Application of BP neural network and genetic algorithm in stress prediction of anchor bolt

Abstract

The bearing capacity detection of anchor bolt system is very important for the supporting effect evaluation. In this paper, back propagation neural network(BPNN) and genetic algorithm(GA) were used to predict the pull force of free bolt. Acoustic stress wave signals of free bolt were collected under different pull forces and analyzed in time domain and frequency domain. The wave velocity, fundamental and secondary frequency of acoustic stress wave signals were selected as inputs of BPNN. The weights and thresholds of BPNN were optimized by GA to avoid local solution. 8 sets of data were used to test the stress prediction effect of BPNN after training. The results indicates that the BPNN optimized by GA can achieve small errors when compared to basic BPNN.