Identification For Parameters Of Mr Damper Model Based On Improved Genetic Algorithm

To the MR damper used in the civil structures, its choice of characteristic parameters should affect in great degrees to the effect in absorbing vibration. Therefore, the following work gradually becomes hot in civil engineering area: how to build the model for the given MR damper, and how to identify the unknown parameters of the model.The research work aims at the optimization algorithm for the identification of damper parameters directly against to the parameters influencing the property of MR damper. There are 3 parts in the thesis.The first part is to build the model for the damper. From the various models of MR damper, the Bouc-Wen model is chosen for research. This model can comprehensively simulate the properties of the damper in reducing vibration at every using stage. It's proved that the model can be according to the experiment in very small error. Therefore Bouc-Wen model can be used as the research target. Its parameters are the ones for identification.The second part is the research of algorithms. The genetic algorithm and the simulated annealing algorithm are studied in theory. The main points in design are determined. For the genetic algorithm, the individual is treated as a binary chromosome containing 10 genes. From the contrative curve between the probability of genetic operation and the optimal result, the value of crossover and mutation probability are determined. For the simulated annealing algorithm, such main points for design as the function of determining initial temperature, the function of annealing and the function of generating new solutions are also determined. Based on these algorithms, the identification results can be achieved. Through the contrast and analysis between the calculation and experiment, the respective advantage and shortcoming for these two algorithms are achieved. For example, the genetic algorithm contains the robustness and parallelism, and it can assure the diversity of population. But in the other hand, it's easy to get in the local optimization, and makes the optimization inefficient. The simulated annealing algorithm can accept the bad solution in stated probability, which is helpful to jump out of the local optimization. But the convergence speed of SA algorithm is too slow.