工矿自动化

Intelligent fault diagnosis of hoist bearing based on feature transfer learning

PAN Xiaobo, GE Kunpeng, DONG Fei

2022, 48(9): 1-7, 32. doi: 10.13272/j.issn.1671-251x.17980

<Abstract>(315) <HTML> (105) <PDF>(72)

Abstract:
The complex actual working conditions of the hoist causes the problems of low accuracy and weak adaptability of existing fault diagnosis methods. In order to solve these problems, an intelligent fault diagnosis method of hoist bearing based on deep transferable feature selection(DTF) and balance distribution adaptation(BDA) is proposed. The bearing fault signals under different working conditions are subjected to time-frequency analysis. The time and frequency domain's statistical characteristics are extracted. The high-dimensional depth characteristics are extracted by adopting a deep belief network. In order to select features that are beneficial to fault mode identification and cross-domain fault diagnosis from a high-dimensional depth feature set, the transferable feature selection based on ReliefF and differences between domains(TFRD) method is adopted. The method carries out the quantitative evaluation of the transitivity of each feature. The TFRD method carries out the quantitative evaluation on the class discrimination and domain invariance of each feature. The ReliefF algorithm processes various feature data to obtain weight values representing class discrimination. This method calculates the maximum mean discrepancy of the same feature between different domains, and constructs a new quantitative index of feature transferability. Based on the TFRD method, depth features with high feature transferability are selected to construct feature subsets. The balance distribution adaptation is applied to carry out distribution adaptation on the feature subsets of the source domain and the target domain, so as to reduce the distribution difference between the two domains. The source domain feature set is used to train the fault pattern identification classifier, and the target domain samples are used for fault identification and classification. Eight fault diagnosis models are constructed by using the classical machine learning method, deep learning method and transfer learning method. The models are used for comparing the fault diagnosis accuracy with the proposed DTF-BDA fault diagnosis model. The results show the following points. ① The DTF-BDA fault diagnosis model can achieve better performance than other models, and the highest fault diagnosis accuracy can reach 100%. ② The TFRD method can effectively improve the performance of the fault diagnosis model based on the transfer learning method. The highest fault diagnosis accuracy can reach 96.46% and 97.67% respectively when combined with the transfer component analysis and joint distribution adaptation.

Belt conveyor deviation and coal stacking monitoring method based on three-dimensional point cloud

XU Shichang, CHENG Gang, YUAN Dunpeng, SUN Xu, JIN Zujin, LI Yong

2022, 48(9): 8-15, 24. doi: 10.13272/j.issn.1671-251x.17948

<Abstract>(209) <HTML> (37) <PDF>(51)

Abstract:
Conveyor belt deviation and coal stacking are common faults of belt conveyor in the coal mine. The traditional contact conveyor belt deviation or coal stacking detection methods can not meet the requirements of coal mine safety production in terms of durability, sensitivity and reliability. However, the detection effect based on the image processing method is greatly affected by image color information, which is prone to false identification. The belt conveyor deviation and coal stacking monitoring method based on 3D point cloud is proposed. The 3D point cloud data of the conveyor belt surface is collected by line laser binocular camera. The real-time monitoring of belt deviation and coal stacking is carried out by analyzing and processing the point cloud data. In terms of conveyor belt deviation monitoring, Euclidean clustering and random sampling consistency algorithm are used to filter redundant point cloud data, and extract edge data points of the conveyor belt. The mean central characterization value is used to characterize the degree of conveyor belt deviation, so as to reduce the influence of the shape change in the width direction of the conveyor belt on the monitoring. In terms of coal stacking monitoring, the equivalent height of coal flow is obtained by processing point cloud data. The height and width information of coal flow is characterized by the equivalent height, so that the coal stacking degree is evaluated in real-time. The test bed of the belt conveyor deviation and coal stacking monitoring system is built. The test results show the following points. When the speed of the conveyor belt is 0.5-3.0 m/s, the detection error of the edge point of the conveyor belt is − 2.84-1.26 mm, and the maximum error is only 2.84 mm. It shows that the system can reliably realize the function of deviation fault monitoring and accurately predict the deviation trend. Coal samples (14-41 kg, in increments of 1 kg) are stacked on the conveyor belt. When the coal mass is within the range of 14-24 kg and 28-41 kg, the coal stacking detection results are correct. There are detection errors in the range of 25-27 kg. The reason is that the coal sample quality in this range is close to the critical value of 27.6 kg for triggering the coal stacking alarm.

Tear detection method of conveyor belt based on fully convolutional neural network

YOU Lei, ZHU Xinglin, CHEN Yu, LUO Minghua

2022, 48(9): 16-24. doi: 10.13272/j.issn.1671-251x.2022040087

<Abstract>(312) <HTML> (49) <PDF>(46)

