A coal mine inspection method and system based on ultrasonic detection for coal mine transfer conveyors.

Abstract

This invention provides a coal mine inspection method and system for coal mine transfer machines based on ultrasonic detection. The invention relates to the field of coal mine inspection technology. The method includes: acquiring multi-source operating data of the coal mine transfer machine at an initial transfer angle; determining multiple suspected wedge-shaped accumulation points at the initial transfer angle based on the multi-source operating data; determining multiple test transfer angles based on the multiple suspected wedge-shaped accumulation points at the initial transfer angle; determining multiple suspected wedge-shaped accumulation points at each test transfer angle based on the multi-source operating data of the coal mine transfer machine at each test transfer angle; and determining the severity of the wedge-shaped accumulation fault in the coal mine based on the multiple suspected wedge-shaped accumulation points at the initial transfer angle and the multiple suspected wedge-shaped accumulation points at each test transfer angle. This method can efficiently and accurately identify suspected wedge-shaped accumulation points on the coal mine transfer machine and determine the severity of the wedge-shaped accumulation fault.

Description

Technical Field

[0001] This invention relates to the field of coal mine inspection technology, and specifically to a coal mine inspection method and system for coal mine transfer machines based on ultrasonic testing. Background Technology

[0002] Material transfer in underground fully mechanized coal mining faces is a core link in ensuring the continuity of coal mining. As a key transfer device connecting scraper conveyors and belt conveyors, the material flow state inside the chute of a coal mine transfer conveyor significantly impacts its lifespan. In actual operating conditions, the physical properties of materials, such as moisture, viscosity, and particle size, often undergo dynamic changes, easily forming wedge-shaped accumulations on one side of the chute. This accumulation alters the material's center of mass and falling trajectory, generating continuous asymmetric impact loads, leading to hidden wear or deformation during equipment operation. Such failures are strongly correlated with the real-time flow state of the material. Traditional coal mine transfer conveyor inspections based on ultrasonic testing mainly rely on manual inspections and static parameter monitoring, judging fault conditions by periodically checking chute wear. However, this method has significant limitations. Manual inspections are constrained by the underground environment, making it difficult to capture the evolution of dynamic accumulation in real time, and relying on experience-based judgment, resulting in a high rate of missed and false detections. Static parameter monitoring only reflects the instantaneous operating state of the equipment and cannot reflect the dynamic response law of wedge-shaped accumulation changing with the transfer angle, making it difficult to accurately assess the severity of the fault. This limitation not only leads to delayed fault warnings and the inability to identify potential accumulation risks in advance, but also affects the scientific nature of equipment maintenance decisions and exacerbates production losses caused by unplanned downtime. Furthermore, traditional methods often rely on single-dimensional data for fault diagnosis, failing to fully integrate multi-source operational information. This makes it difficult to comprehensively characterize the coupling relationship between material accumulation and equipment stress, and to effectively distinguish between slight adhesion and severe accumulation, resulting in insufficient accuracy in fault assessment. Simultaneously, manual inspections and static monitoring rely heavily on manpower, leading to low operational efficiency and an inability to meet the demands of intelligent and unmanned modern coal mining, thus hindering the continuous and stable operation of the fully mechanized mining face.

[0003] Therefore, how to efficiently and accurately identify suspected wedge-shaped accumulation points on coal mine transfer machines and determine the severity of wedge-shaped accumulation faults is an urgent problem to be solved. Summary of the Invention

[0004] The main technical problem this invention addresses is how to efficiently and accurately identify suspected wedge-shaped accumulation points on coal mine transfer machines and determine the severity of the wedge-shaped accumulation fault.

[0005] According to a first aspect, the present invention provides a coal mine inspection method for a coal mine transfer conveyor based on ultrasonic detection, comprising: acquiring multi-source operating data of the coal mine transfer conveyor at an initial transfer angle, wherein the multi-source operating data of the coal mine transfer conveyor at the initial transfer angle includes ultrasonic detection data of the coal mine transfer conveyor; determining multiple suspected wedge-shaped accumulation point information at the initial transfer angle based on the multi-source operating data of the coal mine transfer conveyor at the initial transfer angle; determining multiple test transfer angles based on the multiple suspected wedge-shaped accumulation point information; controlling a transfer conveyor adjustment device to adjust the transfer angles of the transfer conveyor and the belt conveyor based on the multiple test transfer angles, and acquiring multi-source operating data of the coal mine transfer conveyor at each test transfer angle; determining multiple suspected wedge-shaped accumulation point information at each test transfer angle based on the multi-source operating data of the coal mine transfer conveyor at each test transfer angle; and determining the severity of the coal mine wedge-shaped accumulation fault based on the multiple suspected wedge-shaped accumulation point information at the initial transfer angle and the multiple suspected wedge-shaped accumulation point information at each test transfer angle.

