Fault Detection and Isolation in Industrial Processes Based on RNN: Deep Learning Approach

Abstract

Deep Learning is a subset of Machine Learning related where human brain-inspired neural network algorithms learn from a large amount of data across multiple layers for nonlinear transformation.,Deep learning domain techniques allows to extract features automatically from the Industrial fault detection dataset. Since it has the strong learning ability that can improve the utilization of the dataset for the feature extraction. It determines Industrial Processes fault in the input data by simulating the way the human brain detects the fault in the industrial processes.The Existing system automatically detects the faults and isolation based on Deep Learning. The existing system uses the process variable to prevent and detect the failures in the industrial process. Due to the process complexity and sophistication of the Industrial equipments, the existing system is not affordable and unrealistic to implement on large scale. The proposed system is based on the Parameter Optimization Techniques such as genetic algorithm (GA) and particle-swarm optimisation (PSO). These techniques are used to increase the trained model capability. So that the trained model can handle the complex correlations between the input values. To improve the prediction accuracy, the proposed system uses the recurrent neural-networks(RNN). Therefore, the proposed application predicts the future state of the industrial processes based on the previous behaviour while considering the substantial noise in the data.