Inferring rules for adverse load combinations to crack in concrete dam from monitoring data using adaptive neuro-fuzzy inference system

The formation and growth of cracks in concrete dams are mainly induced by hydrostatic and temperature loads. As cracks especially unstable cracks are of great danger to the safety of dams, it is critical to avoid extremely adverse load combinations during the dam operations to achieve the stability of cracks. Conventionally, the adverse load combinations have to be deter- mined empirically by experts based on specific dam site conditions. Therefore, it is attractive to apply quantitative instead of empirical methods to identify the adverse loading conditions. In this study, we employ an adaptive neuro-fuzzy inference sys- tem （ANFIS） to Chencun concrete dam. The ANFIS is able to help us build a relationship between the model inputs （reservoir water level and air temperature） and the model output （crack opening displacement）. Based on this relationship, the rules of the adverse load combinations to the crack are generated directly from the monitoring data. The accuracy of the trained ANFIS is proved by comparing the modeling results and the monitoring data. Our work demonstrates that the ANFIS is a useful approach for accurately recognizing the rules of the adverse load combinations that can be used in the knowledge base of dam safety expert system.