Safety Evalation For The Concrete Filled Steel Tube Arch Bridge Based On RBF Neural Network

Long-span bridge are the hings of traffic system and play an important role in national economic and society. The special mechanics behavior and the superiority in construction make the CFST Arch Bridges have a widely use in long-span bridges. But, with the growth of the span, the safety is becoming more and more important. However, there are no valid methods to evaluate the safety of the CFST Arch Bridge until now. The related researches focus on the extreme load limitation of steel tube. How to evaluate the safety of the CFST Arch Bridges under the permissibility of economic and technique condition is an important and urgent task.In recent years, the approach of neural network finds more successful application, especially the BP neural networks. However, BP neural network has one's own inherent defect, In view of this, people have proposed the neural network of the radial basis (Radial Basis Function Neural Network is abbreviated as RBFNN), The characters of RBFNN make it exactly demonstrate the vitality stronger than BP neural network, become a kind of new-type network which substitute BP neural network in more and more fields. This text has proposed being in the evaluation approaches of the safety of CFST Arch Bridge based on RBF network, applied this approach to the project instance and made certain achievement.Haveing done the following groundwork concretely:1. Based on the deep stuy of security evaluation appraisal of existing bridge, this text has explained the intension on security appraisal of the bridge. In discussing the foundations of the method and neural network technology that the security thinks in the domestic and international bridge at present thoroughly, as to the deficiency in the security appraisal of the bridge to the method of the neural network at present, propose the security evaluation of CFST based on RBF neural network ,which provids an intelligent new thinking of development that offers on security appraisal of the bridge.2. Systemly explain RBF neural network theory, it is appear RBF neural network have an advantage over BP neural network to lie in relatively to go on with BP neural network that RBF neural network has the only answer ,never fall to the district minmus,and trained more fast than BP neural network. Apart from this, owing to RBF neural network adopts one gauss of nuclear function, divided by make the network train and accelerate, so RBF neural network can deal with test instance to study data better also, and takes ability strengthen further to suffused with.It is very effective to what has been dealt with complicated structure such as bridge.3. Give a complete evaluation model to evaluate the safety of the CFST Arch Bridges based on AHP (the Analytic Hierarchy process). First give a hierarchy combination from the load bearing, construct damage and bridge condition and make a three layers safety evaluation model of the CFST Arch Bridges. Further more, give it's complete index system according to the practice of the CFST Arch Bridges.According to different influential factors, select different evaluation indexes rationally. The structural adjusting coefficient is selected as an evaluation indexes based on the load experiments of the bridge according to the load bearing. The crack-splitting degree and steel-rusting degree are selected as evaluation indexes according to the structure damage. As to the bridge condition, we chose the variable of languages to describe, then carry on number value to change according to appraising the classification standard, as the evaluation index amount of flawed or damaged state of the appearance. Getting the training samples based on the inspection of the CFST Arch Bridge, to training the Neural Network, and the result shows that the evaluation result is unanimous to the result given by bridge engineering experts. This method is proved that it can reappear the empirical knowledge and instinctive thinking of bridge engineering experts, reduce the contrived influence and assure the objectivity of the evaluating result better.4. Aim at real...