Gray Neural Network And Application In Lifetime Predication Of Concrete Structure

The paper focuses on gray uncertainty problem. By using the combinative method of gray system and neural network technology, intensive research on gray system and neural network feature has been done. And then the two technologies have been combined organically according to the similarity in information presentation of the two technologies.The paper outlines the research of gray system method and neural network technology and builds the model of gray nerve network system. According to various gray certainty information or the factor number of relative chief behavior, it studies emphatically on six types of neural network model, including GNNM(1,1), GNNM(2,1), GNNM(1,2), GNNM(1,3), GNNM(1,4), DNNM( 1,4).The six models separately define the information processing course and algorithm of one-dimension or many-dimensions gray problems. The first two models are built on the one-dimension gray problems; the later four models are built on two-dimensions, three-dimensions and four-dimensions gray problems. The object (gray uncertainty complex problem), which is applicable to them, is different because of the diversity of their chief behavior.This paper studies the relationship between operational life of concrete construction and environmental corrosion mediator, such as sulfate, etc. Concrete samples of 37 project instances and service condition under different geographic environment in the world are collected and these data is used in our research.A BP neural network model is designed to analyze and solve the problem, the relationship of operational life of concrete construction and environmental corrosion mediator. We can separate damage grade into seven hierarchies based on the diversity of the structure damage degree. According to the different anticipated damage degree, the network output makes an approximate estimate of the length of structure use. The network has four layers. Input layer has 16 nodes. The first hidden layer has 17 nodes; the second hidden layer has 10 nodes and one output node.37 projects' data is used in training samples. By using MATLAB, accurate BP network tool is applied to do the network training. We can appraise the utility of other project items using this network, and the result is valuable.In the end, Microsoft Visual C++ 6.0 associated with MATLAB 6.5 is used to develop BP network concrete structure with the length appraisal diagnose procedure, and this project has been used in Ten-Five Key Project of concrete expert system.