[0006] In one possible implementation, determining multiple test relocation angles based on the multiple suspected wedge-shaped accumulation point information includes: clustering the multiple suspected wedge-shaped accumulation point information to obtain K clusters; determining multiple verification test relocation angles for each cluster based on the K clusters; and determining multiple test relocation angles based on the K clusters and the multiple verification test relocation angles for each cluster.

[0007] In one possible implementation, determining the severity of a coal mine wedge-shaped accumulation fault based on multiple suspected wedge-shaped accumulation point information under the initial transfer angle and multiple suspected wedge-shaped accumulation point information under each test transfer angle includes: constructing a wedge-shaped accumulation analysis map, which includes multiple transfer angle nodes and edges between the multiple transfer angle nodes. The multiple transfer angle nodes include an initial transfer angle node and a test transfer angle node. The node features of the initial transfer angle node include multiple suspected wedge-shaped accumulation point information under the initial transfer angle, and the node features of the test transfer angle node include multiple suspected wedge-shaped accumulation point information under the test transfer angle. The edges between the multiple transfer angle nodes are the differences in transfer angles. The severity of the coal mine wedge-shaped accumulation fault is obtained by processing the wedge-shaped accumulation analysis map based on the wedge-shaped accumulation analysis model.

[0008] In one possible implementation, the multi-source operating data of the coal mine transfer machine at the initial transfer angle also includes transfer machine operating video, material humidity information, vibration sensor data, temperature sensor data, and current sensor data.

[0009] According to a second aspect, the present invention provides a coal mine inspection system for a coal mine transfer conveyor based on ultrasonic detection, comprising: a data acquisition module for acquiring multi-source operating data of the coal mine transfer conveyor at an initial transfer angle, wherein the multi-source operating data of the coal mine transfer conveyor at the initial transfer angle includes ultrasonic detection data of the coal mine transfer conveyor; a first suspected point determination module for determining multiple suspected wedge-shaped accumulation point information at the initial transfer angle based on the multi-source operating data of the coal mine transfer conveyor at the initial transfer angle; and a test angle determination module for determining multiple test transfer angles based on the multiple suspected wedge-shaped accumulation point information at the initial transfer angle; The module includes a degree adjustment and data acquisition module, used to control the transfer machine adjustment device to adjust the transfer angle of the transfer machine and the belt conveyor based on the multiple test transfer angles, and to acquire multi-source operating data of the coal mine transfer machine at each test transfer angle; a second suspected point determination module, used to determine multiple suspected wedge-shaped accumulation point information at each test transfer angle based on the multi-source operating data of the coal mine transfer machine at each test transfer angle; and a fault severity determination module, used to determine the severity of the coal mine wedge accumulation fault based on the multiple suspected wedge-shaped accumulation point information at the initial transfer angle and the multiple suspected wedge-shaped accumulation point information at each test transfer angle.

[0010] In one possible implementation, the test angle determination module is further configured to: cluster the multiple suspected wedge-shaped accumulation point information to obtain K clusters; determine multiple verification test relocation angles for each cluster based on the K clusters; and determine multiple test relocation angles based on the K clusters and the multiple verification test relocation angles for each cluster.

[0011] In one possible implementation, the fault severity determination module is further configured to: construct a wedge-shaped accumulation analysis map, the wedge-shaped accumulation analysis map including multiple transfer angle nodes and edges between the multiple transfer angle nodes, the multiple transfer angle nodes including initial transfer angle nodes and test transfer angle nodes, the node features of the initial transfer angle nodes including information on multiple suspected wedge-shaped accumulation points under the initial transfer angle, the node features of the test transfer angle nodes including information on multiple suspected wedge-shaped accumulation points under the test transfer angle, and the edges between the multiple transfer angle nodes being the difference in transfer angles; and process the wedge-shaped accumulation analysis map based on the wedge-shaped accumulation analysis model to obtain the severity of the coal mine wedge-shaped accumulation fault.

