Fault Diagnosis Based on Unsupervised Neural Network in Tennessee Eastman Process.

In the industrial process, in order to solve the problem that the supervised neural network used for fault diagnosis always needs to compare with the corresponding output value constantly and takes a long time. In this paper, the competitive neural network and the improved self-organizing feature mapping neural network in unsupervised neural network is proposed for fault diagnosis. In the learning process, the active neighborhood between neurons can be gradually reduced without obtaining output values, so as to enhance the activation degree of central neurons, and then the weights and thresholds can be automatically adjusted. In this way, maintenance personnel can get more time for timely maintenance and reduce losses to a great extent. Finally, the feasibility of the proposed method is verified by the simulation of Tennessee Eastman process.