The Study Of Inner Model Control Based On Neural Networks For Interim Reactor

Interim reactor is a kind of common equipments in the chemistry industry, such as fine chemistry and biology chemistry industry. The operation of interim reactor has strong qualities of nonlinearity and time lag. And it is hard to reach ideal control requires using general method to control it. So the method of inter model control based on neural networks has been proposed in this paper, which combined with the theory of neural networks and genetic algorithm.Inner model control is a new control algorithm which is designed based on the math model of the process. Because that it is easy to be designed, has good performance in control and has superiority in system analysis, inner model control is not only a practical advance control algorithm, but also an useful tool to improve the design level of normal control system.Neural network has the satisfactory capability of approximating any nonlinear mapping, furthermore, it can learn and adapt to the dynamical properties of uncertain system. So neural network based control system has fairly strong adaptability and robustness. An inner model control algorithm optimized by using neural network is proposed, in this paper. This algorithm builds a predictive model of the process by using the computational ability and the satisfactory capability of approximating any nonlinear mapping of the neural network. It can simplify the realization of inner model controller by using predictive model to take the place of the math model.Genetic algorithm is a kind of optimization searching algorithm which is based on the natural selecting and the theory of genetics. It just wants to simulate the evolution process of creature and the operation of genes on the computer when genetic algorithm is used to optimize object, otherwise, it does not need to know any character about the object and does not need the searching space of the object is continuous and differentiable. Moreover, genetic algorithm has the abilities to search all round the space and to get the result fast. Parallel genetic algorithm has been used to train the neural networks in this paper. The algorithm can reduce the operation time of evolution, improve the converging speed and precision of the neural networks training, which by maintaining the multiformity of individuals and reducing the sizes of the populations. This can improve the precision and mapping ability of neural networks, which will improve the performance of inner model control algorithm.The inner model control algorithm is used to the temperature and pressure control of ABS resin polymerization reactor. The simulation result indicates that the algorithm can reach a good result of controlling the pressure and temperature of the ABS resin polymerization reactor.