[0012] In one possible implementation, the multi-source operating data of the coal mine transfer machine at the initial transfer angle also includes transfer machine operating video, material humidity information, vibration sensor data, temperature sensor data, and current sensor data.

[0013] According to a third aspect, embodiments of the present invention provide an electronic device, including: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method comprising: acquiring multi-source operating data of a coal mine conveyor at an initial transfer angle, the multi-source operating data of the coal mine conveyor at the initial transfer angle including ultrasonic detection data of the coal mine conveyor; determining multiple suspected wedge-shaped accumulation point information at the initial transfer angle based on the multi-source operating data of the coal mine conveyor at the initial transfer angle; determining multiple test transfer angles based on the multiple suspected wedge-shaped accumulation point information at the initial transfer angle; controlling a conveyor adjustment device to adjust the transfer angles of the conveyor and the belt conveyor based on the multiple test transfer angles, and acquiring multi-source operating data of the coal mine conveyor at each test transfer angle; determining multiple suspected wedge-shaped accumulation point information at each test transfer angle based on the multi-source operating data of the coal mine conveyor at each test transfer angle; and determining the severity of the coal mine wedge-shaped accumulation fault based on the multiple suspected wedge-shaped accumulation point information at the initial transfer angle and the multiple suspected wedge-shaped accumulation point information at each test transfer angle.

[0014] According to a fourth aspect, this embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned coal mine detection method for a coal mine transfer conveyor based on ultrasonic detection. The method includes: acquiring multi-source operating data of the coal mine transfer conveyor at an initial transfer angle, the multi-source operating data including ultrasonic detection data of the coal mine transfer conveyor; determining multiple suspected wedge-shaped accumulation points based on the multi-source operating data of the coal mine transfer conveyor at the initial transfer angle; determining multiple test transfer angles based on the multiple suspected wedge-shaped accumulation points; controlling a transfer conveyor adjustment device to adjust the transfer angles of the transfer conveyor and the belt conveyor based on the multiple test transfer angles, and acquiring multi-source operating data of the coal mine transfer conveyor at each test transfer angle; determining multiple suspected wedge-shaped accumulation points at each test transfer angle based on the multi-source operating data of the coal mine transfer conveyor at each test transfer angle; and determining the severity of the coal mine wedge-shaped accumulation fault based on the multiple suspected wedge-shaped accumulation points at the initial transfer angle and the multiple suspected wedge-shaped accumulation points at each test transfer angle.

[0015] This invention provides a coal mine inspection method and system based on ultrasonic detection for coal mine transfer conveyors. The method includes acquiring multi-source operating data of the coal mine transfer conveyor at an initial transfer angle, including ultrasonic detection data of the conveyor; determining multiple suspected wedge-shaped accumulation points based on the multi-source operating data at the initial transfer angle; determining multiple test transfer angles based on the multiple suspected wedge-shaped accumulation points; controlling the transfer conveyor adjustment device to adjust the transfer angle between the transfer conveyor and the belt conveyor based on the multiple test transfer angles, and acquiring multi-source operating data of the coal mine transfer conveyor at each test transfer angle; determining multiple suspected wedge-shaped accumulation points at each test transfer angle based on the multi-source operating data of the coal mine transfer conveyor at each test transfer angle; and determining the severity of the wedge-shaped accumulation fault based on the multiple suspected wedge-shaped accumulation points at the initial transfer angle and the multiple suspected wedge-shaped accumulation points at each test transfer angle. This method can efficiently and accurately identify suspected wedge-shaped accumulation points on coal mine transfer conveyors and determine the severity of the wedge-shaped accumulation fault. Attached Figure Description

[0016] Figure 1 A schematic flowchart of a coal mine inspection method for a coal mine transfer machine based on ultrasonic detection, provided in an embodiment of the present invention;

[0017] Figure 2 A schematic diagram of a coal mine transfer machine provided in an embodiment of the present invention;

[0018] Figure 3 This is a schematic diagram illustrating the angle relationship between a coal mine transfer conveyor and a belt conveyor, provided in an embodiment of the present invention.

