Time-dependent Reliability Analysis of Bending Resistance of Aqueduct Side Wall by Combining BP Neural Network and High-order Moment Method.

Based on BP neural network and higher-order moment method, combined with measured data, a time-varying reliability analysis method for the bending resistance of reinforced concrete aqueduct side walls is proposed. In order to evaluate the bending time-varying reliability of aqueduct side walls under the influence of steel corrosion, this paper considers the effective cross-sectional area loss effect of steel bars in concrete structures, and derives the bending time-varying performance function under the limit state of bearing capacity of the aqueduct side wall. Then, combined with the measured data of steel corrosion and the principle of BP neural network, the prediction model of steel corrosion rate is designed, and the calculation formula of effective cross-sectional area of steel bars in concrete structures is established through the hemispherical pitting model. On this basis, the point estimation and high-order moment reliability theory is introduced to develop the time-varying reliability analysis method of aqueduct structure, and finally the proposed method is applied to the bending time-varying reliability analysis of the side wall of an actual aqueduct. The results show that compared with the previous eight practical empirical models, the prediction model of BP neural network built in this paper can predict the corrosion rate of rebar in concrete structures more accurately, comprehensively, easily and quickly. By comparing with Monte Carlo simulation method, the time-varying reliability index of this method is efficient and accurate, which can provide an effective way for the evaluation and prediction of time-varying reliability of aqueducts. [ABSTRACT FROM AUTHOR]