Research On Non-destructive Testing Of Bolts Anchorage System Quality Based On Structural Self-organizing Elman Neural Network

The bolts anchorage system is widely used in various geotechnical engineering projects because of its convenient construction,low cost and reliable supporting effect.The supporting effect of the bolts anchoring system directly affects the quality of the project.If the anchorage system has defects,the bolt support capacity will be reduced,which will affect the quality of the project and even cause people’s life and property safety accidents.Therefore,it is of great significance to identify and classify the defects of the anchoring system.The neural network can be used as an intelligent classifier for the classification of anchor system defects.It can complete the network training according to the existing bolts anchorage system defect database,so as to correctly classify the bolts anchorage system defects.The neural network overcomes the shortcomings of the traditional manual classification method with poor accuracy and slow classification speed,and also requires the classification personnel to have higher professional technical level and experience.However,the network structure of the neural network has an important influence on the performance of the classification.The methods determined by the traditional network structure,such as the trial and error method and the empirical formula method,are difficult to obtain the optimal structure of the network and are laborious.Therefore,the difficult problem of neural network structure is difficult to determine,which hinders the large-scale popularization and application of neural network as a smart classifier for anchor identification of bolts anchorage system.In this paper,a self-organizing Elman neural network is constructed for the difficult problem of neural network structure.The neural network is proposed as an intelligent classifier for the experimental data and simulation data obtained by the stress wave anchor non-destructive testing method.Identification,thereby completing the classification of the defect intelligent identification of the bolts.The specific research is as follows:(1)According to the defects of insufficient anchoring density and anchoring cavity in the application of anchor engineering,four kinds of experimental defect identification classification bolt models are constructed.Based on the stress wave nondestructive testing method,the intelligent identification data of bolts are obtained by experimental method and simulation method respectively.(2)According to the characteristics of Elman network structure,the network pruning mechanism is designed based on the definition of network hidden layer node contribution and network pruning threshold.The convergence of network pruning process is analyzed.The adaptive improvement of network learning rate is completed.UCI standard is utilized.The classification data set analyzes the feasibility of the network structure pruning process.The simulation results show that the pruning mechanism can reduce the network redundancy and improve the network classification performance.(3)Based on the contribution degree of the network hidden layer,the Elman neural network growth threshold is defined,and the network split growth mechanism is introduced.Together with the network pruning mechanism,the structure self-organized Elman neural network is constructed.The standard deviation of the contribution distribution of the hidden layer nodes is used.The pruning threshold and the growth threshold are corrected.at the same time,the pruning coefficient and the growth coefficient are adaptively improved to improve the stability of the structural structure of the self-organizing process.The convergence of Elman neural network structure self-organizing process network is analyzed.The UCI standard data set is used to test the network performance.The simulation experiment proves that when the initial structure of Elman neural network is too complicated,the network can reduce the network redundancy by network structure pruning.When the initial structure of Elman neural network is too simple,the network structure can be increased by the growth mechanism to improve the network data processing capability.The network can complete the structural self-organization according to the training data,obtain the optimal structure of the network,and improve the network classification performance.(4)The self-organizing Elman neural network So-Elman neural network is used to identify the bolts defect data intelligently.By analyzing and comparing the network parameters and performance,the experimental results verify that the self-organizing Elman neural network has structural self-organization ability and better classification performance,which can complete the classification and identification of bolts defects,and then complete the non-destructive detection of bolts anchorage quality.