Multi-Resolution Based Fuzzy System And Its Applications In Locomotive Adhesion Control

Fuzzy system identification and control is an important branch of the identification and control of general nonlinear system that is a complex research field. With the increase of the complexity of nonlinear system and control performance, it becomes very hard to acquire the complete and interoperating fuzzy rules only by knowledge of the experts or manipulators. Therefore, it is important to go deep into the study on fuzzy system identification for the sake of the improvement of fuzzy system theory. Adaptive fuzzy system has been proposed on the basis of this requirement. But more studies are focused on parameter identification of fuzzy system. Therefore, in the existing approaches, fuzzy system models are mainly viewed as black-boxes, carrying about the conflict between interpretation of fuzzy rules and control accuracy. In another respect, with the development of railroad transportation in our country, especially the development of over-loading and high-speed transportation, it is urgent to adopt optimal adhesion control for the fully utilization of maximum traction force. In order to solve the above-mentioned problems, this thesis first proposes a novel topology of fuzzy system based on wavelet transform and multi-resolution analysis theories, i.e. multi-resolution analysis based fuzzy system (MAPS). Then, system identification and adaptive control algorithms for MAPS are studied. Eventually, MAPS adaptive adhesion controls are proposed and investigated. The results of this thesis not only enhance the fuzzy system theory but also propose a novel approach for the structure identification of fuzzy systems. Moreover, this thesis also gets some primary results for the final realization of optimal adhesion control. The main contents in this thesis are as following:(1) Based on wavelet and multi-resolution analysis theories, the properties of B-spline scale function and its wavelets are analyzed. Combined the compact support and multi-resolution approximationproperty of B-spline scale function with traditional fuzzy systems, a novel multi-resolution analysis based fuzzy system (MAPS) is firstly proposed in this thesis from the viewpoint of time-frequency localization.(2) The characteristics of MAPS related to multi-resolution are investigated. The approximation capability of MAPS is analyzed by employing function analysis and function approximation methods. MAPS is proved to be a universal approximator on the basis of theory analysis.(3) Several structure and parameter identification algorithms are proposed by combining wavelet network identification with multi-resolution analysis characteristic of MAPS. Among them, system identification algorithms of MAPS are mainly studied. Simulation results show the validity of these algorithms and the superiority of MAPS.(4) Based on adaptive fuzzy control theory, multi-models control theory and system identification algorithms of MAPS, MAPS adaptive control of a class of nonlinear systems are studied, and some algorithms are proposed.(5) The dynamics of locomotive traction are analyzed. Optimal adhesion control algorithms are studied based on MAPS adaptive controls, including of estimation of optimal slip speed, single-model and multi-model based MAPS adaptive adhesion control. Simulation results show that the proposed control algorithms are valid and feasible.