Investigation of damage identification of 16Mn steel based on artificial neural networks and data fusion techniques in tensile test

Abstract

This paper proposes a damage identification method based on back propagation neural network (BPNN) and dempster-shafer (D-S) evidence theory to analyze the acoustic emission (AE) data of 16Mn steel in tensile test. Firstly, the AE feature parameters of each sensor in 16Mn steel tensile test are extracted. Secondly, BPNNs matching sensor number are trained and tested by the selected features of the AE data, and the initial damage decision is made by each BPNN. Lastly, the outputs of each BPNN are combined by D-S evidence theory to obtain the finally damage identification of 16Mn steel in tensile test. The experimental results show that the damage identification method based on BPNN and D-S evidence theory can improve damage identification accuracy in comparison with BPNN alone and decrease the effect of the environment noise. © 2009 Springer.