Online First have been peer-reviewed and accepted, which are not yet assigned to volumes /issues, but are citable by Digital Object Identifier (DOI).
Video object detection of fully-mechanized coal mining face based on deep neural network
Abstract:
The recognition of the main equipment and personnel in the general mining working face is one of the key directions of intelligent perception of coal mine information. The complex environment with serious influence of coal dust and water mist, and single light source with sparse RGB color space features bring great challenge to the equipment and personnel recognition. In order to recognize the video object in the working face with those issues, this paper proposes a solution frame including dataset production, deep neural network construction, video stream management, and engineering deployment. In the frame, the features of object and background is considered deeply to extend the scene coverage of the production processes, which overcomes the challenge of video sample collection and lays the foundation for improving the robustness of object recognition under the complex scenes of variability; a lightweight deep neural network based on the structure of Darknet is proposed to improve the detection accuracy, and the TensorRT model of the neural network is established to accelerate the calculation speed to achieve the real-time performance; the Gstreamer is employed to the encode, infer, decode and display the large number of the video streams in the work face, hence the real time video object recognition in the working face with multi-source is realized effectively. The experiment consequences and application results show that the MAP, recall rate, and IOU of the proposed framework are 95%, 96.48% and 76.02%, respectively in the dataset , meanwhile the FPSin the processing reaches 67 frames/s.
<Abstract>(0)
Research on LiDAR and IMU fusion method for positioning and mapping in underground coal mines
Abstract:
A real-time localization and mapping method is proposed for the fusion of LiDAR (Light Detection and Ranging) and IMU (Inertial Measurement Unit) to accurately estimate the current situation of the underground environment in the event of a mine disaster. The method introduces a strategy adapted to hierarchical positional estimation in underground coal mines, fusing IMU for hierarchical positional estimation during point cloud matching, giving priority to matching surface features and then line features, which improves the real-time positional estimation while ensuring accuracy. Bayesian tree-based factor map optimization algorithm for incremental optimization estimation of variable nodes, thus being able to improve localization and map building accuracy and reduce global cumulative errors in the mine tunnel environment. Finally, the proposed method is experimentally validated in challenging scenarios such as coal mine tunnels and narrow corridors, and a qualitative and quantitative comparison with the mainstream LeGO-LOAM and LIO-SAM is carried out. The results show that the mean and median error values of the method in all three axes are less than 0.32m, and can achieve good positioning and mapping results in the coal mine tunnel environment. The method can provide technical support for autonomous positioning and navigation of mobile robots in underground coal mines.
<Abstract>(0)
Research on the prediction model of overflow water concentration in concentration tank based on improved SSA-LSTM
Abstract:
Concentration monitoring of overflow water from concentration tank is the key to realize intelligent dosing of slime water in coal preparation plant. Aiming at the problems of difficult on-line real-time monitoring and low control accuracy in the current industrial production based on the direct feedback adjustment of the dosing amount by the sensor, this paper proposes an improved sparrow optimization algorithm to optimize the overflow concentration prediction model of the long-term and short-term memory neural network, and adds the prediction model to the feedback process to overcome the time-space lag of the sensor detection data, so as to achieve real-time monitoring and accurate control. The construction of the model and the optimization of the super parameters are realized through the MATLAB platform, and then the correlation analysis of the overflow concentration multi parameter time series and the noise reduction processing based on the digital signal processing technology are carried out. Firstly, the original data are filtered by the moving mean filter and the Savitzky-Golay smoothing denoising method to reduce the noise and random error; Secondly, chaos mapping, spiral predation and firefly disturbance strategies are introduced to jointly improve the sparrow search optimization algorithm, so as to accelerate the convergence speed of the algorithm and improve its global optimization ability, so as to optimize the relevant super parameters of LSTM, so as to build the ISSA-LSTM overflow concentration prediction model; Finally, taking RMSE and MAPE as the evaluation criteria of model performance, and using the actual monitoring data in the process of slime water concentration and sedimentation, the neural network model is experimentally verified, and compared with the typical double-layer LSTM, SSA-LSTM and LSSVM models to test the superiority of this method. The results show that the proposed ISSA-LSTM overflow concentration prediction model has higher accuracy, and has certain theoretical guiding significance for the real-time monitoring of overflow concentration and the precise adjustment of dosing amount in the coal slurry concentration process.
