Research On Structural Damage Detection And Classification Based On Artificial Immune Algorithm

The health status and safety evaluation of large-scale civil engineering infrastructure not only affects the lifeblood of the social economy,but also relates to the people's safety of life and property. The health monitoring of large building structures has been one of the hottest topics. Structural health monitoring system can get the real-time information about structure system using advanced sensing element. Combined with the signal information processing method and modeling method of structural mechanics,structural health monitoring system extracts structure damage characteristic parameters to diagnosis health status of the structure and control or eliminate the unsafe factors in order to avoid the developing of security risks. At present,there are few researches on structural damage detection and classification,which is one of the most important problems in structural health monitoring. This thesis utilizes artificial immune system method to solve structure damage classification problem. The main work of this thesis is as follows:(1)Biological immune system has the extremely powerful capabilities of identification and classification. Thus,basing on the basic concepts,system composition,system characteristics and principle analysis of the biological immune,using the autonomy,initiative,adaptive and the bionic principle between learning and memory.based on the low energy consumption and distributed detection wireless sensor network,this thesis mainly researches the problem between structural damage identification and classification of Structural Health Monitoring System,also the correspondence between biological immune system and structural health monitoring sensor network based on artificial immune system is further discussed in detail.(2)Based on the research of supervised large-scale structural health monitoring and classification,this thesis presents a clone selection algorithm of particle swarm mutation. The algorithm samples data of structure model as antigen which stimulates the antibody sets. In order to improve the quality of memory cells,the antibodies go through learning and evolving process including cloning,mutation and selection. Especially,the particle swarm mutation is introduced to solve the problem of binary encoding complexity and uncertain of mutation direction in clone selection algorithm. At last memory cell sets of high quality are used to detect and classify measured data. The experiment results of the proposed algorithm using Benchmark structure model show that the algorithm can improve the success rate of structure damage classification obviously.(3)For the large span environmental space,complex structure environment and long-term stable operation of SHM,a classification method based on the AiNet immune clustering algorithm for the large-scale structure is presented for the unsupervised structural damage classification. Firstly,the AiNet algorithm is employed to update memory cell sets and get the clustering center. Secondly,these memory cell sets are used to classify measured data. The algorithm is tested using a Benchmark structure putted forward by IASC-ASCE SHM task group. The simulation results discuss the effect of inhibition threshold on clustering and the average classification success rate,the test results show the feasibility of using this algorithm for the unsupervised structure damage classification.