Batch Process Reactor Soft Measurement And Iterative Learning Control Study

As with the flexibility, high value-added features of batch process, batch process has been widely used in modern industry. However, the level of automation of batch processes is relatively low compared to continuous process.On one hand, it is lacking affordable and reliable online sensors to detect a number of variable parameters such as concentration of reactants.Despite trying to develop such sensors, their usefulness and reliability are still very limited and expensive so that they cannot meet the actual needs.On the other hand, the batch process is a complex industrial process with a large hysteresis, nonlinear, time-varying parameters and difficult for modeling, which brings the batch process control system design much difficulties and challenges.This thesis, as for the two difficult issues on concentration measurement and temperature control in the batch reactor which is widely used in chemical process,has done some research on soft measurement technology and iterative learning control. Main work and innovation points as follows:(1) Data communication between Matlab based on OPC technology platform and configuration software WinCC is implemented, and is applied to a batch reactor data acquisition. To improve the quality of collected data, a series of data pre-processing techniques such as excluding outliers, data normalization processing, wavelet filtering and principal component analysis and so on are researched(2) A soft-sensor modeling algorithm based on the combination of an improved adaptive genetic algorithm and BP neural network is proposed. The algorithm fully retains the IAGA and BPNN respective strengths, and improves their shortcomings, so that not only to get the global optimal solution, but also to accelerate the convergence rate of the population. Simulation of the algorithm and its concentration software measurement in the batch reactor application are given. Simulation results show that the algorithm is high precision, faster.(3) The control strategy of adaptive switch and open and closed loop PD-type iterative learning advanced control algorithm is applicable to nonlinear systems which have the state and control delay. It takes full advantage of system information along the time axis and iteration axis.The adaptive control algorithm whose PD parameters change based on the number of iterations is introduced. Introducing the forgetting factor makes full use of the past and current system information in the open loop iterative learning control framework.Then the algorithm's convergence is proved based on the theory of operator.Finally, a comparative simulation study between this algorithm and advanced closed-loop PD-type open-type iterative learning law is made, simulation results show the algorithm put forward in this thesis has feasibility and advantages.(4) Take the producing process of Unsaturated Polyester Resin-a typical batch process for studying, and do reactor temperature control curve fitting study to establish the target temperature control curve.A simulation study which uses the adaptive switch and open and closed loop PD-type iterative learning advanced control algorithm above is made, and the results show the algorithm gives a better solution to the difficult problems of temperature control in batch process reactor and achieve better results.