PID Neural Network Control System of Ball Mill Based on Modified PSO Algorithm

Abstract

Ball mill system is a complex multivariable system, which has characteristics of strong coupling, nonlinearity, large delay and slow time-varying, so it is difficult to build its precise mathematical model and achieve satisfying control effect with conventional control strategy. In view of the problem, the paper proposed a multivariable PID neural network control strategy which is independent of the mathematical model of controlled object based on analysis of dynamic characteristics of the ball mill system. In order to improve the performance of controller further, a modified PSO algorithm was used to off-line optimize training of initial value of weights of PID neural network, and the weights were adjusted by BP algorithm on-line, so as to avoid that network falls into the local minimum,and ensure the system cannot overshoot and shake greatly. The simulation results showed that the strategy can guarantee robustness and adaptability of the control system of ball mill in a large range, and solve problems of coupling and time-varying, which has good coupling mechanism and control quality.

Keywords

• ball mill, pulverizing system, pid neural network, multivariable system, decoupling, modified particle swarm optimization algorithm