On Modified Recursive Empirical Frequency-domain Optimal Parameter Estimate Method And Its Application

For the purpose of improving accuracy and speed of model parameters identification for control systems, this paper focuses on the design of parameters estimation method for disturbed linear systems. Based on reduced-rank matrix approximation technique, a novel recursive empirical frequency-domain optimal parameter(EFOP) estimate method is proposed to solve the contradiction between accuracy and calculation of existing EFOP methods. The application in servo system sufficiently shows that it is practical and effective to apply the proposed method to estimate model parameters of actual control systems. The main results can be summarized as follows.Firstly, the basic principle of the empirical frequency-domain optimal parameter(EFOP) estimate method and its recursive algorithm is expounded. In order to enhance the accuracy of the existing recursive EFOP(REFOP) estimate algorithm in the case of weak input signal, a novel REFOP estimate algorithm is proposed by modifying the recursive form of the vital time-varying weighted matrix. The extensive simulations show that compared with the existing REFOP algorithm, the proposed one can perform more precise parameters estimation whether the input signal is weak or not, which verifies its correctness and effectiveness.Secondly, to improve the computational efficiency of the proposed algorithm, the technique of reduced-rank matrix approximation is employed to alleviate the increased computation burden caused by the modification, and a fast modified REFOP parameter estimate method based on reduced-rank matrix approximation is developed further. The extensive simulations show that it can perform faster parameters estimation while guaranteeing high estimation accuracy, which verifies its effectiveness and superiority.Thirdly, those factors that impact the performance of the modified REFOP estimate algorithm is analyzed. Through deducing the analytical expressions of error matrix, the native factors affecting the estimation accuracy and speed of the proposed algorithm are derived. Furthermore, based on simulations, the influences of input signals characteristics on performance of the proposed algorithm are deeply analyzed from intuitive and practical point of view, which provides a helpful guidance of selecting the input signal for reasonable application.At last, the proposed algorithm is applied to estimate model parameters of some turntable, and combining with the model-based frequency-domain parameterized control method, a model-parameters-estimation-based(MPEB) controller is designed. The experiments are conducted to compare the control performance of MPEB controller with the one of fixed-parameter controller. The results show that the MPEB controller always satisfy the double-ten index performance whether the turntable is installed with massive load or not, which is superior to the fixed-parameter controller. The experiments verifies well the feasibility that the proposed algorithm is used to estimate model parameters of actual control systems.