论文部分内容阅读
核级阀门具有高可靠性、长寿命的特点,阀门历史失效数据的小子样问题突出,并且由冲击、振动、磨损、腐蚀等耗损性因素引起的故障,其故障概率具有时变性,随时间的增加而增大。故障概率p为常数的Jeffreys先验模型不能以合理概率复现观察数据,故满足不了分析p的时变性要求。对服从Binomial分布的阀门故障数据建立广义线性模型,研究概率p的时间趋势;在评价模型复现观察数据能力时不仅进行定性图检验,而且还利用贝叶斯?2统计量进行定量化检验;经过定性和定量的双重检验,表明该模型具有良好的预计能力,可以分析阀门故障概率p的时变性。
Nuclear-grade valves have the characteristics of high reliability and long life. The small sample-like problems of valve historical failure data are prominent, and the failures caused by the wear and tear factors such as impact, vibration, abrasion and corrosion have a time-varying probability of failure. Increase and increase. Jeffreys priors, whose failure probability p is constant, can not reproduce the observed data with a reasonable probability, so they can not meet the time-varying requirements of analyzing p. The generalized linear model of valve fault data obeying Binomial distribution was established to study the time trend of probability p. When the model was used to evaluate the observational data, not only the qualitative graph test but also the Bayesian 2 statistic was used for quantitative test. After qualitative and quantitative double test, it shows that the model has good predictive ability and can analyze the time-varying of probability p of valve failure.