Application of Improved BP Algorithm in Prediction of Axial Tension Performance and Crack Width Control Ability of UHPC

In view of the defects of traditional BP algorithm in neural network-based function fitting, such as low accuracy and slow convergence speed, this paper proposes an improved and optimized BP neural network algorithm (igsaa-bp) in the application of UHPC axial tension performance and crack width control ability prediction. The algorithm in high performance concrete (UHPC) axial tensile properties and fracture width control prediction is adopted in the control variable method, the four kinds of different fiber content of the shaft and characteristic parameters of high performance concrete and crack width control parameters for the comparison and analysis, based on the measured curve, puts forward the relationship between the stress - strain relationship and crack width - contains the basic tensile stress of three phase structure. The improved and optimized BP neural network algorithm can avoid falling into local optimization more effectively, and modify all individuals of the new population species according to the situation, so as to improve the global optimization ability of the algorithm and increase the diversity of individual population. The research results in this paper show that igsaa-bp algorithm not only improves the convergence speed of BP network, but also effectively improves the fitting ability of the network. When the steel fiber content increases from 2.3% to 3.5%, the axial tension performance strain increases by 230%, which indicates that the constitutive relation model of UHPC and the experimental data are in line with expectations.