Data-based adaptive fault prediction method and its application

Multi-level recursive method is an adaptive and data-driven fault prediction process. In terms of input-output equivalence, a nonlinear model can be modified into a multi-level linearized model using the multi-level recursive method. The time-varying characteristics of model parameters are accounted for at the same time. Therefore, the proposed approach obtained satisfied results when utilized in prediction issues. The fault prediction for CSTR(Continuous Stirred Tank Reactor) system has been studied based on the integrated multi-level recursive forecasting method which considering the CSTR system´s dynamic & time-variable characteristics. The optimal match of models and the algorithm of multi-level recursive method have been investigated through simulation. Through used in digital simulation experiments, the proposed method which is specific for CSTR system fault prediction has been validated and proved to be effective. This method can be used to predict the faults in such a class of nonlinear time-varying systems. Hence applying the proposed method in engineering and industry is proved to be feasible.