[0019] Figure 4 A flowchart illustrating the process of determining multiple test retracing angles provided in an embodiment of the present invention;

[0020] Figure 5 A schematic flowchart illustrating the process of determining the severity of a wedge-shaped accumulation fault in a coal mine, provided as an embodiment of the present invention;

[0021] Figure 6 This is a schematic diagram of a coal mine detection system for a coal mine transfer machine based on ultrasonic detection, provided as an embodiment of the present invention. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0023] In this embodiment of the invention, the following are provided: Figure 1 The diagram illustrates a coal mine inspection method for a coal mine transfer machine based on ultrasonic detection. The method includes steps S1 to S6:

[0024] Step S1: Obtain multi-source operating data of the coal mine transfer machine at the initial transfer angle. The multi-source operating data of the coal mine transfer machine at the initial transfer angle includes ultrasonic detection data of the coal mine transfer machine.

[0025] A coal mine transfer conveyor is an intermediate conveying device in a fully mechanized coal mining face used to receive coal transported by a scraper conveyor and transfer it to a belt conveyor. Figure 2 This is a schematic diagram of a coal mine transfer machine provided in an embodiment of the present invention.

[0026] A coal mine transfer conveyor mainly consists of a head unit, chute, scraper chain, drive unit, and adjustment unit. The coal mine transfer conveyor can adjust the material conveying height and angle to ensure continuous and stable coal transfer.

[0027] The transfer angle refers to the relative inclination angle formed between the centerline of the discharge chute of the coal mine transfer machine and the conveying plane of its downstream belt conveyor during the coal mine production and transportation process. Figure 3 This is a schematic diagram illustrating the angle relationship between a coal mine transfer conveyor and a belt conveyor, provided as an embodiment of the present invention.

[0028] The transfer angle determines the trajectory and descent speed of material as it enters the conveyor belt from the transfer chute. A smaller transfer angle results in a gentler discharge direction from the chute, less material impact, and higher transport stability, reducing coal spillage and belt misalignment. However, the material flow is slower, and the gravitational force is insufficient, making it prone to adhesion and asymmetrical accumulation on the chute wall. A larger transfer angle results in a steeper discharge from the chute, smoother material flow, and less stagnation, reducing the risk of adhesion. However, it increases the material throwing distance and descent impact, potentially causing belt damage and increased equipment load.

[0029] Multi-source operational data of a coal mine transfer conveyor at its initial transfer angle refers to the set of information reflecting the equipment's operating status, collected in real time by sensor arrays deployed at key parts of the equipment when the conveyor is at its currently set initial transfer angle. This multi-source operational data includes ultrasonic detection data, conveyor operation video, material moisture information, vibration sensor data, temperature sensor data, and current sensor data.

[0030] The ultrasonic testing data of a coal mine transfer conveyor is a set of real-time echo parameters collected by multiple sets of ultrasonic probes deployed on the side walls, bottom plate, and around the discharge port of the transfer conveyor, which directionally emit and receive high-frequency ultrasonic signals. The ultrasonic testing data of a coal mine transfer conveyor includes core parameters such as ultrasonic wave propagation time, echo intensity, signal attenuation coefficient, reflection angle, and obstacle echo profile.

[0031] In some embodiments, ultrasonic detection probes can be evenly distributed at multiple points along the chute conveying path. Then, the ultrasonic signal penetrates the air medium to complete the scanning detection. When there are situations such as coal wedge accumulation, local material adhesion, or abnormal increase in material thickness on the inner wall of the chute, the accumulation will change the reflection path and loss degree of the ultrasonic waves, thereby causing the echo signal to be significantly attenuated, the reflection peak to shift, and the propagation delay to increase.

[0032] The ultrasonic detection data of the coal mine transfer machine can accurately capture the material accumulation shape, accumulation thickness, accumulation range and spatial distribution characteristics in the hidden locations inside the chute, and can realize the fine detection of wedge-shaped accumulation hazards in hidden areas such as dead corners and side wall gaps in the chute.

[0033] The transfer machine operation video is a continuous video data collected in real time by a camera device positioned above the transfer machine chute in the coal mine. It is used to record the flow status, trajectory distribution, accumulation outline, and adhesion of coal materials inside the chute to the wall.

[0034] The video of the transfer machine in operation clearly shows the material's left and right displacement, local bulges, and single-sided adhesion contours within the chute.

[0035] Material moisture information is a quantitative value of the moisture content of coal materials transported in the chute of a coal mine transfer machine, collected by sensors.

