Application of artificial neural network in bridge deck condition rating

Abstract

Currently bridges are evaluated either a visual inspection process or structural analysis. When bridge evaluation is conducted by visual inspection a subjective rating is assigned to the bridge components with analytical evaluation. The rating is computed based on the load applied and the resistance capacity of bridge component visual inspection is subjective and depends primarily on the experience of the inspector in assigning the rating. analytical rating unable to represent the condition of bridge component since the rating is computed based on the load and bridge capacity only. If a relationship between analytical rating and subjective rating can be found. The estimation of bridges condition can be made only by determining the bridge analytical rating. Several attempts to correlate both methods using the conventional statistical analyses. As well as fuzzy logic have not been very succesful in providing the relationship between those rating methods. However, an attempt to utilize Artificial Neural Network (ANN) to correlate the analytical rating for railroad and bridge parameter with bridge subjective rating in Chicago has produced succesful results. This study describes the application of ANN in developing the correlation between load rating and subjective rating as well as bridge parameter for highway bridges. The subjective rating in this study is limited to deck rating only. The data provided by California Department of Transportation (CALTRAN) is utilized for training and testing session. The results obtained in the first part showed that additional variables are needed along with load rating variable to provide acceptable prediction performance. After several rounds of improvement process the best reults obtained exhibits 77% of the data used for testing are predicted within the acceptable range. Generally, this study showed that the ANN has a potential to be used to predict the subjective rating if the proper input variables are applied.