Fault Prognosis for Data Incomplete Systems: A Dynamic Bayesian Network Approach

For the cases that data samples are partially missing in control systems, analysis are given to determine the type of missing data mechanisms, then a dynamic Bayesian network approach is used to model the general fault prognosis problem in control systems, after that we proposed the method of dynamic Bayesian network to deal with real-time fault prognosis of nonlinear systems with missing data. Our approach is demonstrated on a benchmark continuous stirred tank reactor (CSTR) problem, with which we show the process of constructing the dynamic Bayesian network model and use the model for the simulation of fault prognosis. Results show that though data samples are noisy and partially missing, combined with effective treatment of missing data, dynamic Bayesian networks can efficiently predict the system failures.