Structural Damage Diagnosis Based On GA And Resource Limited Artificial Immune System

The safety of large building structures is closely related to the social economic development and the people's living standards, Some large buildings have been damaged due to the long service time and the damage of natural disasters, such as typhoon, earthquake and other natural disasters. The real-time online health monitoring of large scale structure and the assessment of the safety and damage of the structure to found security risks of large building structures timely is an important subject of research. Structural health monitoring system gather and analys data of building structure based on wireless sensor element,As well as when the building damage, determine the extent of damage and take corrective measures to repair, reinforcement,control and eliminate safety hazards.Structure damage diagnosis is the key problem of structural health monitoring.The biological immune system has good information processing ability, which is very suitable for structural damage diagnosis.The artificial immune system is introduced into the structural damage detection and classification with the study of artificial immune system, bionic mechanism and algorithm model.Useing the distributed adaptive, learning, memory and associative retrieval of artificial immune system, The problem of structural damage diagnosis and classification of supervised and unsupervised the structure health monitoring system based on wireless sensor network is studied:(1) For supervised structures damage diagnosis problem, an artificial immune algorithm of GA evolutionism is proposed. The algorithm will samples data of structure model as antigen which stimulates the antibody sets, the antibodies go through evolutionary process which include celselect, crossove, rmutation and construction of the optimal antibody set in order to improve the quality of memory cell. The trained memory cell set is used to realize the damage detection and classification of measured data. In order to overcome the shortcomings of the binary encoding and useless of excellent model from other individuals,algorithm introduced multi parents crossover, which expand the search scope of algorithm and reduce inbreeding coefficient rate of simple artificial immune algorithm. The experiment results of the proposed algorithm using Benchmark structure model show that the algorithm achieve effective damage mode recognition.(2) In the process of structural fault monitoring, it is difficult to obtain all the damage mode, which limits the application range of supervised's structural fault detection algorithm.For unsupervised structures damage classification problem, an unsupervised structure damage classification method based on the resource limited artificial immune system is proposed. The algorithm make unsupervised samples data as antibody to composite ARB network, the ARB gothrough select, mutation, distribute B cells and resource limition in order to get high quality ARB net,which is used to realize the damage classification of measured data. A new calculation method of NAT is proposed to achieve effective connection and silhouette index is introduced to determine the network stability and clustering effect. The algorithm is tested using a Benchmark structure. The test results show the feasibility of using this algorithm for the unsupervised structure damage classification.