Study On Decision-making System Of Bridge Components Maintenance Cycle Based On Dynamic Fuzzy Neural Network

It is necessary to monitor structures’ state because finding the damage to structures’ components timely and taking effective measures are important for ensuring structures’ normal operation and safety.At present,the bridge health monitoring system in practical engineering has accumulated a large amount of information about bridge state.However,the results of manual inspection which will be affected by inspectors’ knowledge and experiences are the basis of bridge condition assessment nowadays.In the same time,the data collected by bridge health monitoring system has not been fully utilized.In addition,most of the bridge maintenance work is based on the existing codes and standards.Although the codes and standards stipulate the inspection cycle for each type of bridge’s components,the importance,environmental conditions and stress state of different components are different and these characteristics will change during the life cycle of the bridge,which leads to the deviation of the maintenance cycle decision.In this paper,a decision-making system for bridge components’ maintenance cycle is established by dynamic fuzzy neural network.And the incremental learning for dynamic index weight is realized by introducing sliding window to traditional learning algorithm of dynamic fuzzy neural network,which can provide a more accurate method for determining the maintenance cycle of bridge components.The main work of this article is listed below:1.Based on the analysis of the traditional dynamic fuzzy neural network learning algorithm,the sliding window is introduced to improve its learning process so that it can learn incrementally.The Hermite function is learned by dynamic fuzzy neural network with two learning methods respectively.The results show that the dynamic fuzzy neural network learning algorithm with sliding window has the same learning accuracy as the traditional learning method,which means the learning ability of dynamic fuzzy neural network will not be reduced by using this method.2.Based on the existing research,the dynamic fuzzy neural network with sliding window is employed to establish bridge components’ maintenance cycle decision system.The bridge state evaluation system based on fuzzy analytic hierarchy process(F-AHP)is used to simulate experts’ evaluation process under actual conditions.Then,the decision-making system proposed in this paper is trained by the samples before being tested.The results show that the indexes’ weight will be extracted after a certain amount of data are studied,which makes the system make decision like the exports.3.Taking Tsing Ma Bridge as an example,the decision-making system proposed in this paper is used to make decision on the maintenance cycle of each component.The correctness of the decision-making system is proved by comparing its results with those coming from the rating system based on fuzzy analytic hierarchy process.The decision-making system’s adaptability to dynamic weight index is verified by its learning and prediction for changing indexes’ weight.