Reliability Analysis of Multi-Stress Accelerated Life Testing Based on Genetic Neural Network
To deal with the difficulties of traditional life prediction methods in establishing an accelerated model and solving pluralism likelihood equations, a new model is proposed to predict the life of the items in multi-stress genetic neural network model based on their predictability and convergence. The accelerated stress levels and reliability are used as training input vectors. By the least square fitting, the regression equation of the original data can be obtained, with which a large number of simulation data can be generated as training target vectors. Then a three-layer BP neural network is set up and trained, and failure data can be predicted by putting the normal stress levels and required reliability into the model, and the predicting curves can be drawn. The stimulation data demonstrates that the model can reflect the relationships among the stress levels, the reliability, and the life of the items. It provides a new route for life-prediction in multistress accelerated life testing.
BP neural network Genetic algorithm constant stress reliability accelerated life test
Xiaoying Wang Jihong Shen
College of Automation, Harbin Engineering University, Harbin, China College of Automation, Harbin Engineering University, Harbin, China College of Science, Harbin Engin
国际会议
太原
英文
477-480
2011-02-26（万方平台首次上网日期，不代表论文的发表时间）