The Research On Multivariable Internal Model Control Algorithms

With the advantages of simple to design, excellent robustness, and good prediction ability, the internal model control(IMC) shows the strong vigor to solve the control problems of the multivariable systems. This paper mainly discussed and compared some different kinds of multivariable IMC strategies. The main study contents and conclusions are as follows:For the unstable time delay processes, a 2-degree-of-freedom IMC structure with two PID controllers was adopted, and the design methods of controllers for first order and second order unstable time delay processes were discussed. The simulation results indicate the validity of this method.For the square systems, three kinds of IMC design methods, namely, main loop IMC, process decoupling IMC, controller decoupling IMC, were analyzed and compared in detail. For the second kind, the essence and conditions of decoupling were analyzed, and a new expression of diagonal element for the generalized process was put forward, which contained the effective information of the original process. For the third kind, a new design method was proposed for two-input two-output processes. The static gain of the process transfer function's determinant and the inversion of the irrational part which was divided from the ratio of non-diagonal element to the corresponding diagonal element on the same row were used to design the internal model controller. It could solve the multiple time delays problem effectively. The NLJ optimizing algorithm with an improved ISE criterion was used to tune the filter's parameters, and it was realized by combining the M file and SIMULINK model. The multivariable IMC-PID formulas were deduced by Maclaurin series. Simulation results show that the third IMC method has the best decoupling effect and robustness.For the non-square systems, two kinds of IMC design methods, namely, main manipulated variable method and generalized inverse method, were discussed. The former method was to choose one main manipulated variable of the MISO system, and compensate other inputs. The latter method was to use the generalized inverse of the non-square process to design IMC controller. The simulation results show that both of the two methods were better than the squared-down method.