Abstract:
The existing conveyor belt tear detection methods have problems, such as poor underground visible light imaging quality, lack of tear physical size measurement means, and poor generalization capability. In order to solve these problems, a conveyor belt tear detection method based on fully convolutional neural network is proposed. The method collects images based on a line-structured light imaging principle, and can effectively solve the problem of poor lighting conditions in a coal mine. The improved maximum method is used to detect line laser stripes, which can effectively eliminate the breakpoints of stripes, accurately extract stripes, and fit the missing points. The U-net network in the fully convolutional neural network is selected to segment the line laser stripe. The tear detection problem is converted into a semantic segmentation problem. The U-net network is optimized through dimension reduction, so as to reduce the number of parameters and calculations. The segmentation result is back-projected to the original image. The physical size of the tear is measured using the line-structured light calibration data. The experimental results show that the improved maximum method can effectively deal with the breakpoint area of line laser stripes without false detection and missed detection. The performance is superior to the Steger method and gray-weighted centroid method. The convergence speed of the U-net network is faster than that of the SegNet and FCNs network. The iteration stability is strong, and the evaluation index is optimal. The performance of the U-net4 network is better than that of U-net3 and U-net5. The test results on the verification set show that the recall rate of tear detection is 96.09%, and the precision is 96.85%. The measurement results on the experimental platform show that the maximum relative error of tear physical dimension measurement is −13.04%.

Audio fault diagnosis method of mine belt conveyor roller

WU Wenzhen, CHENG Jiming, LI Biao

2022, 48(9): 25-32. doi: 10.13272/j.issn.1671-251x.2022070071

<Abstract>(352) <HTML> (58) <PDF>(45)

Abstract:
In the existing fault diagnosis method of mine belt conveyor roller, the roller signal is decomposed and converted to the frequency domain. The fault diagnosis is carried out by extracting characteristics from the frequency domain. The common signal decomposition methods include wavelet decomposition and empirical mode decomposition. The methods have the problems of difficult selection of wavelet basis, frequency spectrum aliasing and endpoint effect, resulting in low fault diagnosis accuracy rate. In view of the above problems, an audio fault diagnosis method of mine belt conveyor roller mine based on variational modal decomposition (VMD)-BP neural network is proposed. Firstly, the audio signal of the roller along the mine belt conveyor is collected by the audio sensor. The audio signal is preprocessed to suppress the noise signal in the audio information. Secondly, VMD is used to decompose the audio signal into different IMF (intrinsic mode function) components according to the center frequency. The method extracts characteristic values of the kurtosis, gravity frequency, frequency standard deviation of each IMF component. Finally, the characteristic values are input into the trained BP neural network. According to the difference in IMF component characteristic values, it is possible to diagnose the mine belt conveyor roller fault through audio, and determine the position of the faulty roller according to the sensor number corresponding to the audio signal. The audio information of the roller of the belt conveyor collected in a coal mine is used to analyze and verify the audio fault diagnosis method of mine belt conveyor roller based on VMD-BP neural network. The results show that the method can avoid spectrum aliasing and endpoint effect in the decomposition process when decomposing and extracting audio signal characteristics. The overall fault diagnosis accuracy rate reaches 96.15%. Compared with the fault diagnosis method based on BP neural network and the fault diagnosis method based on wavelet decomposition and BP neural network, the proposed method has improved the fault diagnosis accuracy rate by 26.92% and 15.38% respectively. The false detection rate has also been significantly reduced.

Construction of health index and condition assessment of coal mine rotating machinery

LI Man, PAN Nannan, DUAN Yong, CAO Xiangang

2022, 48(9): 33-41. doi: 10.13272/j.issn.1671-251x.18004

<Abstract>(184) <HTML> (28) <PDF>(38)

Abstract:
The monitoring parameters of coal mine equipment are time-series data, and the time-series characteristics have great influence on health assessment. The traditional mechanical equipment health assessment has the problems of incomplete extraction of signal spatiotemporal characteristics, high dependence on human experience, and difficult assessment of early condition change of equipment. In order to solve these problems. A two-dimensional array long short-term memory denoising convolutional autoencoder (2D-LSTMDCAE) model is constructed, and a health index (HI) construction and condition assessment method of coal mine rotating machinery based on 2D-LSTMDCAE is proposed. The one-dimensional vibration data is converted into a two-dimensional array. The two-dimensional convolution network model is used to fully learn the information contained in the original data, so the learning capability of the model on data characteristics is enhanced. The samples are input into convolution and long short-term memory (LSTM) units in parallel to obtain complete signal spatiotemporal characteristics. The unsupervised learning denoising convolutional autoencoder (DCAE) model is constructed for sample reconstruction. The similarity between the original sample and the reconstructed sample is calculated by Bray-Curtis distance to obtain the HI. It solves the problem that it is difficult to obtain the condition tag during the operation of the equipment, and improves the adaptability of the model in strong background noise. The characteristic learning capability of the 2D-LSTMDCAE model is verified by using XJTU-SY bearing data set. The two indexes of correlation and monotonicity are adopted to evaluate the condition assessment method based on HI. The test results show the following points. The two-dimensional input sample construction method and the HI construction method of learning the time series characteristics of data are more sensitive to the performance degradation of bearings. The 2D-LSTMDCAE model can detect the early failure of the equipment earlier. On the test bearing, the HI and RMS constructed by the 2D-LSTMDCAE model are about 7 min earlier than that of the LSTMDCAE and DCAE models. Compared with the HI and RMS constructed by the LSTMDCAE and DCAE models, the HI constructed by the 2D-LSTMDCAE model has higher correlation and monotonicity, and it can better reflect the degradation of bearings. The health assessment experiment is carried out by using the accelerated degradation experimental data of the reducer. On the test reducer, compared with RMS, it can detect early failure 8 min in advance by using the HI constructed by the 2D-LSTMDCAE model. The correlation is improved by 0.007, and the monotonicity is improved by 0.211, which can better reflect the degradation situation of the reducer.

