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