Prediction of Parameters of Vehicle Dynamics Simulation Model using modeFRONTIER

Asahiko Otani (CD-adapco Japan), Nobuhiko Suzuki (ADVICS), Hisao Shibakawa (Virtual Mechanics Corp.)

Vehicle Dynamics Simulation is widely applied to many stages of vehicle development, accuracy of the simulation is strictly required. Since accurate vehicle dynamics simulation requires correct vehicle parameter value, but they are hardly be obtained in many cases, during the stages of the vehicle development system suppliers are usually in such circumstances due to measurement facilities / confidentiality. However, running vehicle behavior can be measured and logged as chronological data. Since “Data to be compared” are not simple index but chronological curve, decision criteria (good or bad etc.) they tend to be subjective, thus hard to be consistent. The procedure to predict reasonable value of each parameter could be defined only by well experienced engineer in Vehicle Dynamics Simulation field, skillful engineer has to repeat “Trial and Error”,this work to be done isn’t creative, thus it is really wasting time and money. To avoid subjective judgment, fitness of chronological data were statistically evaluated and the indexes were used as objectives and constraints. Configuring optimization problem using them, Vehicle Model Parameters have been successfully predicted. Work efficiency was fairy improved, although preparation initially took a week or more. The efficiency of similar optimization to predict model parameters will be much better because the project file can be reused so that no preparation will be needed.