A Study Of Batch Reaction Process Modeling Based On Combined B-Spline Neural Network

In recent years, with the rapid development of Chinese market economy, the demand of chemical process market promoted the use of chemical process modeling. Its application areas have been expanding and become more and more important. Because of the adaptive ability, variety specifications and high quality of market demand, batch process has been more and more important and has been widely used in many areas such as special chemicals, electronics materials, polymer materials, food biochemical products and medication. However, it has features of high nonlinear, uncertainty, disturbances, limited process variable and online measurement information. According to these difficulties, this paper studied on the key problem of batch process modeling.This paper's main contents are:According to nonlinear dynamic system, such kind of complex modeling problem, in the past, the modeling method mainly by using the recursion form. The solution for the sequence of optimal control variables usually need to use differential or integral and is very complex. This paper mainly studies on using the method of combined B-spline neural network to model nonlinear dynamic model and ites optimal control and implementation. Considering the batch reaction process is time fixed, use the time variable as B-spline neural network's input, let its share the function of describing object's dynamic characteristic. The relationship between input and output variables are operating by another B-spline neural network. Two neural networks combined to establish the nonlinear batch reactor dynamic model.The simulation experiment results show that combination B-spline neural network(CBSNN) by using least-square method of learning can built a nonlinear dynamic system model with higher precision and quicker speed. Through comparison with other neural network modeling methods shows that CBSNN can reduce the complexity of the single network. CBSNN improves using single neural network modeling of the existence of the insufficient precision and long training time, etc. The experimental results of suspension polymerization of styrene reaction conversion optimal control proved that the optimal control strategy is effective. Because of the combined neural network modeling ideas, optimal control strategy can be decomposed. The batch reaction time is pre-defined so that one neural network can be regarded as part of the output value known. The decomposition method for solving the optimal control sequence brings great convenience.By using CBSNN modeling for the time variable characteristics, this paper built a model for a typical nonlinear time-varying system. Through comparison advantages and disadvantages with different modeling methods shows that the feasibility of the CBSNN model for time variable related objects.This paper studies the batch reaction and the nonlinear and time-varying system modeling and optimal control problems. It's innovation and extension. It has actual value in chemical industry modeling and optimization control, and shows great advantages.