Fault diagnosis of rigid guide based on wavelet transform and improved convolutional neural network

DU Fei, MA Tianbing, HU Weikang, LYU Yinghui, PENG Meng

2022, 48(9): 42-48, 62. doi: 10.13272/j.issn.1671-251x.17964

<Abstract>(198) <HTML> (37) <PDF>(22)

Abstract:
Some of the existing fault diagnosis methods for rigid guide are only suitable for small sample data sets. Although some methods are suitable for large sample data sets, they ignore the multi-condition background in the actual working environment. The method of rigid guide fault diagnosis based on the convolutional neural network has the problems of huge data and computation, and easy to produce over-fitting. In order to solve these problems, a fault diagnosis method of rigid guide based on wavelet transform and improved convolutional neural network is proposed. Firstly, two kinds of defects, dislocation and gap, are set in the rigid cage guide. The vibration acceleration signals of the hoisting container under multiple working conditions are collected. Secondly, the collected vibration acceleration signals are converted into two-dimensional time-frequency images by wavelet transform. The time and frequency resolution of the two-dimensional time-frequency images processed by the Complex Morlet wavelet basis function is determined to be the best by trial and error method. Thirdly, the structure of the convolutional neural network model is adjusted. The first pooling layer and the fifth pooling layer are reserved. The second pool layer, the third pooling layer and the fourth pooling layer are replaced by small-scale convolutional layers to prevent the over-fitting phenomenon. Finally, the two-dimensional time-frequency image is input into the improved convolutional neural network model. The experimental results show the following points. ① After training, the average accuracy of the improved model is about 99% on the training set and 99.5% on the test set. ② When the training data reaches 200 steps, the accuracy of the improved model is more than 99%, and the loss function of the improved model approaches 0. These results show that the improved model has good convergence performance, and the generalization of the model is enhanced. The inhibition effect on over-fitting in the learning process is obvious. ③ On the confusion matrix of the validation set, the identification rate of gap defect and dislocation defects is 100%. The identification rate of no defect is 92%, and 8% of the defect are mistakenly identified as gap defects. ④ Compared with EMD-SVD-SVM, wavelet packet-SVM, EMD-SVD-BP neural network and wavelet packet-BP neural network, the accuracy of rigid guide fault diagnosis method based on wavelet transform and the improved convolutional neural network reaches 99%.

Bearing intelligent fault diagnosis

WU Dongmei, WANG Fuqi, LI Xiangong, TANG Run, ZHANG Xinjian

2022, 48(9): 49-55. doi: 10.13272/j.issn.1671-251x.17986

<Abstract>(273) <HTML> (57) <PDF>(45)

Abstract:
Bearing vibration signal is a kind of time series data, and its time dimension characteristic plays a key role in classification. Using convolutional neural network (CNN) alone to diagnose bearing fault will cause the loss of time dimension information. This results in the decline of diagnosis accuracy. To solve the above problems, a bearing fault diagnosis model combining one-dimensional CNN, bidirectional gated recurrent unit (Bi GRU) and attention mechanism is proposed. Firstly, CNN is used to adaptively extract the local space characteristic of one-dimensional vibration signals. Secondly, the characteristic information is taken as the input of the Bi GRU. Bi GRU is used to perform time dimension fusion on the extracted characteristic information. The attention mechanism is introduced to weigh the characteristic information of a plurality of moments so as to extract a more critical fault characteristic. Finally, the fault characteristic is input into a full connection layer to obtain a classification result, so as to realize intelligent fault diagnosis of the bearing. The experimental result shows the following points. ① On the confusion matrix of the test set, the classification of the bear running state is basically correct. Only some mark types are not completely classified correctly. But the recall rate is more than 95%, and the total fault recognition accuracy rate is 99.3%. ② The t-SNE technology is used to visualize the data after dimensionality reduction processing. The data of each running state of the bearing are well gathered in their own space. Only a small amount of data are mixed into other areas, which shows that the model has strong characteristic extraction capability. ③ Under the condition of constant load, the average accuracy of fault diagnosis of this model is 0.8%, 0.6% and 0.3% higher than that of one-dimensional CNN, Bi GRU and attention CNN models respectively. ④ Under the condition of variable load, this model has better stability than SVM, one-dimensional CNN, Bi GRU, attention CNN and other models. When the load is 2.25 kW, the accuracy rate is more than 85%. The model has the capability to extract one-dimensional CNN local characteristics and the capability to model Bi GRU time-dependent information. The model can further fuse time dimension information among the characteristics after acquiring the bear signal local complex characteristics. And the attention mechanism can further pay attention to the characteristics more relevant to faults. Therefore, the model has better precision.

