IN VEHICLE FUEL CONSUMPTION Kromer, M.A., and J.B. Heywood. 2007. Electric powertrains: opportunities Ricardo, Inc. 2009. A Study of Interaction Effects Between Light Duty and challenges in the U.S. light-duty vehicle fleet. Publication No. LFEE Vehicle Technologies. Prepared for the NRC Committee on Assessment 2007-02 RP. Sloan Automotive Laboratory, Massachusetts Institute of of Technologies for Improving Light-Duty Vehicle Fuel Economy by Technology, Cambridge, Mass. Ricardo Inc., Van Buren, Mich., February 27. Markel, T., A. Brooker, T. Hendricks, V. Johnson, K. Kelly, B. Kramer, Rousseau, A. 2007. Designing advanced vehicle powertrains using PSAT. M. O’Keefe, S. Sprik, and K. Wipke. 2002. ADVISOR: A systems Presentation to the Committee on Fuel Economy of Light-Duty Vehicles, analysis tool for advanced vehicle modeling. Journal of Power Sources National Research Council, September 10. 110(2):255-266. Sierra Research, Inc. 2008. Technology and retail price implications of HR6 Moskwa, J. 2008. Powertrain Control Research Laboratory, University of CAFE standards based on vehicle simulation modeling (preliminary Wisconsin, Madison, Wisc. results). Presentation to the committee by James M. Lyons, January 24. NESCCAF (Northeast States Center for a Clean Air Future). 2004. Reducing Sovran, G., and D. Blaser. 2003. A contribution to understanding automotive Greenhouse Gas Emissions from Light-Duty Motor Vehicles. March. fuel economy and its limits. SAE Paper 2003-01-2070. SAE Interna- NRC (National Research Council). 2002. Effectiveness and Impact of tional, Warrendale, Pa. Corporate Average Fuel Economy Standards. National Academy Press, Sovran, G., and D. Blaser. 2006. �uantifying the potential impacts of re- Washington, D.C. generative braking on a vehicle’s tractive-fuel consumption for the U.S., Patton, K.J., A.M. Sullivan, R.B. Rask, and M.A. Theobald. 2002. Aggregat- European, and Japanese driving schedules. SAE Paper 2006-01-0664. ing technologies for reduced fuel consumption: A review of the technical SAE International, Warrendale, Pa. content in the 2002 National Research Council Report on CAFE. SAE Sovran, G., and M.S. Bohn. 1981. Formulae for the tractive-energy require- Paper 2002-01-0628. SAE International, Warrendale, Pa. ments of vehicles driving the EPA schedules. SAE Paper 810184. SAE Ricardo, Inc. 2008a. A Study of Potential Effectiveness of Carbon Dioxide International, Warrendale, Pa. Reducing Vehicle Technologies. Prepared for the U.S. Environmental Van Schalkwyk, J., W. Gazda, K. Green, D. Pickrell, and M. Shaulov. 2009. Protection Agency, EPA420-R-08-004, Contract No. EP-C-06-003, Corporate average fuel economy compliance and effects modeling Work Assignment No. 1-14, Ann Arbor, Mich. system documentation. DOT HS 811 012. U.S. Department of Trans- Ricardo, Inc. 2008b. A study of potential effectiveness of carbon dioxide portation, Research and Innovative Technology Administration, Energy reducing vehicle technologies. Presentation to the Committee on Technology Division, John A. Volpe National Transportation Systems Fuel Economy of Light-Duty Vehicles, National Research Council, Center, Cambridge, Mass., April. January 24. Waters, W.C. 1972. General purpose automotive vehicle performance and economy simulator. SAE Paper 720043. January 10. SAE International, Warrendale, Pa.