Research On Static Load Prediction Method Of Bolt Structure Based On Machine Learning

Bolt support technology can not only improve the overall strength and stability of the supported structure,but also greatly reduce the space volume and the weight of the structure.Therefore,it has been used in building foundation pits,mine tunnels,and highway slope and other engineering applications.However,with the increase in service life of anchor bolt structures,coupled with the influence of harsh natural conditions and human factors,the damage problems of anchor rods and anchoring have gradually emerged.Due to the high load and large volume of the bolt supporting structure,a small amount of bolt failure will trigger the butterfly effect,leading to the large-scale destruction of the supported structure,resulting in huge loss of personnel and property.Therefore,it is very important to monitor the health of the bolt structure.The static load,anchoring quality,structural integrity and corrosion of the bolt structure are the key issues in bolt health monitoring.Among them,the static load monitoring of the bolt is the top priority,and the quantitative monitoring of the bolt load can ensure the safe operation of the bolt support system and help predict the service life of the bolt.In view of the structural characteristics and mechanical characteristics of the bolt,the smart washer based on lead zirconate titanate(PZT)is adopted in this paper and placed in the position of the ordinary washer.The self-actuating and self-receiving circuit is used to isolate the excitation signal and the response signal,and finally the static load detection based on single PZT is realized.Aiming at the shortage of static load prediction methods for bolt structures,this study combined with machine learning algorithm with strong selflearning and adaptive ability to carry out quantitative evaluation method research on static load.In this paper,multiple linear regression(MLR)and BP neural network(BPNN)are used as the basic models for static load prediction.At the same time,in order to fully consider the possibility that the piezoelectric response and static load have both linear and nonlinear relationships within a larger range of static loads,a composite network of MLR and BPNN based on the average method is proposed.In order to verify the effectiveness of the algorithm proposed,this paper designs a simulation experiment for the static load detection of the bolt,and the experiment obtains the piezoelectric response data under different static load conditions.The piezoelectric response signal sequence after deaveraging is used as the network input,and the normalized static load is used as the expected output.Three types of networks are used for training and testing.The test results show that the combined model has the best fit degree,with an average relative error of only 0.544,which can meet the needs of quantitative evaluation of static loads.In summary,aiming at the shortcomings of the anchor rod static load quantitative evaluation method,this paper proposes and implements a load prediction method based on machine learning.The test results verify the feasibility and accuracy of the method,indicating that the combined network of the network can establish the mapping model of the piezoelectric response signal and the static load.