Fault diagnosis of planetary gearbox based on multi-information fusion and convolutional neural network

SHI Zhiyuan, TENG Hu, MA Chi

2022, 48(9): 56-62. doi: 10.13272/j.issn.1671-251x.2022060011

<Abstract>(258) <HTML> (36) <PDF>(39)

Abstract:
The fault diagnosis method of planetary gearbox based on machine learning relies on the artificial selection of the eigenvectors. The quality of eigenvectors selection largely determines the accuracy of the diagnosis method. The convolutional neural network (CNN) can extract characteristics automatically. But it is difficult to accurately diagnose the fault from a single vibration signal when it is used for the planetary gearbox fault diagnosis. To solve the above problems, a fault diagnosis method of planetary gearbox based on multi-information fusion and CNN is proposed. The method performs data layer fusion on three-dimensional (horizontal radial direction, vertical radial direction and axial direction) vibration signals and sound signals of the planetary gearbox. The one-dimensional vibration signals and sound signals are integrated into two-dimensional signals in a parallel connection mode. The two-dimensional signals are used as the input of CNN. The multiple convolutional layers and maximum pooling layers are used for depth characteristic extraction and information filtering. Finally, the Softmax classifier is used to achieve fault classification. The fault diagnosis experiment platform of the planetary gearbox is built. The vibration signals and sound signals of normal and fault states of the planetary gearbox under different speed and load conditions are collected and input into CNN for training and verification. Four single-source information of horizontal radial vibration signal, vertical radial vibration signal, axial vibration signal and sound signal are selected under the same conditions and combined with CNN respectively for comparison. The experiment is used to verify the superiority of the fault diagnosis method for planetary gearbox based on multi-information fusion and CNN. The experimental results show that the fault identification accuracy of the two methods of axial vibration signal+CNN and sound signal+CNN is 74.07% and 75.13% respectively. The fault identification accuracy of the two methods of horizontal radial vibration signal+CNN and vertical radial vibration signal+CNN is 89.70% and 87.09% respectively. The method based on multi-information fusion and CNN has the fastest convergence speed and the highest fault identification accuracy, which is 93.33%.

Fault diagnosis method of rolling bearing based on MTF and DenseNet

JIANG Jiaguo, GUO Manli

2022, 48(9): 63-68. doi: 10.13272/j.issn.1671-251x.17985

<Abstract>(284) <HTML> (72) <PDF>(28)

Abstract:
The fault diagnosis methods of rolling bearing based on model and signal processing and analysis have the problems of modeling difficulty and signal characteristic extraction difficulty. The rolling bearing fault diagnosis method based on shallow machine learning has limited capability to learn the characteristics of complex data. The convolutional neural networks are often used in rolling bearing fault diagnosis methods based on deep learning. But with the deepening of the network, gradient dispersion or disappearance will occur. And directly converting the rolling bearing vibration signal into one-dimensional or two-dimensional images as network input will not preserve the time correlation between the signals, resulting in the loss of signal information. To solve these problems, a fault diagnosis method for rolling bearing based on Markov transition field(MTF) and densely connected convolutional networks(DenseNet) is proposed. The vibration signal of the rolling bearing is coded by MTF to generate a two-dimensional image. The time sequence information and the state transition information of the signal are preserved. The two-dimensional image is taken as the input of DenseNet, and the fault characteristics of the rolling bearing vibration signal are extracted through DenseNet. The method enhances the propagation of characteristic information, makes full use of characteristic information, and then realizes fault classification and identification. The data on the Case Western Reserve University bearing dataset is used for the test. The results show that the method can effectively identify the fault types of rolling bearings, and the accuracy of fault diagnosis is 99.5%. In order to further verify the fault diagnosis capability and superiority of this method when the motor load changes, four kinds of network input images, namely, gray-scale image, envelope spectrum image, cepstrum image and MTF generation image, are selected for comparative experiments with the method of combining three networks, namely, Inception, ResNet and DenseNet. The results show that the fault diagnosis accuracy of different methods is higher when the motor load is unchanged than when the motor load is changed. The fault diagnosis accuracy of MTF+DenseNet method is higher than that of other methods. The proposed method still has a high fault diagnosis accuracy when the motor load changes, with an average value of 94.53% and good generalization performance.

Health condition assessment of centrifugal pump based on AHP-TOPSIS comprehensive evaluation method

QIAO Jiawei, TIAN Muqin

2022, 48(9): 69-76. doi: 10.13272/j.issn.1671-251x.17984

<Abstract>(128) <HTML> (29) <PDF>(20)

Abstract:
In view of the problem of centrifugal pump impeller wear, the existing research mostly concentrates on numerical simulation analysis, vibration signal analysis and wear detection of impeller wear. There are few studies on the working condition parameters of centrifugal pump under impeller wear. In order to solve this problem, a health condition assessment method of centrifugal pump based on analytic hierarchy process (AHP) and technique for order preference by similarity to ideal solution (TOPSIS) is proposed. Firstly, according to the operating conditions of centrifugal pump and the performance requirements of centrifugal pump, the operating parameters (flow, head, efficiency and shaft power) characterizing the health condition of centrifugal pump are determined. Secondly, the outer diameter and the specific speed of the impeller are taken as the evaluation indexes of the working condition parameters. The weight values of each working condition parameter to the health condition of the centrifugal pump are determined by using the AHP and the super-transitive approximate method. Finally, the working condition parameters of the centrifugal pump are evaluated by TOPSIS comprehensive evaluation method. The health condition score of the centrifugal pump is obtained by weighted summation, to realize the health condition evaluation of the centrifugal pump. Taking the IS-100-80-125 single-stage centrifugal pump as the research object, the outer diameter of the impeller is worn from 125 mm to 87 mm, and the working parameters of the centrifugal pump with different outer diameters of the impeller are obtained. The working parameters and the health condition of the centrifugal pump are evaluated by the AHP-TOPSIS comprehensive evaluation method. The results show that there is a linear relationship between impeller wear and centrifugal pump health. With the increasing wear of impeller, the drainage capacity of centrifugal pump decreases. The evaluation results are in line with the actual situation of the centrifugal pump, which proves the rationality and feasibility of this method.

