Non-sample based parameters design for system performance reliability improvement

Design methods for quality generally help to improve quality over time, but do not consider change of system performance over time, resulting from degradation in components. As design methods for quality over time (performance reliability), which minimizes effects of unavoidable component degradations as well as component variations on system performance change, system model-based sampling methods using Monte-Carlo simulations have been used. But, there are main concerns related to computational efficiency and optimization in applying the sampling methods. To overcome the concerns, we propose a non-sample method for quality over time. Based on the proposed method, the process of allocating design parameters, which could minimize the noise effects with the consequence that both quality and performance reliability are optimized, is discussed. Reliability metrics such as mean time to failure and standard deviation of time to failure are optimized simultaneously for reliability improvement. Desirability functions for the metrics are introduced to perform the simultaneous optimization. The proposed method is applied to an electrical system design and compared to a sampling based design method.