Design of experiments for reliable operation of electronics in automotive applications

Starting from pre-sillicon verification steps, automotive electronics demand for the assurance of a high degree of reliability. A dependable system must comply with requirements even in the presence of sources of variation, which can come from internal or external influences. Therefore, one must study the impact of such variations on the responses of the system and find safe margins for them. The problem amplifies when the effects of variations and the conditions which drive the system into a worst-case change in time. Statistical Design of Experiments and metamodelling strategies are adopted and extended to cope with the problem. A reduced number of simulation runs is found sufficient to characterize time-variant effects and to predict the system's behaviour between the simulated points. Then, the bounds of the signals of interest are estimated and confirmed. The proposed methods are demonstrated on a window lifting automotive system.