Study On Nondestructive Detection Of Anchorage Quality Of Rock Bolts Based On Improved Elman Neural Network

The quality and working condition of anchor anchoring system determine the safety performance of the whole construction project to a great extent.In order to ensure the use of the anchor in the support system works to prevent the occurrence of major disasters,anchor quality inspection research has become particularly important.In actual engineering,the non-destructive testing technology is used to carry on the quality inspection research on anchor rods,In this thesis,the stress wave method and the neural network are combined to predict the quality of the anchor.Specific research content are as follows:Firstly,according to the practical engineering application,four different anchor states of bolt anchoring system models are established based on the finite element analysis software ANSYS/LS-DYNA,and the dynamic response of the stress wave in different anchoring states is analyzed,the anchorage length and the defect position and the length of the anchored bolt are determined,and the accuracy of the model is verified.It shows that the four different anchor anchorage system models can be used in the follow-up study.Secondly,the thesis introduces the method of modal analysis of anchor bolt by ANSYS software.The modal analysis of anchor anchoring system model with different length,different material parameters and force or not is carried out,and the natural frequency is analyzed and compared.The natural frequency is independent of the initial condition and only relates to the length of the anchor anchoring system,the geometric parameter setting,and the material parameter setting.Thirdly,an output-input Elman neural network based on the ant colony optimization(ACC-OIF-Elman neural network)is proposed because of the shortcomings of Elman neural network,such as slow learning speed and likeliness to fall into local minimum.Based on the original network structure,the feedback of the output layer to the hidden layer is increased,and the OIF-Elman neural network is obtained,which improves the ability of the network to process the information dynamically and adapts to the time-varying characteristics.It takes advantage of the ant colony clustering algorithm to improve the function of network intelligent search and the overall optimization on the basis of the OIF-Elman neural network,effectively improve the OIF-Elman neural network convergence speed,and improve the accuracy.Accordingly,the ACC-OIF-Elman network structure for bolt defect recognition is established.Finally,the Elman neural network,OIF-Elman neural network and ACC-OIF-Elman neural network are used in the intelligent prediction of the bolt: First,the Elman neural network,OIF-Elman neural network and ACC-OIF-Elman neural network are used to identify the defect of the simulated bolt and the actul engineering bolt.The input of the neural network is the characteristic value of the acceleration signal of the simulated anchor and the actul engineering anchor extracted from the wavelet packet energy spectrum;Second,Elman neural network,OIF-Elman neural network and ACC-OIF-Elman neural network are used to predict the length of the bolt.The results of modal analysis of multiple anchors are used as input of the neural network.It is verified that the ACC-OIF-Elman neural network has higher classification rate and prediction accuracy.