Fault diagnosis of the IFAC Benchmark Problem with a model-based recurrent neural network

Abstract

Demonstrates the utility of a model-based recurrent neural network (MBRNN) in fault diagnosis. The MBRNN can be formatted according to a state-space model. Therefore, it can use model-based fault detection and isolation (FDI) solutions as a starting point, and improve them via training by adapting them to plant nonlinearities. In the paper, the application of MBRNN to the IFAC Benchmark Problem is explored and its performance is compared with 'black box' neural network solutions. The benchmark problem represents the nonlinear model of an electromechanical governor used in speed control of large diesel engines. For this problem, the MBRNN is formulated according to the eigenstructure assignment (ESA) residual generator. The results indicate that the MBRNN provides better results than 'black box' neural networks, and that with training it improves the results from the ESA residual generator.