[0036] Material moisture information can be used to characterize the moisture level and adhesion tendency of materials. The higher the moisture content, the easier it is for the material to adhere to the chute wall and form a buildup.

[0037] Vibration sensor data is collected by vibration sensors fixed to the side wall of the transfer conveyor chute and the head frame, which collect data on vibration acceleration, vibration amplitude, and vibration frequency.

[0038] Vibration sensor data can be used to reflect the impact state and load symmetry of materials on the chute. If unilateral accumulation occurs, the values ​​in the vibration sensor data will show significantly higher values ​​or periodic abnormal fluctuations.

[0039] Temperature sensor data is obtained by non-contact collection of real-time temperature values ​​of the chute wall and material surface through infrared temperature sensors arranged above the coal mine transfer conveyor chute.

[0040] Temperature sensor data can reflect heat generated by wall friction, material accumulation, or changes in ambient temperature. Areas with long-term localized accumulation or material stagnation may exhibit temperatures slightly higher than normal continuous conveying areas; temperature sensor data can be used to help identify suspected accumulation locations.

[0041] The current sensor data is obtained by collecting the operating current of the drive motor of the coal mine transfer conveyor, resulting in a current fluctuation curve. This data can characterize the magnitude of the drive load and indirectly reflect material accumulation, jamming, or impeded flow.

[0042] Multi-source operational data from coal mine transfer conveyors at the initial transfer angle can directly reflect the material flow state and asymmetric distribution characteristics within the chute. For example, the material flow trajectory can be observed through transfer conveyor operation videos; current sensor data and vibration sensor data can indirectly reflect the impact load of the material on the equipment; and temperature sensor data and material humidity information can characterize material adhesion characteristics and changes in environmental conditions. This multi-source operational data from coal mine transfer conveyors at the initial transfer angle can be used for subsequent identification and trend assessment of wedge-shaped accumulation hazards.

[0043] Step S2: Based on the multi-source operating data of the coal mine transfer machine at the initial transfer angle, determine the information of multiple suspected wedge-shaped accumulation points at the initial transfer angle.

[0044] In some embodiments, a first suspected point determination model can be used to determine information on multiple suspected wedge-shaped accumulation points at the initial transfer angle. The first suspected point determination model is a Transformer model. The input to the first suspected point determination model is multi-source operating data of the coal mine transfer machine at the initial transfer angle, and the output of the first suspected point determination model is information on multiple suspected wedge-shaped accumulation points at the initial transfer angle.

[0045] The Transformer model is a deep learning architecture based on a self-attention mechanism. Through multi-head self-attention, the Transformer model projects input data into different dimensional spaces, thereby learning the complex nonlinear correlation features within the data. The Transformer model mainly consists of an encoder and a decoder. The encoder preserves the temporal information of the data through positional encoding and uses layer normalization and feedforward neural networks to abstract the features layer by layer. When processing data from coal mine transfer conveyors, the Transformer model can accurately capture subtle state shifts caused by changes in the physical properties of materials and establish a mapping relationship between the material flow trajectory and equipment wear and deformation.

[0046] The information on multiple suspected wedge-shaped accumulation points under the initial transfer angle was determined through analysis of the first suspected point determination model. This information identifies the abnormal locations and characteristics of suspected asymmetric material accumulation within the transfer machine chute. Each suspected wedge-shaped accumulation point under the initial transfer angle includes its three-dimensional spatial coordinates, the estimated thickness of the suspected accumulated material, the intensity of the asymmetric impact load at the suspected accumulation point, and the probability that the suspected accumulation point represents a wedge-shaped accumulation fault.

[0047] The asymmetric impact load intensity at a suspected accumulation point refers to the magnitude of the uneven impact force generated on the chute wall where the suspected accumulation point is located when the material is accumulated in a wedge shape on one side of the chute.

[0048] Information on multiple suspected wedge-shaped accumulation points at the initial transfer angle can reflect the abnormal adhesion state of the material at the current transfer angle due to changes in humidity or viscosity, such as the preliminary accumulation profile information formed on the baffle on one side of the chute.

