Research And Implementation Of The Truss Damage Identification Based On Intelligent Algorithm

With buildings,roads,highways,and other projects in the frequent occurrence of accidents,the quality of the project has aroused increasing attention.But in projects,the existence of initial imperfection,the load and the combined effect of the environment,the cumulative long-term decline in performance often tend to be ignored,it is potential safety problems and may lead to heavy losses or even accidents in some particular situations.As load-bearing structural elements,truss is very important in the whole project,and it is meaningful to identification the damage of the truss.The current mainstream technologies for the truss damage identification are based on vibration or artificial intelligence.Since the parallel distributed processing,fault-tolerance and a high degree of robustness and self-learning adaptive capabilities,artificial neutral network become an effective method for damage identification of truss.Feedforward neutral network are researched on the most mature and used most widely.As the basic BP algorithm with slow convergence,there are some improve technology for it.The LMBP algorithm,using numerical optimization technology,has the fastest convergence rate. Because of its gradient-based search strategy,BP algorithm is vulnerable to local optimum.Particle Swarm Optimization as a Swarm Intelligence algorithm,has strong global search capability,can be used for training neural network to overcome the defect of BP algorithm.In particle swarm optimization algorithm,the precocious also occurred in some time,constraining the performance of the algorithm.In artificial immune system,suppression mechanism of antibody and genetic variation mechanism can be to maintain the diversity of antibodies effectively,vaccination can make use of priori knowledge to guiding the search process,speed up the convergence of the algorithm.In this paper,introducing immune operator to particle swarm optimization algorithm,construct an immune particle swarm optimization algorithm(ImPSO). For the four benchmark functions,testing the immune particle swarm optimization algorithm,and comparing with the standard particle swarm optimization algorithm, the results show that the immune operator overcomes the defects of precocious effectively and enhances global optimization capability,improved performance of algorithm significantly.The paper also uses such immune particle swarm optimization algorithm to train the feedforward neutral network,the results are similar to the results of function test,immune operator improves the performance of algorithm.The paper compares the ImPSO with some variations of BP algorithm,for the neutral network which has trained with ImPSO,using LMBP algorithm to further training,combining the two algorithms can make use of the global search capabilities of ImPSO in the early,and also make use of the fine local search capabilities of LMBP in the late,performance of the network is improved greatly. Using the network which trained with the two algorithms to identify the damage of truss,the results can guide to find the damage location.On this basis,the paper realized the simulation system of damage location identify for truss.