Failure Rate Prediction Method Based on BP Neural Network and Two-Parameter Weibull Distribution-A Case Study of Liquid Chlorine Storage System
The liquid chlorine storage system is especially prone to leakage accidents due to component defect and equipment failure, which may result in explosion, poisoning and other severe consequences.Combined with limited historical fault data of liquid chlorine storage system from a chemical enterprise,this paper presents a method to predict failure rate in the future based on the BP (back propagation) neural network and the two-parameter Weibull distribution.Firstly, the BP neural network is applied to expand the data set through training and testing for known historical data.There are two methods for predicting the corresponding fault condition of data set, namely using BP neural network again and contrasting anti-normalization prediction parameters with pre-setting threshold of fault status, and this paper chooses the former since their errors are 16.67% and 33.33% respectively.Then the results of BP neural network are combined with some mathematical methods (such as the middle-rank-method, the Nielsen method) to determine the parameters of two-parameter Weibull distribution and the failure rate function, based on which the mean failure rate can be calculated for each phase.Finally, compared with the traditional way of failure rate calculation and the method merely considering original process fault data into two parameters Weibull distribution, this paper shows the proposed failure rate prediction method for calculating the mean failure rate at various stages is closer to the actual failure rate on the basis of the expansion of total data in the early stage and can achieve a continuous dynamic prediction for future, which has more practical meaning for the safety production of enterprise and can help to avoid unnecessary risk.
failure rate prediction BP neural network two-parameter Weibull distribution liquid chlorine storage system
LI Chenyang WANG Jinghong ZHI Youran WANG Zhirong JIANG Juncheng GONG Junhui
Jiangsu Key Laboratory of Hazardous Chemicals Safety and Control, College of Safety Science and Engi College of Mechanical Engineering, Nanjing Institute of Technology, Nanjing 210009, Jiangsu, China
国际会议
The 2016 International Symposium on Safety Science and Technology(2016安全科学与技术国际会议)
昆明
英文
25-35
2016-10-17（万方平台首次上网日期，不代表论文的发表时间）