Cognitive Tuning Method Of PID Parameters Based On Case-based Reasoning

PID controller is widely used in industrial process control with the simplestructure, robustness. And the parameter (proportional, integral, and differential)values directly affect the tuning control system performance, which will cause a majorimpact on the smooth operation of the production process and the final product quality.So far, theoretical tuning method, or engineering tuning method has been well studied,however the PID parameters achieved by the methods above are difficult to get usedin actual control, which still needs the experience of the operator to adjust theparameters, therefore, PID parameter tuning method has been widespread concernedin the control community. This paper analyzes the characteristics of existing tuningmethods, aiming at the difficulty in the model identification, the long regulation timeand the difficulty in the determination of the critical point etc, a systematic study ofPID parameter tuning methods and experiments are carried out, the main contents areas follows:(1) Through the analysis of human cognitive mechanism, a case-based reasoningcognitive PID parameters technology tuning method (CTCBR) is proposed.According to the cognitive reasoning model, a CBR tuning structure is designed,which could achieve the self-tuning of PID parameters by conducting the featureextraction, multi-attribute decision case retrieval, multi-objective case assessment,case reuse, case correction modules. The proposed cognitive tuning method couldavoid the difficulties on the identification of model parameters and determination ofthe critical point.(2) According to the introspective learning principles, a method withself-learning ability for the construction of the tuning case baseis proposed. In thismethod, the discovery and extraction of PID control systems operating results fromthe PID tuning values result is put into operation to fix the setting value, and thesuccessful and efficient experience cases will be stored in tuning the case base, whilefailing and inefficient experience will be forgotten, so as to achieve optimized storageof the tuning case base and self-learning function.(3) Learn from multi-attribute decision making thoughs, an improved multi-attribute decision making case retrieval methods is designed. According to preferencetuning program, he method uses entropy weight method to establish preferencerelations of each performance indicator and conducts the comprehensive evaluationthrough the gathering way of the performance-based indicators, then retrieves asatisfactory tuning result for the real implementation, which could improve the overallretrieval success rate of a given problem. (4) Aiming at the real-time adjustment problem for PID parameters, a modifiedmulti-objective evaluation of case-based approach is studied. This method adopts aprior evaluation strategy on the setting value to correct PID parameters based onperformance trends, which realizes the cognitive function of the setting valueadjustment.(5) According to the proposed cognitive tuning method, a cognitive PIDparameter tuning comprehensive simulation platform is developed withMATLAB-SIMULINK tool. By typical first-order, second-order system andcomparative experiments of TE process, the result shows that the method doesimprove various performance indicators, which proves the effectiveness and theapplication advantages of the proposed method. Meanwhile, the integrated simulationplatform lays the foundation for the continuing application research on the PIDparameter tuning problem with CBR.