| The development of theory analysis of rock mechanics and numerical simulation is restricted to some degree by constitution model and parameter of rock. Back-analysis is a good method to get constitution model and parameter of rock. System identification of rock is based on model class that is assumed by transcendental information of rock. Some research of the elastoplastic model identification among one model type is made. In view of the disadvantages of Evolutionary Algorithms in astringency, some betterment is done making use of the mechanism of thickness control of Immunity Algorithms. The main works is listed as follow:1 , Expatiating the concept of system identification of rock., summarizing the elastoplastic model, and introducing the method of optimizing back analysis using displacement.2 , In order to improve the diversity of individuals in colony, selection operation of Evolutionary Programming is mend by using the mechanism of thickness control of Immunity Algorithms. In addition, initial colony is created by the method of even test design.3, Combining Immune Evolutionary Programming with BP Neural Networks, the model of evolving Neural Networks and optimizing back-analysis is brought forward.It includes three parts:1) Neural Networks structure optimization;2)Neural Networks test;3)model and parameter back analysis.4, Applying the method mentioned above to practical engineering to back-analyse the elastoplastic model and parameters of rock. Validating and utilizing the results identified.
|
PDF Full Text Request