The Research On Some Chemical And Physical Test Methods In The Chemical And Fiber Production Process

A well-developed transportation network can play an important role in promoting economic development,but as the national road traffic network continues to expand,road collapse accidents due to underground cavities are becoming more and more frequent.This poses a serious threat to the safety of people’s lives and property.In this context,Ground Penetrating Radar(GPR)has been widely used as an important tool for nondestructive testing of urban roads.However,due to the complexity of the structure of the detection area and electromagnetic wave propagation mechanism,there is a huge challenge in the data interpretation of ground Penetrating Radar.In view of this,this paper addresses the ground-penetrating radar detection data interpretation problem under urban roads from both forward and inverse ground-penetrating radar simulations.The specific research contents are as follows:(1)To address the problem of low computational efficiency of existing ground-penetrating radar orthorectified simulation methods,this paper investigates the ray theory-based ground-penetrating radar orthorectified simulation method,writes ground-penetrating radar signal simulation program based on MATLAB platform and simulates ground-penetrating radar simulation images.Then,for the shortcomings of Multistencils fast marching methods(MSFM)in computational efficiency and transmission ray computational accuracy,improvement strategies are proposed:using deep learning methods to improve the computational accuracy of transmission ray by 1 order of magnitude;using parabolic interpolation algorithm to improve the computational efficiency of a single reflection point,the computational The calculation time is reduced from 0.02 s to 3.95×10-6 s.The velocity model partitioning algorithm is studied to improve the calculation efficiency.After the simulation experiments,the profile generated by this method is basically consistent with the simulation image generated by gprMax,and achieves a greater advantage in the calculation efficiency.Finally,the ground-penetrating radar imaging law is studied according to the simulated image and the actual radar measurement data,and the influence of factors such as relative permittivity and conductivity on ground-penetrating radar imaging is analyzed.(2)By combining spatial pyramid pooling with a one-dimensional U-net network and introducing a cosine annealing learning rate adjustment strategy,we successfully inverse-perform the electromagnetic wave propagation velocity from the ground-penetrating radar data with zero offset distance,and solve the problem of matching the electromagnetic wave propagation velocity in the time domain with the ground-penetrating radar data.The inversion problem is solved.Compared with the physically driven inversion method,this method can complete the inversion efficiently and accurately.The ground-penetrating radar simulation dataset is constructed to train the model and compared with GPRNet.The experiments show that compared with GPRNet,GPRUnet improves the R~2 value by about 0.6%and the inversion results are more consistent with the actual velocity model,which verifies the accuracy and superiority of GPRUnet.Finally,the inversion is performed on the measured data,and this measured data is analyzed.