Fault diagnosis of mine drainage system based on fuzzy Bayesian network

SHI Xiaojuan, YAO Bing, GU Huabei

2022, 48(9): 77-83. doi: 10.13272/j.issn.1671-251x.18014

<Abstract>(121) <HTML> (26) <PDF>(26)

Abstract:
The mine drainage system is developing towards automation and intelligence. The system's structure and function are becoming more and more complex, and the abnormal function and failure of a single component may cause the failure of the whole system. The existing fault diagnosis methods of the mine drainage system have problems, such as difficult implementation, no consideration of the integrity of the system, and low fault diagnosis efficiency. In order to solve the above problems, a fault diagnosis method of mine drainage system based on fuzzy Bayesian network is proposed. Firstly, the fault tree analysis method is used to decompose the fault causes of the system layer by layer, and find out the root cause of the system fault. Secondly, the events in the fault tree are transformed into the nodes of the Bayesian network. The logic gates are transformed into the directed edges and conditional probabilities of the Bayesian network. The Bayesian network is constructed according to the mapping relationship between the fault tree and the Bayesian network. Thirdly, the fuzzy set theory is introduced into the Bayesian network. The correlation strength between fault and symptom is determined by expert evaluation. After fuzzification by triangular fuzzy number, averaging and defuzzification, the conditional probability of the fuzzy Bayesian network is obtained. Finally, according to the prior probability and conditional probability, the fuzzy Bayesian network is used to judge the probability of each root node fault. The simulation software Genie3.0 is used to establish the fuzzy Bayesian network, and the reasoning analysis and diagnosis test are carried out. The results show that the diagnosis accuracy of the method for each fault symptom is above 80%, and the average accuracy is 82.7%. The method can not only determine the specific position and specific components of the fault source, but also find out the weak nodes of the mine drainage system, eliminate potential faults, and improve the reliability and safety of the system.

Analysis of weak magnetic nondestructive testing for cracks in the top beam of hydraulic support

LIU Ning, XIN Song, HE Min, CHEN Qiuyan

2022, 48(9): 84-91, 133. doi: 10.13272/j.issn.1671-251x.2022060108

<Abstract>(112) <HTML> (17) <PDF>(15)

Abstract:
In view of the complicated operation of the current crack detection method of hydraulic support, a weak magnetic nondestructive detection method for cracks in the top beam of hydraulic support based on geomagnetic field excitation is proposed. Firstly, the stress field of the crack in the top beam of hydraulic support is analyzed by using COMSOL Multiphysics simulation software. Secondly, based on the principle of weak magnetic nondestructive testing, the geomagnetic field is loaded in the simulation space. The superimposed field formed by the magnetized material with crack defects and the geomagnetic field is obtained and analyzed. The results show the following points. ① There is stress concentration phenomenon in the top beam defect of hydraulic support. The lower the stress is, the greater the stress is. The crack expands from the outside to the inside to both sides. The failure risk of the top beam can be effectively reduced by finding the surface crack in time. ② When the length and depth of the crack defect are fixed, the horizontal distance between the valley and the peak and the peak value of the magnetic flux density mode curve of different width cracks increase with the increase of the width. The valley value decreases first and then increases with the increase of the width. The amplitude of the flux density mode increases with the crack width. The variation rate of the difference between the adjacent maximum valley and peak of the flux density mode decreases with the crack width. The change amplitude of magnetic flux density is positively correlated with the change of crack width. ③ When the length and width of the crack defect are fixed, the valley value of the magnetic flux density mode curve of different depth cracks has little difference. The left peak value increases with the increase of the width. The right peak value decreases with the increase of the width. With the increase of the crack depth, the change amplitude of the flux density mode increases. The change rate of the difference between the adjacent maximum valley and peak of the flux density mode is almost constant. The change amplitude of magnetic flux density is positively correlated with the change of crack depth. The magnetic flux density is more sensitive to the change of crack width than to the change of crack depth. ④ When the length, width and depth of the crack defect are fixed, the change in the direction of the crack does not affect the judgment of crack defects.

