Unsupervised Hybrid Models Integrating Deep Autoencoders and Process Controllers' Models for Enhanced Process Monitoring and Fault Detection

This paper introduces a novel hybrid process monitoring model that integrates long short-term memory autoencoders with process controllers' models. The parameters of the hybrid model are optimized by minimizing a novel loss function, which combines the mean square error (MSE) between controlled variables and their reconstructions from the LSTM-AE model, along with the MSE of manipulated variables and their reconstructions obtained with the numerically implemented and exactly a priori known controller equations. The effectiveness of the proposed method is evaluated on the benchmark of an industrial-scale penicillin process as a batch case study and the Tennessee Eastman plant process under a decentralized control strategy as a continuous case study. A comparative analysis of the proposed hybrid model with an equivalent nonhybrid LSTM-AE model, which does not utilize process controllers' equations, highlights the superiority of the proposed hybrid monitoring model in fault detection. These improvements result from the use of an LSTM-AE network with fewer parameters, thus making it less susceptible to overfitting.