Research On Intelligent Adaptive Force/Motion Control Methods For Condenser Cleaning Mobile Manipulator Robot

Condenser is the vital apparatus of the thermal power plants, nuclear power plants and other industrial plants. Commonly, the industrial-condenser-system is constructed as a bank of horizontal or vertical small-condenser-tubes that play the role of heat exchangers. In fact, the condensers usually operate under impact of environmental-conditions for very long periods of time, and the hazards to which they are prone during normal service always exist. When the ills appear on the condensers, the operating-effectiveness of the condensers will be reduced. In addition, the risks from the condensers will affect not only unit heat rate and other equipment of plants but also the operating-cost. To overcome this problem, a condenser cleaning mobile manipulator robot for large condenser-cleaning has been researched and tested in recent years. By inheriting this advantage, this dissertation focuses on intelligent control methods-based hybrid neural networks (NNs) in order to improve the operating-effectiveness for this robotic control system. Main results and contributions of this dissertation are presented as follows.The first contribution proposes an adaptive intelligent tracking control method based on fuzzy wavelet neural networks (FWNNs) and recurrent FWNNs (RFWNNs) for robot manipulators control. Here, there are two intelligent controllers, adaptive RFWNNs and FWNNs controllers, are proposed to examine and compare. In this method, the control system is designed without the knowledge of robotic system. The unknown dynamics of the robotic system is approximated by the proposed RFWNNs/FWNNs. By combining the wavelet technique with the fuzzy neural networks (FNNs), the proposed controllers-based the FWNNs can achieve higher tracking-performances in comparison with the FNNs/NNs controllers. In addition, the recurrent technique is also considered in the first contribution, the second proposed RFWNNs-controller, to improve the flexibility/adaptation for the first proposed FWNNs-controller under parameter variation conditions. The RFWNNs structure is a dynamic network structure. Thus, it can overcome the static-structure problem of the FWNNs structure, and it can be applied more convenient in highly nonlinear systems in the presence of time-vary ing uncertainties. The effectiveness and robustness of the proposed methods are confirmed by comparative simulation and experimental results. The robot manipulators-control is first examined here to set the premise properly for the mobile manipulator robot control (MMR).In the second contribution, an adaptive intelligent force/motion control system based on the RFWNNs is applied for the condenser cleaning mobile manipulator robot (CCMMR) control. The purpose of this contribution is to improve the flexibility and tracking errors of the previous controllers-based neural networks for the CCMMR or MMR control under time-varying uncertainties. By inheriting the good results from the first contribution, the RFWNNs/FWNNs are also utilized to relax the unknown-dynamics-requirements of the CCMMR control system. When compared with the robot manipulators structure, the dynamics of CCMMR are more complex with impact of the holonomic/nonholonomic constraints. Therefore, an adaptive robust control strategy in the proposed control system is also developed for the nonholomic constraint force from the CCMMR. In addition, the online-learning algorithms of the proposed control-parameters are obtained by the Lyapunov stability theorem such that the stability of the proposed control systems is guaranteed. The effectiveness and robustness of the proposed methods are verified by comparative simulation and experimental results.The third contribution proposes a novel recurrent fuzzy wavelet cerebellar model articulation control neural networks (RFWCMACNNs) structure. This structure is the combination of the wavelet technique, the Takagi-Sugeno-Kang (TSK) fuzzy structure, WNNs, and the recurrent cerebellar model articulation control (RCMAC) neural networks structure. Therefore, the RFWCMACNNs are more generalized networks and they are suitable for the control of the CCMMR. The purpose of this contribution is to improve the effectiveness of controllers-based NNs for the robotics-control. In addition, this contribution also presents the improvement for the approximation processes-based cerebellar model articulation control (CMAC) techniques, such as fuzzy CMAC (FCMAC), RCMAC and recurrent FCMAC (RFCMAC). The RFWCMACNNs are applied in the proposed position tracking-controller to approximate the unknown dynamics of the CCMMR control system. Based on the design of the position-tracking controller, an adaptive robust control scheme is also developed for the nonholonomic constraint force from the CCMMR/MMR. All the online-learning algorithms of RFWCMACNNs control-parameters are derived by the Lyapunov stability theorem. Therefore, the stability of the proposed methods is guaranteed. Moreover, this contribution also presents the design brief of the controller-based fuzzy wavelet CMAC NNs (FWCMACNNs) to compare with the proposed method. The effectiveness and robustness of the proposed intelligent controllers are verified by comparative simulation and experimental result that are implemented in the CCMMR control system.In the fourth contribution, an intelligent control system is proposed for the CCMMR by inheriting the advantage of the conventional backstepping control system (BCS). Based on the good control-results from the second and third contributions, the RFWCMACNNs are applied in the position-tracking-BCS to approximate the unknown dynamics of the CCMMR/MMR control system. The purpose of this contribution is to further improve the effectiveness and robustness of the previous controllers-based NNs and CMAC for the CCMMR control. Similar to two previous contributions, the control of the nonholonomic constraint force from CCMMR is also considered by an adaptive robust controller. All the online-learning algorithms of control-parameters in the proposed controllers are obtained by the Lyapunov stability theorem such that the stability of the controlled system is guaranteed. In addition, comparative simulation and experimental results are provided to confirm the effectiveness and robustness of the proposed control systems.