Segmentation method of the abnormal area of coal infrared thermal image

ZHAO Xiaohu, CHE Tingyu, YE Sheng, TIAN He, ZHANG KAI

2022, 48(9): 92-99. doi: 10.13272/j.issn.1671-251x.2022030086

<Abstract>(204) <HTML> (30) <PDF>(21)

Abstract:
Infrared radiation can reflect the damage of coal and rock under load, and can be used to monitor and prevent the dynamic disaster of coal and rock. But the infrared thermal image generated by the infrared thermal imager has low pixel resolution and large noise, which leads to the detection result being greatly affected by subjective factors. Therefore, the damaged area of the coal body cannot be accurately identified. It has become a trend to combine deep learning with infrared thermal imaging for nondestructive testing. But the research on the identification and detection of coal damage under load by combining deep learning and infrared thermal imaging is relatively few. In order to solve the above problems, a segmentation method of the abnormal area of coal infrared thermal image based on multi-scale channel attention module (MS-CAM) U-Net model is proposed. The MS-CAM is introduced into the encoder of the traditional U-Net model, and the U-Net model structure based on MS-CAM is designed. The model not only pays attention to the major characteristics of the coal infrared thermal image abnormal area, but also pays attention to the small target characteristics of the abnormal area, so as to improve the segmentation accuracy of the abnormal area. In order to reduce the influence of the lack of coal infrared thermal image data set on the accuracy and applicability of the model, the data enhancement operation is carried out on the created coal infrared thermal image data set. The MS-CAM-based U-Net model is pre-trained by using the MS COCO data set. Then the coal infrared thermal image data set is used for training to obtain the final network weight. The experimental result shows that the method can effectively segment the abnormal areas of the infrared thermal image of the coal body. The accuracy rate, the F1 score, the Dice coefficient and the average cross-combination ratio are 94.75%, 94.94%, 94.65%, and 90. 03% respectively. The results are superior to the Deeplab model, the U-Net model and the U-Net model based on the attention mechanism of the SENet.

Optimization of structural parameters of wire rope flaw detector based on orthogonal test

TIAN Jie, SUN Ganggang, LI Ruifeng, WANG Wei

2022, 48(9): 100-108. doi: 10.13272/j.issn.1671-251x.2022050017

<Abstract>(213) <HTML> (33) <PDF>(21)

Abstract:
In the detection of wire rope damage, the structure design of the flaw detector is very important to the detection precision of wire rope damage. The research on the structural parameters and their combinations of the existing electromagnetic wire rope flaw detector is insufficient. In order to solve the above problems, an optimization method of structural parameters of wire rope flaw detector based on orthogonal test is proposed. Based on the theoretical mathematical model of magnetic field distribution of radial magnetic ring and the theoretical model of the equivalent magnetic circuit, the structural parameters affecting the detection precision of wire rope flaw detector are analyzed and obtained. The parameters include the length of the magnet, the thickness of the magnet, the thickness of the armature, the length of the armature and the chamfer parameter. The influence grade and significance of each parameter are studied by the orthogonal test. The influence grade of each parameter factor on the detection precision of the wire rope flaw detector is the thickness of the magnet, the length of the magnet, the length of the armature, the thickness of the armature and the chamfer. The thickness of the magnet, the length of the magnet and the length of the armature have significant effects. These should be given priority when designing the wire rope flaw detector. The thickness of the armature and the chamfer are not significant and can be ignored. The influence of the thickness of the magnet and the length of the magnet is positively correlated with the increase of level. With the increase of the thickness of the magnet and the length (＜70 mm) of the magnet, the detection precision will be significantly improved. The length of the armature shows a negative correlation trend as a whole. The longer the length, the worse the detection precision. According to the analysis results, the optimized values of the parameters of the steel wire rope flaw detector are determined. The magnetic line distribution, the magnetic field distribution and the radial and axial phase magnetic induction intensity distribution of the steel wire rope flaw detector before and after the optimization are compared and verified. The results show that the optimized steel wire rope flaw detector based on the orthogonal test has uniform magnetic flux distribution. The excitation effect of steel wire rope is more than 2 T. The magnetic flux leakage signal is obvious. The damage signal under different phases is quite different. Compared with the steel wire rope flaw detector before optimization, the steel wire rope flaw detector has the following advantages. The magnetic flux leakage intensity is greatly improved. The spatial distribution is obviously improved. The requirements for the position (lift-off value) of the sensor are relatively broad. The radial detection precision is improved by about 40%, and the axial detection precision is improved by about 80%. The perception effect on the damage of the steel wire rope is obviously improved.

The overall design for unmanned transportation system in the mining area

TIAN Chen, DING Zhen, LI Zhenjiang, GAO Yu, AI Yunfeng, CHEN Long

2022, 48(9): 109-115. doi: 10.13272/j.issn.1671-251x.18000

<Abstract>(287) <HTML> (88) <PDF>(44)

Abstract:
At present, the research on unmanned mine trucks mainly focuses on specific control methods such as path tracking and aided decision-making for driving safety. The research is insufficient to support the unmanned operation of the whole mining area transportation system. To solve this problem, an overall design scheme for unmanned transportation system in the mining area is proposed. The system is divided into three layers of cloud control center, edge side and intelligent terminal. The system mainly comprises a mine truck automatic driving system, a mining area unmanned transportation cloud control dispatching platform, an unmanned transportation simulation system, a remote emergency takeover system, a health management system, a collaborative operation management system and a positioning/communication system. The automatic driving system of mine trucks has the function of intelligent management of a single vehicle. It realizes the functions of obstacle perception, intelligent decision-making, path planning, high-precision positioning and precise control. The cloud control dispatching platform is used for unified dispatching management. It realizes collaborative automatic loading and unloading of unmanned mine trucks, electric shovels and bulldozers. It also realizes collaborative mixed operation of unmanned mine trucks and various manned auxiliary operation vehicles. The unmanned transportation simulation system simulates different dispatching operation modes according to the real-time environment of the mine. It forms an optimal schedule scheme to guide actual production. The remote emergency takeover system realizes remote control of the unmanned mine truck in extremely dangerous scenarios. It ensures that the vehicle can be taken over urgently when the automatic driving system of the mine truck encounters a failure. The health management system realizes the functions of equipment self-inspection, fault diagnosis, maintenance result reporting and health state evaluation through the relevant sensors installed at the terminal of the unmanned mine truck. The collaborative operation management system assists the unmanned mine truck to complete the operation together under the condition of ensuring the safety of the driver and the vehicle. It is realized through the software and hardware systems installed on the manned mining equipments, engineering equipments and auxiliary production equipments. The positioning/communication system is the communication channel connecting the cloud control center and the intelligent terminal. It supports multiple communication modes such as 4G/5G/Mesh, and realizes V2N (vehicle-network), V2V (vehicle-vehicle) and V2I (vehicle-infrastructure) communication.

