Nonlinear Chemical Process Monitoring And Fault Detection Based On Modified Lstm Model

• Main Author: Zambri, Muhammad Ridzuan
• Format: Monograph
• Language: English
• Published: Universiti Sains Malaysia 2022
• Subjects: T Technology; TP155-156 Chemical engineering
• Online Access: http://eprints.usm.my/55494/1/Nonlinear%20Chemical%20Process%20Monitoring%20And%20Fault%20Detection%20Based%20On%20Modified%20Lstm%20Model.pdf; http://eprints.usm.my/55494/

With the development of the chemical industry, fault detection of chemical process has become hard challenge due to the high-dimensional data and complex chemical process and increasing number of equipment. The standard feedforward neural network is not particularly effective at solving these issues. This study proposed a fault detection model based on modified Long Short-Term Memory (LSTM) model. The simulation experiment of the Tennessee Eastman (TE) chemical process for modified LSTM model will be using MATLAB software. The investigation the performance between the LSTM model with the Artificial Neural Network (ANN). The modification of the LSTM will be made by comparing different type of faults that will be used for the fault detection. The percentage of the training and validation also has a great influence towards the accuracy of the fault detection. The link to determining the optimum number of hidden layer nodes by manipulated the value of each hidden layers on the LSTM network is added since the number of hidden layer nodes in the LSTM network impacts the diagnosis outcome. Then, the optimized LSTM model will be obtained in order to get higher accuracy of the fault detection in chemical process. Finally, through the simulation in the MATLAB software, the results show that the modified LSTM model has a better performance in chemical fault detection than ANN and the higher accuracy that can be achieved by the LSTM model is 99.69%.