Identification And Control For Processes With Time-Delay

The problem of identification and control for processes with time-delay is always encountered in industry. Because of the time-delay, the controlled processes can not reflect the disturbance of systems in time, and which brings obvious overshoot. In addition, the adjusting time becomes longer and the difficulty of control becomes larger. This paper proposed two methods to solve this problem.Based on modified internal model control (MIMC) scheme, a control strategy which introduce to the inner-loop control named feedforward-feedback control is proposed in the first part of this paper. The separately design method can be used to design the feedforward-feedback controller for disturbance rejection and the MIMC controller for set-point tracking. Furthermore, a method of choosing the inner-loop controller is applied when normalization of time delay is different. Moreover, by tuning the filter time constant, the robustness is improved when model mismatch between process and plant exits. The results of simulations show good disturbance rejection and robustness of the controller. The principle of SVM is introduced in the second part of the article, which has performed excellent ability of non-linear function approximation. Thus, the Gauss RBF kernel-based function is applied to identify the positive-model and inverse-model of the processes with time-delay. Combining the IMC with the identification of inverse-model, the internal model controller of SVM (SVM-IMC) is proposed. Firstly, according to the SVM inverse identification, the inverse-model can be acquired. Then using the principle of internal model control, the identification model can be seen as the internal model controller. Moreover, by tuning the filter time constant, the stability of processes can be maintained. Simulation results proved the good performance of SVM which based on positive-model and inverse-model identification, and SVM-IMC in many aspects, including identification precision; identification speed; less dependence on sample, and stronger generalization ability.