Efforts to predict vehicle performance probably began shortly after Karl Benz drove his first three wheeler on the streets of Mannhiem, Germany in 1885. A century later, computers have accelerated and improved the precision of the predictive process. The challenge has been to achieve the highest level of predictive accuracy for the greatest variety of vehicle configurations operating in different environments, and to do it with the greatest ease and least burden for the person needing the answer.
A computational method has been developed that meets those requirements and can be executed on a personal computer with as few as forty-seven inputs. At their discretion, users can input up to eighty-five specifications. The program is capable of generating its own values when certain specific information is not available.
The performance predictive process follows the methodology published four decades ago by J. J. Taborek [1], but adds features and refinements to accommodate the more diverse technology of current vehicles. Additionally, fuel economy is computed for city and highway driving cycles defined by the U. S. Government.
The methodology starts with a determination of tractive power, but limits the tractive force to conditions established by the tire/road interface and the chassis weight transfer dynamics. Three optional methods are presented for the determination of roadload power requirements based on the type of information available. Linear and rotational inertia are combined into a single value of equivalent mass which is then adjusted to reflect the clutch/torque converter slippage and driving wheel spin.
The predictive accuracies were validated by comparing the computed performance and economy with physical test data. Specifications for forty-five vehicles were obtained from the vehicle producers, while performance/economy data was generally obtained from independent testing agents. Statistical analysis of all compiled data evaluated the variances between computed and measured performance and economy. Average variances of the nine studied factors were generally less than one percent with standard deviations in a similar range.