Research and Application of Process Object Intelligent Learning Modeling

Abstract

Industrial process objects are often multi-order inertial systems. Least squares identification algorithm needs prior knowledge to determine the order of the object model and then identify it. The identified model is prone to large deviations. In this paper, an intelligent learning modeling method for process objects is proposed. Two DNN deep learning networks are used to train and model the input and output data of the model, and an intelligent modeling system is designed based on the third-order inertial discrete structure. Firstly, the pseudo-random sequence excitation signal is added to the object open-loop system, and the first-order inertia plus delay object is used to fit the characteristics of the object, and the first-order inertia time constant T is obtained by genetic optimization algorithm. Secondly, a third-order inertia link and DNN deep learning network are embedded in the discrete structure of the third-order inertia model to build an intelligent model structure. To ensure that the third-order inertia link and the object inertia time are similar, the inertia time constant of each link is set to T/3, and the input and output data are sent into the intelligent model for training, and the DNN1 model of the object is obtained. Finally, the DNN1 model is embedded in the intelligent model structure again, and a more accurate identification model is obtained by training the input and output data for the second time. The validity of the intelligent model is verified by testing the simulation data. This method has great practical significance for the design and application of intelligent learning modeling.