<Abstract>(0)
Research and application of path optimization of gangue selection robot
Abstract:
The separation of coal and gangue is an important step in the process of coal washing. In order to realize the automatic separation of coal and gangue, liberate labor, improve the quality of selling coal, the application of manipulators to realize gangue separation has become a popular trend. The gangue sorting manipulator designed by our team has been successfully applied in the field of gangue sorting industry. In order to further improve the efficiency of the manipulator,a dynamic sorting path planning method for coal and gangue carried by conveyor belt is proposed. The principle of priority processing in multi-objective state is studied, under the condition of meeting the rated torque. Based on 3-4-5 degree polynomial,two dynamic sorting path planning methods of "ladder" operation path and "V" operation path are proposed. The two dynamic sorting paths are simulated and analyzed in MATLAB. The results show that the separational efficiency of "V" path is 42% higher than that of "ladder" path. Finally, the proposed "V" path is applied to the gangue sorting manipulator used in the industrial field. The application results show that after using the "V" operation path, the sorting efficiency of the gangue sorting manipulator is improved by 19%.
<Abstract>(20)
More
[Special of Intelligent Mining and Excavation Equipment Technology and Application in Coal Mine]
Digital twin: meeting the technical challenges of intelligent fully mechanized working face
GE Shirong, WANG Shibo, GUAN Zenglun, WANG Xuesong, AN Wenlong, LYU Yuanbo, CHEN Shuhang
2022, 48(7): 1-12.   doi: 10.13272/j.issn.1671-251x.17959
Abstract: The goal and task of intelligent fully mechanized working face are to independently complete the reliable coal cutting of the fully mechanized working face, maintain the geometric relationship of the working face and reliable roof support. According to the goal and task, the key technologies of intelligent control of fully mechanized working face are proposed. The technologies include shearer positioning technology, working face visualization technology, hydraulic support electro-hydraulic control technology (device), working face communication technology, collaborative control technology of fully mechanized mining equipment, autonomous height adjustment technology of shearer, autonomous straightening technology of working face and surrounding rock support control technology of working face. Among these technologies, the first three technologies belong to the perception and execution layer of intelligent fully mechanized working face. The working face communication technology is the transmission layer of intelligent fully mechanized working face. And the last four technologies belong to the decision-making layer of intelligent fully mechanized working face. The challenges faced by the intelligent fully mechanized working face are pointed out, which are that the autonomous decision-making capability of the decision-making layer cannot adapt to the complex and changeable working conditions, and the perception and execution layer cannot support the information demand of the decision-making layer and the reliable execution of the decision-making instructions. In order to solve the above challenges, the digital twin system architecture of fully mechanized working face is proposed by use of the simulation-based digital twin modeling method. The virtual entity of the digital twin system of the fully mechanized working face comprises a mechanism model and a behavior model. The unmeasurable data of a physical system of the fully mechanized working face equipment can be obtained by the mechanism model. The behavior model can provide holographic information reflecting the running state of the physical equipment for an intelligent control system of the fully mechanized working face. Thus the problem of the lack of data information in the decision-making layer is solved. The off-line run mode of the combination of the mechanism model of fully mechanized mining equipment and its control system forms the hardware in the loop simulation system of fully mechanized working face, which provides a test platform for intelligent control algorithms based on process rules. The off-line run mode of the combination of the mechanism model, behavior model and its control system of the fully mechanized mining equipment forms the calculation experimental system of the fully mechanized working face, which provides a test platform for the development of the real independent decision-making complex algorithm of the intelligent control system of the fully mechanized working face.
