Research On Shearer Positioning Technology With Heterogeneous Multi-source Information Fusion

With the developing of coal mining industry toward safe,efficient and green mining,improving the intelligent level of mining equipment and automation level of longwall mining is an effective way to realize the intelligent and unmanned mining,which is also an inevitable choice for the transformation and upgrading of coal production.As one of three kinds of mining equipment in the longwall face,the shearer undertakes the important task of cutting coal-wall.The armored face conveyor provides the running track for the shearer,the roof support system supports the cutting space,the armored face conveyor and roof support system both serve the shearer,so the shearer is the core equipment in the longwall face.The shearer positioning technology is an important research content of mining equipment intelligentization.In addition,the shearer automatic height-adjustment and automatic face-alignment of the armored face conveyor based on the shearer positioning technology are the necessary functions of longwall mining automation.Therefore,the shearer positioning technology is of great significance for the intelligent and unmanned mining.In this thesis,the shearer positioning technology with heterogeneous multi-source information fusion is studied,in which the inertial navigation system(INS)is the core element,and the auxiliary information sources are the axial encoder,dynamic zero-velocity update technology,kinematic model,and calibration information at the end of longwall face,respectively.The main research contents and conclusions are as follows:(1)The integrated navigation with inertial navigation system and axial encoder is selected as the basic positioning method.By analyzing the dead reckoning algorithm,the shearer positioning accuracy is mainly affected by the INS attitude angle drift errors.The allan variance is used to analyze the noise source of the measurement error of INS gyroscopes.It is found that the angle random walk is the main component,that is,the attitude angle drift error is a random error caused by the integration of white noise introduced into the gyroscopes.Through analyzing the shearer motion characteristic along the longwall face,the armored face conveyor restricts the velocity direction and the velocity only has component along the running track direction.Therefore,based on the constraint condition of dynamic zero-velocity update technology,the kalman filter model is established.The state vector is the platform attitude errors in the east,north and up,and the measurement vector is the velocity values along the shearer horizontal and vertical direction provided by INS.The simulation experiment of shearer running with 54 m length is carried out.The results show that the maximum errors during the third cutting cycle in the east and north decrease from 0.6397 m,0.8567 m to 0.4564 m,0.5942 m,the maximum errors during the fourth cutting cycle in the east and north reduce from 0.6444 m,0.9106 m to 0.4665 m,0.6030 m,respectively.The positioning accuracies in the east and north increase by about 30%.(2)According to the longwall mining method,the shearer kinematics model is obtained.The shearer motion is restricted by the velocity constraint in the longwall face and position constraint in the advancing direction.Based on the information filter which is suitable for the multi-observation system,shearer positioning method aided by kinematic model is proposed.The state equation is established according to the differential equations of INS position,velocity and attitude errors.Based on the constraint conditions in the kinematics model,the velocity and position measurement equations are derived.Using the information filter to fuse two measurement information,the attitude errors are estimated,which are applied to compensate the INS attitude output.Then,the shearer positioning trajectory aided by kinematic model is calculated by dead reckoning algorithm.The simulation experiment with 105 m length is performed.In the east,the variations of positioning error during four cutting cycles are 0.8465 m,1.9637 m,1.9497 m and 1.9780 m,respectively.The variations of positioning error in the north are 0.6184 m,1.4223 m,1.4182 m and 1.4238 m,respectively.The experiment results show that the variations of positioning error during third and fourth cutting cycles no longer increase continuously,which indicates the information filter fusion model basically restrains the drift errors of INS attitude angle.(3)In order to obtain the high-resolution radar image in near-field environment of underground coal mine,the range Doppler algorithm adopts the slant range model with fourth order.A new denoising algorithm based on maximum a posteriori estimation and modified guide filter is developed in wavelet domain to remove speckle noise in radar image.Because Gaussian mixture model performs better in describing the statistical characteristics of the wavelet coefficients of high-frequency subbands compared with Gaussian model and two dimensional-generalized autoregressive conditional heteroscedasticity(2D-GARCH)mode,the maximum a posteriori estimation method based on Gaussian mixture model is proposed.For the speckle noise in the low-frequency subband,the modified guide filter is adopted to protect the feature information in the process of noise suppression,in which a new edge-protecting factor is introduced.The simulation and real SAR experiment results show that the proposed algorithm is effective in removing speckle noise and protecting the feature information,which is beneficial to subsequent image registration and other operations.(4)A method measuring the calibration information based on radar imaging is developed to independently and accurately obtain the longwall retreat and creep displacements.In this method,the calibration information is obtained by registering two images which are formed before and after the movement of longwall face.According to this,an image registration based on improved synthetic aperture radar scale invariant feature transform(SARSIFT)is proposed to accurately match the feature points.The improved SARSIFT uses bilateral filter instead of Gaussian filter to blur the image.The circular neighborhood around the feature point is adopted to calculate the gradient,and the cosine value of Euclidian distance between the feature points is applied to achieve the initial matching.In order to eliminate the wrong matching points in the initial matching,an enhanced feature matching algorithm based on principal component analysis is developed in the improved SARSIFT.The results of simulation experiment and underground field test in calibration information measurement method show that the average measurement errors of longwall retreat and creep displacements are less than 10 mm.(5)Drawing on the back-end optimization method in the simultaneous localization and mapping,a global optimization method of shearer positioning trajectory is proposed.When the shearer runs to the end of longwall face,the shearer positioning trajectory during the longwall face is back-end optimized with the calibration information,so as to realize the integrated positioning of five information sources such as INS,axial encoder,dynamic zero-velocity update technology,kinematic model,and calibration information.The results of simulation experiment with 105 m length show that the average positioning errors with global optimization method in the north and east are 0.0634 m and 0.0786 m,respectively.The shearer positioning system with heterogeneous multi-source information fusion is developed.The hardware mainly includes the shearer positioning device with inertial navigation system and the shearer calibration information measurement device.The upper-computer program and its interface are written and designed by Qt software under Linux system,and the program language is C++.The results of simulation experiment of 300 m length with the shearer positioning system show that the spherical error probability(SEP)values during the first and second cutting cycle are0.0634 m and 0.0786 m,respectively.A full-scale equipment experiment with 27 m length in the State Key Laboratory for Mining Equipment and Intelligent Manufacturing is performed.The SEP values in the first group experiment are 0.0097 m and 0.0110 m,and SEP values in the second group experiment are 0.0084 m and0.0103 m,respectively.There are 109 figures,23 tables and 199 references in this dissertation.