Robustness optimization of heterogeneous systems in multi-objective scenarios

Systems grow in complexity but their behavior has to meet the specifications and to be robust in all or most situations and conditions. When variations affect more responses of a system, achieving robustness becomes a hard to fulfill task because different responses may demand concurrent settings for the system's factors. In this paper, we approach multi-response robustness optimization by estimating the dependence of the response distribution on some controllable factors. The estimation is made based on a number of simulations with different factors settings and it is able to predict the response distribution. We used three methods for choosing the optimal factor settings. The first uses a weighted cost function based on the importance of each response, while the last two methods constraint the factors space by imposing a restriction for each response. The methods are applied on a beam-leveling system used in automotive and proved to be successful even if the number of heterogeneous factors is greater compared with other systems used in literature. The optimal settings are validated by comparing the estimated cost functions with the simulated ones.