<HTML> <PDF>(21329KB)
[Special of Intelligent Mining and Excavation Equipment Technology and Application in Coal Mine]
Research on pose measurement system of cantilever roadheader based on laser target tracking
XUE Guanghui, LI Yuan, ZHANG Yunfei
2022, 48(7): 13-21.   doi: 10.13272/j.issn.1671-251x.17967
Abstract: The accurate and rapid measurement of the roadheader pose is the premise and foundation of intelligent heading in coal mine roadway. The existing cantilever roadheader pose measurement has the problem of non-absolute pose measurement, low measurement precision, complex arrangement, or only a few pose parameters being measured. The measurement can not meet the intelligent heading requirement. In order to solve the above problems, based on the pose measurement method of the cantilever roadheader based on laser target tracking, a pose measurement system of cantilever roadheader based on laser target tracking is designed. The system consists of a laser tracking device and a laser target. The laser tracking device is arranged behind the roadway, emits laser to the laser target arranged on the cantilever roadheader fuselage, and tracks the movement of the laser target. By solving the conversion matrix between coordinate systems such as laser tracking device, laser target, roadheader and roadway, six absolute pose parameters such as heading direction position, offset, height, deviation angle, pitch angle and roll angle can be measured. The full parameter real-time measurement of absolute pose of cantilever roadheader in roadway coordinate system is realized. The error influence factors of the system are analyzed, and the error distribution law is obtained by simulation. With the increase of heading distance, the pose measurement error of roadheader changes within a certain range. The measurement errors of offset and height increase linearly. Within the measurement range of 5-80 m, the measurement errors of the deviation angle, pitch angle and roll angle of the roadheader are less than 1.4, 1, 0.03° respectively. The measurement error of the heading direction position is less than 5 mm, and the measurement errors of offset and height are less than 20 mm. Using the crawler robot chassis, an experimental system for pose measurement is built. The pose measurement experiment is carried out in the simulated roadway. The results show that the measurement errors of the heading direction position, offset and height are less than 5 mm. The measurement error of the deviation angle is less than 1°. The measurement error of the pitch angle is less than 0.6°. And the measurement error of the roll angle can be ignored. The results are consistent with the simulation results, which proves the reliability of the system.
<HTML> <PDF>(8790KB)
[Special of Intelligent Mining and Excavation Equipment Technology and Application in Coal Mine]
Research on coal-rock interface distribution perception based on near-infrared spectra
YANG En, WANG Shibo, XUAN Tong
2022, 48(7): 22-31, 42.   doi: 10.13272/j.issn.1671-251x.17950
Abstract: Near-infrared reflectance spectra can distinguish coal and rock based on the difference of reflectance spectra characteristics caused by different intrinsic material attributes of coal and rock. This method has high identification accuracy and good real-time performance. But it has not been used for identification of coal-rock interface position distribution. According to the demand for self-determination of the coal-rock interface in the subsequent cutting cycle of shearer memory cutting, the precise distribution sensing technology of coal-rock interface based on near-infrared reflectance spectra technology is studied. A coal-rock interface platform is built by using gas coal and carbonaceous shale cutting block samples. A spectrum detector integrated with optical fiber collimator and tungsten halogen light source is designed and installed on the shearer's body. The near-infrared (1 000-2 500 nm) backward reflectance spectra curves of coal and rock near the coal-rock interface are measured at three walking velocities of the shearer (0, 3, 7 m/s) and four scanning angular velocities of the spectrum detector (3, 4, 5, 6 °/s). For all the reflectance spectra collected by the spectrum detector in each scanning track on the coal wall, the unsupervised identification of coal-rock reflection spectra is carried out based on cosine distance fuzzy C-means clustering (CFCM) in the differential characteristic wave bands of 2 150-2 250 nm. According to the detection results of each position on each scanning trajectory, the theoretical detection position of the coal-rock interface point is determined based on the height difference weighting method and scanning trajectory equation. The research result shows that under each movement state of the shearer and the spectrum detector, the near-infrared reflectance spectra in the backward direction of gas coal and carbonaceous shale collected by the integrated optical fiber collimator tungsten halogen light source spectrum detector have obvious differential absorption valley bands around 1400, 1900, 2200 nm. The reflectance spectra curves of coal and rock all show a downward trend with the increase of the detection incident angle. With the increase of the scanning angular velocities of the spectrum detectors under the same walking velocity of the shearer, and with the increase of the walking velocity of the shearer under the same scanning angular velocity of the spectrum detector, the reflectance spectra curves of coal and rock tend to be flat as a whole. Based on CFCM, height difference weighting method and coal wall scanning trajectory equation, rapid and precise detection of coal-rock interface points under the movement of the shearer and spectrum detector can be realized. Among them, the root mean square error of the detection results of coal-rock interface points under three scanning angular velocities of spectrum detectors 3, 4, 5 °/s is not more than 1.5 cm. The research provides a reference for the application of near-infrared reflectance spectra technology to the precise and efficient perception of coal-rock interface distribution.
