Designing supervised local neural network classifiers based on EM clustering for fault diagnosis of Tennessee Eastman process

This paper proposes a supervised local multilayer perceptron (SLMLP) classifier integrated with independent component analysis (ICA) models for fault detection and diagnosis (FDD) of industrial systems. The interest of this paper is to improve the performance of single neural network (SNN) by dividing the fault pattern space into a few smaller sub-spaces using Expectation-Maximization (EM) clustering technique and triggering the right local classifier by designing a supervisor agent. To detect both known and new faults of the system, two ICA models are integrated with the proposed classifier. The performances of this method are evaluated on the data of Tennessee Eastman (TE) process, a benchmark chemical engineering problem. The results from the experiments show the superiority of the proposed method compared to other well-known published works.