Neural Networks In Active Structural Control In The Theory And Application

In this thesis, the neural network identification (NNI) and the neural network control (NNC) are investigated in detail in view of the particularities of neural network and architecture structure.Because structures of architecture are very huge, the action force is difficult to be brought to bear. The structure parameters are complex, time depended, random and unpredictable, so that it is very difficult to take the traditional control into the actual system. When the couple effects of the actuator and the structure are taken into the system, the system become more complex. The more actuators are taken into consideration, the more complex the system is.In Chapter 2, the base theories and the merits of the neural network are discussed and demonstrated Firstly. The feasibility that the neural network control is taken into the factual system is analyzed. The optimization methods of the neural network are discussed, and the shortcoming and the improvement of the methods are summarized. To illuminate the velocity of various methods, an example is considered.The NNI is a modal-free method, and on the data of the input/output. In Chapter 3, the selection of the NNI model, input signal and error rule are discussed. Then the NNI of the nonlinear system is studied. The static and the dynamic problem are clarified separately, and two examples are presented correspondingly.There are many forms of the NNC in practice, and each NNC is used in different field. In order to get the appropriate NNC form used in the active control of architecture structure, the merit and the drawback of various NNC are compared. Then, the parallel control between CMAC and PID is elaborated, so does the neural network adaptive control, and simulations show that the methods are very effective. Finally, in order to solve the problem of getting the sampleof input/output, a neural networks training algorithm is proposed that is based on instantaneous optimal control method.In the end, the couple system of the structure and the actuator is controlled. The full process from the identification to the control is given. And an ideal form of the actuator position is selected from many through the simulation.