<HTML> <PDF>(5013KB)
[Special of Intelligent Mining and Excavation Equipment Technology and Application in Coal Mine]
Development of coal and rock identification device based on near-infrared spectroscopy
LYU Yuanbo, WANG Shibo, GE Shirong, ZHOU Yue, WANG Saiya, BAI Yongtai
2022, 48(7): 32-42.   doi: 10.13272/j.issn.1671-251x.17953
Abstract: The current near-infrared spectroscopy identification of coal and rock is to collect spectral data in a static state for offline identification. The technology cannot meet the need for real-time identification of high-speed moving coal and rock on conveyor during caving operation. In order to solve this problem, a coal and rock identification device is developed based on near-infrared spectroscopy technology. The device consists of a data acquisition and processing device, and a light source and probe integrated device. The light source and probe integrated device is used to collect the reflected light of coal and rock. The improved coal and rock identification algorithms (cosine angle algorithm and correlation coefficient method) in the data acquisition and processing device is used to analyze the spectrum data. The spectrum information can be analyzed immediately after obtaining a coal and rock spectrum curve. Then the current coal and rock type can be determined. In order to obtain the best characteristic band and standard spectral library size of the improved coal and rock identification algorithms, the effects of different characteristic bands and standard spectral library sizes on the identification accuracy are obtained through experiments. The characteristic widths of 1 300 -1 500 nm, 1 800-2 000 nm and 2 100-2 300 nm are suitable for most coal and rock samples. The size of the standard spectral library is positively correlated with the accuracy. It is necessary to increase the number of curves in the standard spectral library during identification. In order to improve the spectral quality collected by the coal and rock identification device, the relative motion of coal and rock and the light source and probe integrated device is simulated in the laboratory. The influence law of different spectral acquisition parameters on spectral quality is explored. The integration time mainly refers to the light intensity of the light source. When the acquisition conditions are good, the integration time should be set to be slightly higher than the lower limit by 5-10 ms. For the fully mechanized top coal caving face, the real-time requirement of coal and rock identification is high, and the coal and rock on the scraper conveyor change rapidly during the coal caving process. The integration number is set to one for the best. The smoothing times mainly refer to the speed of environmental fluctuation, which can be set to eliminate the change of ambient light. In order to improve the identification accuracy of coal and rock identification device in the coal flow movement state of working face, the identification accuracy of improved cosine algorithm and correlation coefficient method in the relative movement of coal and rock and light source and probe integrated device is explored. The improved correlation coefficient method is more suitable for the identification algorithm used in working face, and the accuracy rate is 91.3%. The results of the coal and rock identification test in coal mine show that after collecting the spectral curves of coal and rock in a coal drawing cycle, the device immediately analyzes the spectral information and determines the current coal and rock category by the improved identification algorithm. The device realizes the real-time identification of coal and rock in the coal drawing process.
<HTML> <PDF>(5913KB)
[Special of Intelligent Mining and Excavation Equipment Technology and Application in Coal Mine]
Research on the cantilever roadheader positioning based on near-infrared binocular stereo vision
WANG Xue, ZHOU Hongxu, ZHANG Lei, WANG Huaying
2022, 48(7): 43-51, 57.   doi: 10.13272/j.issn.1671-251x.17896
Abstract: The existing roadheader has problems, such as unable real-time positioning, inaccurate positioning, and positioning failure caused by camera view occlusion in visual positioning. In order to solve the above problems, a positioning scheme of the cantilever roadheader based on near-infrared binocular stereo vision is proposed. A near-infrared LED target is arranged on the fuselage and arm of cantilever roadheader. Taking LED as the near-infrared target, the characteristic information of the roadheader is constructed. The three-dimensional spatial positioning of the roadheader fuselage and the cutting part is realized through image processing and pose calculation. The binocular stereo vision camera is arranged at the top of a roadway. The distance between the roadheader and the binocular stereo vision camera gradually increases as the roadheader continues to advance. It leads to the failure of binocular image