Unmanned vehicle fusion positioning method based on 3D point cloud map and ESKF

CUI Wen, XUE Qiwen, LI Qingling, WANG Fengdong, HAO Xueer

2022, 48(9): 116-122. doi: 10.13272/j.issn.1671-251x.17997

<Abstract>(172) <HTML> (36) <PDF>(30)

Abstract:
The precision of the unmanned vehicle positioning method based on map matching depends on the precision of the created map. The precision is less affected by the outside world. The method is suitable for unmanned vehicle positioning in complex scenarios. However, the common laser radar point cloud matching algorithm takes single characteristic as the core for matching. The matching accuracy rate is low for large-scale point clouds. Therefore, the 3D point cloud map deviates greatly from the actual environment, resulting in the low precision of the unmanned vehicle positioning method based on map matching. To solve the above problems, an unmanned vehicle fusion positioning method based on 3D point cloud map and error state Kalman filter (ESKF) is proposed. The method is composed of two parts: 3D point cloud map construction and ESKF fusion positioning. In the 3D point cloud map construction part, inter-frame point cloud matching is performed through the normal distribution transform (NDT) algorithm to improve the accuracy of large-scale point cloud matching. The closed-loop constraint is added to construct graph optimization problem on the basis of the vertex and the constraint edge of the position and posture graph established by the laser odometer data. The Levenberg-Marquardt (LM) algorithm is used for the graph optimization solution in order to reduce the cumulative drift of the position and posture and improve the 3D point cloud map precision. In the ESKF fusion positioning part, ESKF is used to fuse inertial measurement unit (IMU) data and 3D point cloud map data. The correction of the prior position and posture (position, attitude and velocity) of the unmanned vehicle is realized and the posterior position and posture are output. The experimental results show that, compared with the method based on map matching, the maximum relative position and posture error, average error and root-mean-square error of the proposed method are reduced by 0.176 9 m, 0.027 1 m and 0.059 4 m respectively. The proposed method has better performance in positioning precision and stability.

Development of integrated localization and wireless communication and its application in the underground coal mine

HU Yanjun, ZHAO Yingzhu, YANG Yixue, ZHAI Yushuang, LI Shiyin

2022, 48(9): 123-133. doi: 10.13272/j.issn.1671-251x.17945

<Abstract>(164) <HTML> (26) <PDF>(37)

Abstract:
Integrated localization and wireless communication(ILWC) is a new information technology based on hardware resources and software information sharing to realize the coordination of location and communication functions. This paper briefly introduces the research progress of wireless communication technology and localization technology, and reveals that ILWC technology is the inevitable result of the bearer services expansion of wireless communication system. This study summarizes the definition and connotation of ILWC technology in different research. The core idea of ILWC is defined as "hardware integration and software sharing". The research progress of ILWC technology is summarized according to the two stages of equipment reuse and deep fusion. Combined with the particularity of the scene of the coal mine, the concept of ILWC in the coal mine is put forward. Based on the sharing of time, space, spectrum, computing and other resources, the technology is the fusion technology of communication function and localization function with an automatic scene perception and dynamic and adaptive resource allocation mechanism. The adaptability of underground ILWC in roadway, central substation, underground parking lot, coal working face and other scenes is discussed. It is pointed out that the challenges faced by underground ILWC are the complexity of underground wireless channel, unbalanced deployment of base stations and precise recognition of complex underground scene.

Research on gas concentration prediction driven by ARIMA-SVM combined model

FAN Jingdao, HUANG Yuxin, YAN Zhenguo, LI Chuan, WANG Chunlin, HE Yanpeng

2022, 48(9): 134-139. doi: 10.13272/j.issn.1671-251x.2022030024

<Abstract>(275) <HTML> (36) <PDF>(38)

