Tunnel Lining Diseases GPR Detection Intelligent Inversion And Identification Methods Based On Deep Learning

With the increasing number of tunnels in China's water conservancy,transportation,municipal and other fields,the importance of tunnel lining structure for healthy service and long-life operation is becoming increasingly prominent.During the operation period of the tunnel,cracks,voids,cavitys,water seepage and many other structural diseases often occur,which seriously threaten the tunnel health service and long-life operation.It is of great scientific significance and engineering value to study the detection and diagnosis methods of internal structural diseases of tunnel lining,to ensure the safety of tunnel operation,and to realize the predictive maintenance of tunnel lining diseases.Ground-penetrating radar is a non-destructive testing technology that uses the reflected waves of high-frequency electromagnetic pulses to detect the distribution and characteristics of underground objects.It has the advantages of fast and efficient,intuitive results.However,due to the complexity of the tunnel lining structure,the diversity of diseases and the interference of radar signals in the tunnel environment,the ground penetrating radar data identification of lining diseases has problems such as poor precision,strong dependence on experience and low automation.In view of above problems,this paper aims to realize high-precision inversion of tunnel lining dielectric model and automatic identification of disease type,position and contour,relying on the strong nonlinear mapping ability of deep neural network,by combining theoretical analysis,numerical simulation and model.Experiments,field tests and other means to study radar intelligent inversion based on deep learning and tunnel lining disease diagnosis,and proposed radar data intelligent inversion method based on deep neural network,which realizes the high precision of tunnel lining structure disease dielectric model.Inversion;based on convolutional neural network,the tunnel structure anomaly recognition and disease automatic classification are realized,and the results of intelligent inversion are supported.Based on the intelligent inversion results,the SegNet deep learning model is used to realize the automatic extraction of disease shape location contours..Finally,an intelligent inversion method and automatic identification classification method suitable for tunnel lining disease detection data were formed.The model test and field test were carried out to verify the results,and good results were obtained.The main research work and results of this paper are as follows:(1)For the inversion problem of radar detection data to relative permittivity model,this paper proposes an intelligent inversion method for radar detection data based on deep learning network,which is used to approximate the nonlinear relationship between radar detection data and the relative permittivity model through deep neural network.To realize the mapping of detection data to dielectric constant model,optimize the deep learning model by designing network architecture and parameters,improve the indicators of intelligent inversion,and then determine the deep neural network model for intelligent inversion of radar data..(2)For the high-precision inversion problem of the dielectric model of tunnel lining structure,based on the the intelligent inversion of radar data,the tunnel lining disease model conforming to geological significance is designed.By the finite difference time domainmethod(FDTD),wo studied the radar response characteristics of typical lining structure diseases,and generated corresponding large amount of radar detection data,and then trained the deep neural network to realize the intelligent inversion of tunnel lining disease detection data,and the intelligent inversion results were better than the traditional full waveform.Inversion.(3)According to the automatic identification problem of internal disease types and distribution profiles of tunnel lining,the automatic identification and classification method of tunnel diseases based on radar observation data is studied.By constructing convolutional neural network,the internal identification of tunnel lining and various diseases are realized.,supported by the intelligent inversion results.Aiming at the problem of tunnel lining disease type and contour identification,the semantic segmentation deep network model based on SegNet is studied.The intelligent inversion prediction model is classified by pixel level to realize the automatic extraction of disease morphology location.(4)Based on the above research,the intelligent inversion program based on deep neural network is designed.The intelligent inversion based on numerical simulation is carried out,and the convolutional neural network is constructed to realize the automatic identification and classification of radar data.Furthermore,the validity and reliability of the proposed method are verified by designing model tests and collecting field test data.