acquisition, which leads to the failure of the visual solution of the pose of the cutting part. In order to solve this problem, a magnetic field assisted positioning method of the cutting part based on one-dimensional convolution neural network (1D-CNN) is introduced. Three-axis digital magnetometers are arranged on two sides of the fuselage of the roadheader. The permanent magnet is arranged on the machine arm. The strength component of a magnetic field and pose data obtained by binocular stereo vision camera are used as training data to construct the 1D-CNN model, and the pose of a cutting part of the roadheader is output under the condition that vision measurement fails. The cantilever roadheader positioning based on near-infraared binocular stereo vision scheme is tested from the aspects of depth information and its fuselage of the roadheader and the cutting position are verified. The results showed that the measurement error of the fuselage is within ±11 mm, and the relative error is within 0.4%. The measurement error of the cutting part is within ±50 mm, and the relative error is within 1%.The relative pose error between the roadheader fuselage and the cutting part is within ±2.5°, the root-mean-square error of the pitch angle is 0.930 1°, and the root-mean-square error of the yaw angle is 0.922 0°. The errors are within the allowable range of roadway operation. The results show that the cantilever roadheader positioning scheme based on near-infrared binocular stereo vision is effective and reliable. The effectiveness of the magnetic field assisted positioning method based on 1D-CNN is verified. In order to simulate the complex magnetic field environment in coal mine underground, the interference magnetic source is randomly added near the roadheader. The results show that the predicted values of the pitch angle, yaw angle and rolling angle of the cutting part of the roadheader by this method are basically consistent with the measured real values. The determination coefficients of the predicted pitch angle, yaw angle and rolling angle are 0.992 4, 0.995 9 and 0.917 4 respectively. It shows that the magnetic field assisted positioning method of the cutting part of the roadheader based on 1D-CNN can better meet the positioning requirements of the roadheader in the case of visual positioning failure.
<HTML> <PDF>(9661KB)
[Special of Intelligent Mining and Excavation Equipment Technology and Application in Coal Mine]
Shearer positioning method based on non-holonomic constraints
SONG Danyang, YANG Jinheng, TAO Xinya, LU Chungui, TIAN Muqin, SONG Jiancheng
2022, 48(7): 52-57.   doi: 10.13272/j.issn.1671-251x.2022020006
Abstract: At present, the shearer positioning method is based on the combination of the inertial navigation system and odometer. The method directly uses the output of the odometer to correct the shearer forward speed calculated by the inertial navigation system. However, the capability of suppressing the error divergence of the inertial navigation system is very limited. The shearer in the process of movement meets the characteristics of the non-holonomic constraints. When the shearer does not jump and sideslip, the lateral velocity and vertical velocity at the connection between the traction gear and the crawler are zero. Based on this characteristic, a new shearer positioning method based on non-holonomic constraints is proposed on the basis of the combination of the inertial navigation system and odometer. The output of the inertial measurement unit arranged in the middle of the shearer's body is mechanically arranged, so as to obtain the attitude, speed and position information of the shearer. The output of the odometer installed on the traction gear of the shearer is used to calculate the instantaneous velocity of the shearer. The Kalman filtering state equation is established by using a mechanical arrangement result of the inertial navigation system and an error propagation model. The non-integrity constraint is introduced at the joint of a traction gear and a crawler of the shearer. The Kalman filtering observation equation is established by using the difference between the velocity projected at the joint by the inertial navigation system and the velocity output by the mileometer as an observation vector. The output of the inertial navigation system is modified by using the results of the Kalman filtering algorithm as error feedback. Then the optimal estimation of the attitude, speed and position of the shearer is obtained. The experimental results show that compared with the traditional combined positioning method of inertial navigation system and odometer, the positioning error does not diverge with time after the non-holonomic constraint is added. The positioning method has good tracking performance on the actual trajectory. The positioning errors of the shearer in the forward, lateral and vertical directions are reduced by 66%, 62% and 67% respectively.
