Research On Downhole Channel Modeling Of Mixed Model Based On Ray Tracing Method

As one of the main energy sources in China,coal mine plays an irreplaceable role in China's economic construction and development.However,the environment in mine roadway is complex,the electromagnetic wave transmission is severely limited,and the development of mine communication system is slow.In this paper,a series of researches are carried out on the electromagnetic wave transmission characteristics of underground coal mine.It provides a theoretical basis for the research and development of underground communication in coal mine and has practical significance.In this paper,the tunnel in a complex environment is simplified into a straight rectangular tunnel,the inherent transmission characteristics of electromagnetic wave in the tunnel are studied according to the waveguide theory,and two types of electromagnetic wave transmission field are obtained:^?36???mhnand?36???mvn</sup>mode.According to the formula derived transmission cut-off frequency of transmission mode,electromagnetic wave propagation constant and attenuation formula,further analysis of the electromagnetic wave energy attenuation.Then,in view of the special environment of the underground roadway,several factors affecting the electromagnetic wave transmission are studied and summarized,and expressed by formulas.Specific factors include:loss caused by rough roadway wall,loss caused by slope of roadway wall,loss caused by dust scattering in the roadway and fog drop loss caused by damp air.There is a serious multipath effect on the transmission of electromagnetic waves in underground tunnels.This paper introduces the forming factors of multipath effect and introduces the classical electromagnetic wave simulation algorithm-ray tracking method.This paper introduces the forming factors of multipath effect and introduces the classical electromagnetic wave simulation algorithm-ray tracking method.Ray-tracking algorithms treat electromagnetic waves as light,the propagation path of electromagnetic wave is simulated by optical principle,in the ray-tracking algorithm,the field intensity of electromagnetic wave in each path is calculated by vector superposition at the receiving point,and finally the field strength value simulation prediction graph was obtained,the least square algorithm was used for curve fitting to obtain a log curve of downhole attenuation,and the performance of the algorithm was verified.In this project,the actual data were collected in the underground pipe corridor which is similar to the mine tunnel environment,and it was found that the trend of the ray tracking simulation is consistent with the measured data curve,but there are certain errors.Modeling method in the paper considering this modeling method requires high precision of scene data,large amount of data and complex algorithm.The problem of channel modeling is considered as a nonlinear mapping problem between signal strength and receiving distance in the roadway and relevant parameters in the environment,and a BP neural network with strong data processing ability is introduced,which is combined with ray tracking to construct a hybrid model.On the basis of the ray-tracing method,the difference value between the simulation data and the measured data is obtained,the difference value is taken as the input of BP neural network,the field intensity value is taken as the output of the neural network,training network,the BP neural network can master detail information in the coal mine environment's influence on the field intensity value.The construction of the hybrid model makes the subtle information's influence on the field value in the scene is considered in the model,at the same time,the running speed and prediction accuracy are improved obviously.There are still some errors in the prediction model,which are greatly improved after using genetic algorithm to optimize the neural network.In order to compare with other machine learning algorithms,support vector machines are studied in this paper.In view of its advantages of strong generalization ability and global optimization,LS-SVM is used to train the measured field strength,and the model is obtained,and compared with the algorithm results in this paper.By comparison,it is found that the error of ray-tracking-BP network is-3.368?dbm?,the error of ray-tracking-GA_BP is-1.206?dbm?,and the error of LS-SVM is-1.320?dbm?.It can be seen from the results that the optimized hybrid model is the most accurate,followed by LS-SVM,and the error of ray-tracking-BP is relatively large.It can be shown that support vector machines?SVM?are indeed much better than BP neural networks for small sample data.The prediction of the above three schemes verifies the performance of the model.The simulation results show that the mixed model can predict the field strength effectively and accurately.</sup>