摘要
提出一种基于人工神经网络的可靠度预测方法,以系统故障时间和对应的中位秩估计值训练网络,在系统故障时间范围内选取大量的故障时间点并求其预测的累积分布函数值,然后结合样条回归法求出系统累积分布函数曲线,概率密度函数曲线及故障率函数曲线。为验证人工神经网络模型的优越性,以婴儿培养箱等5个系统的故障数据为例,用决定系数R2、均方误差和对数似然函数,与Weibull、Fréchet、Logistic等统计分布模型进行对比,结果表明人工神经网络拟合效果最优。
A system reliability prediction method based on artificial neural network is proposed.The network is trained with the system failure data and the corresponding estimate of median rank estimate.A large number of failure time points are selected in the system failure duration,and their predicted cumulative distribution function values are found.Subsequently,the cumulative distribution function curve,probability density function curve,and failure rate function curve of the system are derived using spline regression method.To verify the superiority of the artificial neural network model,taking the failure data of 5 systems such as infant incubator as an example,the proposed model is compared with the statistical distribution models such as Weibull,Fréchet and Logistic using the coefficient of determination R~2,mean square error and log-likelihood function.The results show that the artificial neural network fits the best.
作者
樊立天
江金达
夏景涛
崔飞易
缪吉昌
王婷婷
夏红林
王胜军
陈宏文
FAN Litian;JIANG Jinda;XIA Jingtao;CUI Feiyi;MIAO Jichang;WANG Tingting;XIA Honglin;WANG Shengjun;CHEN Hongwen(Department of Medical Engineering,Nanfang Hospital,Southern Medical University,Guangzhou 510515,China)
出处
《中国医学物理学杂志》
CSCD
2023年第2期232-237,243,共7页
Chinese Journal of Medical Physics
基金
国家重点研发计划(2019YFC0121908)
广东省科技计划项目(2017ZC0068)
广东省自然科学基金(2017ZC0069)。
关键词
人工神经网络
统计分布
可靠度预测
拟合精度
artificial neural network
statistical distribution
reliability prediction
fitting accuracy