Abstract
Today the vehicle attribute performance especially in the fields of vehicle dynamics is an important purchase criterion for vehicle customers. Driven by an increased market penetration of hybrid-, fully-electric and autonomous vehicles, the vibration isolation of passenger compartments will gain even more importance in the future. Therefore, it is vital to balance various vehicle attributes early on in the development process to meet the customer demands of future vehicles. In the first place, advanced multi-body full vehicle simulation models are generated and validated with the aid of measurement data from physical road measurements as well as rig tests. The multi-body simulation environment (MSC Adams/Chassis) was coupled to a multi-objective optimization environment (Esteco modeFrontier). Various objective target metrics are typically defined as critical to customer satisfaction within the simulation environment. As input variables a variety of parameters like for example stiffness and damping parameters of tunable vehicle components were selected. Various design of experiments (DOE) were run to generate response surfaces based on radial basis functions. Finally, optimization studies were conducted to produce Pareto Fronts for two and more target metrics. The coupling of the multi-body simulation environment to the multi-objective optimization framework creates a powerful tool to virtually optimize vehicles for various different attributes like for example vehicle dynamics or driving comfort early on in the development phase. A huge number of input parameter variations can be assessed in relatively short time. In a first phase the results from the DOEs based on a random sequence with uniform distributions for all input parameters are used to generate scatter matrix charts to identify sensitivities between independent input variables and individual dependent responses. Next to sensitivities correlation coefficients are identified for all parameter combinations to determine positive, negative and neutral interrelations between controllable input factors. In the final optimization phase Pareto diagrams are established for two or more dependent response metrics using the non-dominated sorting genetic algorithm II (NSGA-II).