Model-Free Adaptive Control With Noise Attenuation Based On Modified Wavelet Threshold Denoising Methods

With the increasing complexity of industrial process and the higher requirements of control precision, model based modern control theory cannot meet the requirements of industrial production any longer. Meanwhile with the development of computer technology, large amounts of data from production process were stored in industrial computers, these data contains significant information about the system. On this basis, data-driven control algorithms emerged as the time require. Model free adaptive control (MFAC) is a typical data-driven control method, the design of MFAC controller only using input and output (I/O) data of the controlled system, none of the mathematical model information in the controlled process is included in the controller. After more than twenty years of development, many achievements of MFAC have been obtained in theory and in practice. For practical systems, the noise disturbance of the measured data is an inevitable problem. And I/O data is exactly the design basis of MFAC controller, therefore it has great significance in both theory and application to find a method which can attenuate the noise disturbance. The main work of this paper is as follows:First, the application of MFAC in various fields is reviewed in this paper. The literatures on applications of MFAC were summarized and classified into 15 major categories according to the application domain. Moreover, the application situations of each field were briefly introduced.Secondly, for the problems caused by measurement disturbance in practice, a MFAC based on wavelet threshold denoising method was proposed using modularized designing. In order to realize adaptive real-time denoising for the output data, adaptive decomposition level and sliding time window are applied. The convergence of the proposed scheme was theoretically proved, and its validity was verified by simulations.Finally, the proposed control strategy was improved from two perspectives.On one hand, a modified control strategy based on sliding time window algorithm with variable length was proposed to restrain the overshoot and maintain the stability of output. On the other hand, a modified control strategy based on the median filtration was proposed to reduce the negative influence of white gaussian noise and impulse noise. Futhermore, the effectiveness of the.control strategies was verified by simulations.