A Hybrid Strategy Enhanced Sub-Sampled Recursive Second Order Algorithm for Chemical Process Intelligent Identification

• Published in: International journal of computational intelligence systems Vol. 16; no. 1; pp. 1 - 20
• Main Authors: Wang, Yaxin, Xu, Baochang
• Format: Journal Article
• Language: English
• Published: Dordrecht Springer Netherlands 20.07.2023; Springer
• Subjects: Advanced process control; Artificial Intelligence; Chemical process modeling; Computational Intelligence; Control; Deep learning; Engineering; Intelligent identification; Mathematical Logic and Foundations; Mechatronics; Research Article; Robotics; Second order learning algorithm; Deep learning; Intelligent identification; Chemical process modeling; Second order learning algorithm; Advanced process control
• ISSN: 1875-6883; 1875-6883
• DOI: 10.1007/s44196-023-00296-5

The accurate dynamic model of the chemical process is an important condition for the successful implementation of advanced control in the plant. In this paper, an efficient second order algorithm for long short-term memory (LSTM) network training is proposed for chemical process intelligent identification. A novel Hessian inverse recursion method is adopted to achieve fast convergence and avoid the high-cost operation of the classic second order optimization method. Besides, more information is back propagated since the proposed method retains the real curvature information of the neural network. Considering the large amount of chemical process data, a sub-sampled recursive second order-stochastic gradient descent (SRSO-SGD) algorithm which uses sub-sampling method and hybrid strategy is proposed. The identification experiment on a delayed coker fractionator shows that the proposed sub-sampled neural network second order training algorithm has better performance than other learning algorithms in terms of model identification accuracy and convergence speed. By adopting a hybrid strategy that performing Hessian inverse estimation every 3 training epochs, the expensive Hessian inverse calculation cost in the identification process is further reduced while low training and testing errors are maintained.