<HTML> <PDF>(1951KB)
[Special of Intelligent Mining and Excavation Equipment Technology and Application in Coal Mine]
Study on dynamic modification method of 3D model of coal seam in fully mechanized working face
LIANG Shua, WANG Shibo, GE Shirong, BAI Yongtai, XIE Yang
2022, 48(7): 58-65, 72.   doi: 10.13272/j.issn.1671-251x.17956
Abstract: The high-precision coal seam geographic information of fully mechanized working face is the key to realizing intelligent unmanned mining. However, the vertical precision of 3D model of coal seam constructed at this stage is low. The model cannot meet the actual needs of intelligent mining. In order to solve this problem, a dynamic modification method of 3D model of coal seam in fully mechanized working face is proposed. The static data of the initial coal seam 3D model and the dynamic data generated by the shearer cutting in the mining process are fused. The method is based on the prediction algorithms of long-short term memory (LSTM) network and its improved algorithm. The improved algorithms are based on the convolutional long-short term memory network (Conv LSTM) and encoder-decoder long-short term memory network (Encoder-Decoder LSTM). The coal seam floor curved surface and the coal seam thickness of the unmined area in the next stage are dynamically predicted according to the coal seam data of the previous mining stage. The parameters of the above three prediction algorithms are optimized by using the grid search method of double-layer loop nesting. The obtained high-precision vertical distribution data of the coal seam floor curved surface and the coal seam thickness of the unexploited area are taken as the coal seam 3D model correction value. The correction value is used to dynamically correct the coal seam 3D model of the unexploited area in the next stage. With the continuous mining of the working face, the newly obtained correction data is used to continuously and dynamically correct and update the initial coal seam 3D model, so as to improve the precision of the initial coal seam 3D model. Therefore, the dynamic modified coal seam 3D model can reflect the actual coal seam distribution of fully mechanized working face more accurately. Taking the coal seam 3D model of 18201 working face of a coal mine in Lvliang, Shanxi Province as an example, the proposed dynamic correction method is used to correct the coal seam 3D model. Within the range of 16-23.2 m in the advancing direction of the working face, the average error of the coal seam floor after the dynamic correction is 0.068 5 m. The average error of the coal seam roof is 0.076 m. Compared with the average floor error of 0.20 m and vertical average error of 0.40 m of the coal seam thickness before correction, the precision of the coal seam 3D model after dynamic correction is greatly improved. The results confirm the effectiveness of the correction method.
<HTML> <PDF>(4068KB)
[Special of Intelligent Mining and Excavation Equipment Technology and Application in Coal Mine]
Research on large flow intelligent liquid supply system in fully mechanized working face
SI Ming, WU Bofan, WANG Ziqian
2022, 48(7): 66-72.   doi: 10.13272/j.issn.1671-251x.2022030033
Abstract: The liquid supply system in fully mechanized working face has the problems of insufficient liquid supply capacity, large pressure fluctuation and poor system operation stability. In order to solve the above problems, an immune particle swarm optimization fuzzy neural network PID (IPSO-FNN-PID) algorithm is proposed. The IPSO-FNN-PID controller is designed to stabilize the pressure of the liquid supply system. In the IPSO-FNN-PID algorithm, a particle swarm optimization (PSO) algorithm and an immune algorithm (IA) are introduced into a fuzzy neural network (FNN) PID controller. The immune particle swarm optimization (IPSO) algorithm is used to solve the problem that the FNN algorithm is easy to fall into local optimization. The IA is added to the PSO algorithm to improve the convergence of the PSO algorithm. Therefore, the output of the optimal PID parameters is realized. In order to verify the effectiveness of the IPSO-FNN-PID controller, traditional PID controller, Fuzzy-PID controller and FNN-PID controller are selected to compare. The simulation results show that the IPSO-FNN-PID controller has the best control effect on the emulsion pump. The rise time, peak time and regulation time of the other three controllers are longer than the IPSO-FNN-PID controller. The maximum overshoot is greater than the IPSO-FNN-PID controller. After adding the disturbance signal, the IPSO-FNN-PID controller has good adaptability and robustness, and it takes only 1.2 s to restore to a stable state. When traditional PID and Fuzzy-PID controllers are used to control the emulsion pump, the oscillation is obvious and the overshoot is large, which are 41.2% and 22.3% respectively. When the FNN-PID controller is used to control the emulsion pump, the oscillation is significantly weakened, the overshoot is reduced to 17.6%, and the adjustment time is reduced to 2.68 s. When the IPSO-FNN-PID controller is used to control the emulsion pump, there is almost no oscillation. The overshoot is only 5.22%, the adjustment time is shortened to 2.61 s. And the stability is stronger when encountering interference signals. When the disturbance signal is received, the load disturbance has little effect on the IPSO-FNN-PID controller, the convergence is rapid, and the robustness is greatly improved. The results show that the IPSO-FNN-PID controller has good anti-disturbance and disturbance compensation capability, and can meet the pressure stabilization control requirements of the liquid supply system.
<HTML> <PDF>(2349KB)
More

Video RecommendationMore>

Industry NewsMore>