A dynamic failure rate prediction method for chemical process system under insufficient historical data

Chemical process system is especially prone to accidents due to equipment failures. Considering that in some circumstance the historical fault data is insufficient to perform statistical analysis, this paper proposes a method to predict the dynamic failure rate of chemical process system based on BP neural network and two parameter-Weibull distribution. The BP neural network is applied to expand the limited amount of process data and determine the fault states. Combining the expanding data set, the parameters of Weibull distribution and the failure rate function are determined. A liquid chlorine storage system is introduced to demonstrate the method. Results show that the failure rate of the system calculated by the method is more consistent with the actual situation, especially in the early stage of system operation. Moreover, this method can achieve a continuous dynamic prediction for future failure time points, which has practical meaning for the prevision of accident risk.