Cyclic Subspace Regression Modeling And Genetic Algorithm Optimization Based On Multi-Agent

It is important for modern industries to increase productive efficiency, cut down production cost, and enhance market competitive power by means of optimization. In particular, in process industry, optimization should be the first choice when the production flow scheme and installations had been built.There is actual significance to explore the relation between operation conditions and production input-output, and find out the rules from the fluctuation of the production process, then build the mathematical model and optimize the operation conditions according to the model. The complicacy of the relation between the dependent and independent variable in chemical process challenges the current modeling and optimization methods used in chemical process. The researchers all over the world are devoting themselves to develop new methods to solve the questions of the complicated chemical process modeling and optimization. This work researched and discussed the modeling and optimization methods and their applications in chemical process. The main contributions of this work are as follow:1) Discussed the Partial Least Square Regression (PLSR) and Multi-Cyclic Subspace Regression (MCSR), investigated Radial Basis Function Neural Network (RBFNN), analysed Genetic Algorithm (GA) and its weakness. Proposed a new integrated method and provided experiments to show the results of analysis.2) Proposed the RBF-MSCR method which integrates the RBF network and MCSR. It provides a novel modeling method for the complicated non-linear object of chemical process.3) Considering the defect of the Standard Genetic Algorithms (SGA) occurred when it deals with the large scale and complicated problem,the intelligent Agent was explored in this work, and the genetic algorithm based on multi-Agent System was worked out. A new MultiAgent-GA (MA-GA) method was proposed, which carries out genetic algorithms by the aid of a Multi-Agent System. This method enhances effectively the capability of GA to deal with the complicated optimization problem.4) Modeling the xylene isomerization process using the RBF-MCSR method, and optimizing the operation conditions using the MA-GA method based on the model built was elaborately discussed. The result was satisfactory. It showed these methods are good for modeling and optimization in applying to the complicated chemical process.