Neural Network Forecasting of the Load-Bearing Capacity of Reinforced Concrete Elements

Abstract

The article is devoted to the study of the force resistance of reinforced concrete bendable elements at various stages of their life cycle, taking into account corrosion damage to the reinforcement and subsequent composite reinforcement. The paper considers changes in the characteristics of reinforced concrete structures under the influence of operational and aggressive environmental factors, which leads to a decrease in their bearing capacity. The main attention is paid to the corrosion of reinforcing steel, which causes cracking of the concrete coating and a decrease in the cross section of the reinforcement, reducing the durability of the structure. An approach based on the use of artificial neural networks is proposed to predict the residual strength and bearing capacity of reinforced concrete elements. The prediction was carried out using artificial neural networks in the Statistica software package, models of a multilayer perceptron were developed and tested using data on tests of reinforced concrete beams with various damages and reinforcements. The models of artificial neural networks were validated using cross-validation and showed high resistance to changes in input data, which confirms their suitability for practical use in civil engineering to assess residual load-bearing capacity. The use of artificial neural networks makes it possible to take into account non-linear dependencies between parameters and improves the predicted results, which makes them a more reliable tool compared to traditional methods.