Technique of Detecting the Degradation of Failure Based on Cross-Entropy

Abstract：To solve the problem of electronic equipment test and failure diagnosis, we propose a degradation failure detecting algorithm based on Cross-entropy. First, we choose appropriate degradation model as trajectory based on characteristic of electronic equipment and obtain distribution curve over time with degradation data and the initial of data using cross-entropy algorithm, and by establishing the relationship between the probabilities of overrun, the threshold of the failure is determined. Finally, taking +12V power supply as test object, we make simulation by generating random samples of time-varying value: The cross-entropy method can accurately detect equipment failure and the aging of the slowly changing, and with respect to the detection of overrun probability, cross-entropy improves the detection accuracy and reduces the amount of computation; Meanwhile, the cross entropy uses discrete points measured directly for failure detection, to avoid calculation errors of fitting distribution curve compared with overrun probability; Finally, we analyze the impact of the number of samples using cross-entropy, indicating that the number of samples is not better in order to take the stability and accuracy into account.