Complex System Reliability Estimation Methodology in the Absence of Failure Data

摘要

This paper presents a comprehensive system reliability estimation methodology for cases when failure data are unavailable, at least initially. In this methodology, the laws of physics and thermal fundamentals are used to establish a mathematical model that relates the influential input operating characteristics, such as material properties and geometry, to system performance measures. Probability distributions for each influential operating characteristic, identified from the available manufacturing data, information found in instruction manuals and related technical journals, and expert knowledge, are used to simulate the system behavior with Monte Carlo simulation. An initial reliability estimate is obtained by comparing the simulated system performance with the permissible system performance. Fuzzy logic is used to incorporate the impact of environmental factors on the performance of the simulated system performance and hence the system reliability. Finally, with the use of Bayesian analysis, initial system reliability is updated to take into account the effect of environmental factors. The proposed methodology is applied to estimate the reliability of the hazardous gas detection system used in aerospace shuttles for the timely detection of explosive gases.