Abstract:
The single gas prediction model has weak capability in mining all characteristics of the mine gas concentration time sequence. In order to solve the problem, a combined prediction model based on autoregressive intergrated moving average (ARIMA) model and support vector machine (SVM) model is proposed. The model is used to predict gas concentration. Firstly, the prediction results of the two single models are obtained by using the ARIMA model and the SVM model to predict and analyze the experimental data respectively. Secondly, combining the autocorrelation function, partial autocorrelation function and Bayesian criterion, the optimal ARIMA model is obtained as ARIMA(1,1,2). According to the optimization of kernel function and other parameters, the optimal SVM model is established, and then the ARIMA-SVM combined model is established. The ARIMA model is used to process the historical data of the gas concentration time series and obtain the corresponding linear prediction result and the residual sequence. The SVM model is used to further analyze the nonlinear factors in the data residual sequence and obtain the unlinear prediction result. The prediction results of the two models are combined to obtain the final prediction result of the target gas concentration time series. The experimental results show the following results. ① The fitting degree of the prediction results of the ARIMA-SVM combined model is better than that of the ARIMA model and SVM single model. ② Compared with the ARIMA model and SVM model, the error of the ARIMA-SVM combined model is greatly reduced, and the prediction result is obviously better than that of the single model. ③ The mean absolute error, mean absolute percentage error and root mean square error of the ARIMA-SVM combined model are the smallest. This result indicates that the prediction precision of the ARIMA-SVM combined model is higher.

Action recognition method for mine kilometer directional drilling rig

XIANG Xueyi, LEI Zhipeng, LI Linbo, REN Ruibin, LI Jie, WANG Feiyu

2022, 48(9): 140-147, 156. doi: 10.13272/j.issn.1671-251x.2022030103

<Abstract>(222) <HTML> (10) <PDF>(16)

Abstract:
At present, the walking and drilling operations of the mine kilometer directional drilling rig are all realized by the manual operation of drillers. The intelligence level is low. At present, there is no research on the correlation between the action type of kilometer directional drilling rig and the vibration state of the hydraulic pump station. Therefore, it is difficult to remotely identify the action type of the kilometer directional drilling rig. In order to solve the above problems, an action recognition method for mine kilometer directional drilling rig based on empirical wavelet transform (EWT) and fuzzy C-means (FCM) clustering algorithm is proposed. Firstly, the EWT method is used to analyze the frequency characteristic information of the three key parts (motor, hydraulic pump and coupling) of the hydraulic pump station when the kilometer directional drilling rig performs five different actions (the start of the kilometer directional drilling rig, the rotation of the power head without drill pipe, the rotation with drill pipe, the forward slow drilling with drill pipe and the forward fast drilling with drill pipe). The vibration signals in the most obvious direction of each vibration characteristic are selected to form the original signal group for action recognition. Secondly, the combination of EWT decomposition and correlation coefficient selection rules is used to extract eigenvectors containing drill action information in the original signal group for action recognition. The weight of different eigenvectors is confirmed. The standard recognition eigenvector is constructed. Finally, the membership degree between the action eigenvector to be identified and the five action recognition standard eigenvectors is obtained by using the FCM clustering algorithm. The intelligent recognition of the action types of the kilometer directional drilling rig is realized. Taking the ZYL-17000D type mine kilometer directional drilling rig as the research object, the reliability of the action recognition method based on EWT and FCM clustering algorithm for mine kilometer directional drilling rig is verified by experiments. The vibration data of the motor, hydraulic pump and coupling in the axial, horizontal and vertical radial directions under five actions are collected in the experiment. The results show that the empirical wavelet functions of the vibration signals of the motor, hydraulic pump and coupling of the drilling rig show different characteristics when it performs different actions. The clustering performance of the eigenvectors of the axial vibration signals of hydraulic pumps is the best. According to the difference of extracted eigenvectors under different actions, action types can be identified. The results of action recognition based on test data show that this method can effectively identify the action type of kilometer directional drilling rig, and the recognition accuracy is 96.8% when the membership degree is greater than 0.9.

Roadheader combined positioning method based on strapdown inertial navigation and differential odometer

WANG Haoran, WANG Hongwei, LI Zhenglong, FU Xiang

2022, 48(9): 148-156. doi: 10.13272/j.issn.1671-251x.17993

<Abstract>(145) <HTML> (35) <PDF>(30)

Abstract:
Autonomous and accurate positioning of roadheader is the key to realize intelligent and unmanned heading face. The combination of strapdown inertial navigation and odometer is an ideal positioning scheme for roadheader. The positioning error of strapdown inertial navigation accumulates with time and the position and posture perception capability of single odometer is limited. In order to solve the above problems, a roadheader combined positioning method based on strapdown inertial navigation and differential odometer is proposed by introducing differential odometer as auxiliary positioning on the basis of strapdown inertial navigation. The realization of the method consists of three parts: position and posture perception based on strapdown inertial navigation, dead reckoning based on differential odometer, and data fusion based on Kalman filtering. The reference displacement and attitude angle of roadheader in navigation coordinate system are obtained by strapdown inertial navigation. The differential odometer is composed of two odometers installed on the left and right tracks of the roadheader, and the dead reckoning of the roadheader is calculated by the differential odometer. Kalman filter is designed according to the error equation of strapdown inertial navigation and differential odometer. The difference between the position and posture data of roadheader obtained by strapdown inertial navigation and differential odometer is used as the input of Kalman filter. The output value of Kalman filter is used to correct and compensate the strapdown inertial navigation data. The validity judgment of odometer data is integrated into the combined positioning method, so that the influence of track slipping on the positioning precision is avoided. The positioning experiment of roadheader is carried out in the simulated roadway. The results show that the heading angle error measured by the combined positioning method can be controlled within 0.6° and the position error can be controlled within 0.19 m. The method has high positioning precision.

2022 Vol. 48, No. 9

Submission System

Latest AcceptedMore>