Evaluation of bridge conditions using artificial neural networks

The working condition of a bridge is evaluated periodically for the purpose of safety. In the conventional assessment procedure, there is significant influence of subjective factors involved. A feasibility study on the use of neural networks in the bridge condition assessment is presented in this paper. A neural network that consists of five subnets is designed to simulate the current bridge evaluation process. For all of the training cases the network converges very well, and for the test cases the network prediction is consistent with the expert's in about 60 percent. From these case studies, it is observed that neural networks are capable of simulating the bridge condition evaluation process and modeling the input-output relationship. The presented method has a potential in real world applications.