Research Of Remote Intelligent Identification Methods For Structural Damage Based On Neural Network

Neural network is one of the most important branches of AI. The research about intelligent diagnostic technology based on artificial neural network has the strong background of engineering and the great value of the practical application. Following the impetuous development of Internet, how to setup a remote intelligent diagnostic system is becoming a focus area of research. This thesis discusses the present state of structure damage identification research and its tendency. A remote intelligent diagnostic system and the computing methods are developed by Java.In this thesis, the main studies include: (1) Discussing the advantages of several network models, and finding one which is suitable for large scale science computing; (2) The methods and process of establishing the identification system model are demonstrated. And the structural damage identification methods based on BP network for unilateral damage plates and multiple beams are demonstrated; (3) Based on the B/S model which sets client as computing center, a remote intelligent diagnostic system is developed by Java. The system consists of graphic user's interface module, sample processing module, neural network structural module and neural network computing module. The process is like as following: the client downloads the JavaApplet program, then the client can set up the neural network by himself, inputs the sample data and preprocesses these data. After the training, the program will show client the results and save it.The studies mentioned above show that the system of structural damage identification based on BP has acquired relatively high accuracy and rapid convergence. Meanwhile, the remote intelligentdiagnose model based on client as scientific computing center can solve the problem in some extents. And it is valuable and worthy of further research.