[0049] Multi-source operational data from the initial transfer angle of the coal mine transfer machine recorded the dynamic physical feedback of the machine under actual load. Abnormal pulsations from the current sensor and spectral shifts from the vibration sensor contained the underlying characteristics of material centroid displacement. Temperature sensor data and material humidity information reflected material adhesion trends and environmental conditions, while the transfer machine's operational video provided intuitive visual criteria for material trajectory changes. This multi-source operational data provides the model with fundamental information to mine wedge-shaped accumulation fault characteristics from multiple dimensions, including equipment operating status, environmental material characteristics, and visual features. This enables the model to identify suspected wedge-shaped accumulation points that are difficult to detect through static inspection.

[0050] The multi-source operational data of coal mine transfer machines from the initial transfer perspective can fully reconstruct the actual flow state, wall adhesion, and asymmetric distribution patterns of materials throughout the entire chute from multiple dimensions, including ultrasonic full-domain profile detection, visual imaging, equipment dynamic operation, material physicochemical properties, and environmental conditions. Once a wedge-shaped accumulation fault develops, it will simultaneously trigger ultrasonic echo parameter distortion, visual material trajectory deviation, and equipment sensor parameter anomalies, thus forming multi-dimensional, quantifiable, and identifiable linked abnormal characteristics.

[0051] Ultrasonic detection data is less susceptible to dust obstruction and can compensate for the limitations of visual monitoring. It accurately reflects the thickness, density, and initial wedge-shaped formation of material accumulation in hidden corners of the chute, making it core preliminary data for identifying hidden accumulation hazards. Ultrasonic detection data can complement visual video data to cover the entire detection area of ​​the chute, effectively improving the completeness of comprehensive detection through multi-source data analysis.

[0052] The transfer machine operation video can intuitively provide visual information such as material flow trajectory, offset status, and local accumulation outline. The transfer machine operation video is the direct basis for model identification of suspected wedge-shaped accumulation points.

[0053] Current sensor data can reflect changes in the drive load. For example, when asymmetrical accumulation occurs in the chute, the motor load will fluctuate abnormally. Current sensor data can indirectly characterize the material's centroid shift and accumulation trend.

[0054] Vibration sensor data can reveal the asymmetric impact characteristics of the chute walls and can be used to locate suspected accumulation sites. For example, unilateral accumulation of material can cause spectral shifts and vibration anomalies.

[0055] Temperature sensor data and material humidity information can characterize the adhesion properties of materials. In particular, areas with high humidity and abnormal temperature are more prone to material retention and adhesion accumulation.

[0056] The Transformer model, through its internal multi-head self-attention mechanism, can perform in-depth analysis of multi-source operational data of coal mine transfer machines at the initial transfer angle. The Transformer model first normalizes and maps heterogeneous data, including ultrasonic detection data, current sensor data, vibration sensor data, temperature sensor data, material humidity information, and transfer machine operation videos, into a unified high-dimensional vector space. During the encoding phase, the Transformer model uses a self-attention matrix to calculate the correlation between data within different time windows. For example, when the ultrasonic echo intensity continuously decreases, the local ranging thickness increases synchronously, and the model detects periodic small oscillations in the motor current accompanied by asymmetric vibration signals, the self-attention mechanism simultaneously strengthens the ultrasonic and electromechanical anomalies, assigning higher weights to the combined features. Multiple attention heads within the model can capture anomaly patterns at different scales in parallel. One set of attention heads focuses on analyzing dynamic anomaly components in vibration and current signals, while another set focuses on extracting material accumulation, trajectory deviation, and adhesion features from the transfer machine operation video. Through stacked Transformer blocks, the model can fuse multi-dimensional features and identify wedge-shaped accumulation signals hidden within normal fluctuations. In the output stage, the Transformer model, based on the learned material dynamic distribution pattern and fault correlation law, can traverse and evaluate the spatial grid in the chute, determine whether there is material adhesion and centroid shift trend in each grid cell, and calculate the anomaly confidence of each point. Finally, the model can identify multiple suspected wedge-shaped accumulation points under the initial transfer angle.

[0057] Step S3: Determine multiple test relocation angles based on the information of multiple suspected wedge-shaped accumulation points under the initial relocation angle.

[0058] In some embodiments, Figure 4 This is a flowchart illustrating the process of determining multiple test re-entry angles according to an embodiment of the present invention. The determination of multiple test re-entry angles includes steps S31 to S33:

[0059] Step S31: Cluster the multiple suspected wedge-shaped accumulation points to obtain K clusters.

[0060] The clustering method described is K-means clustering, an unsupervised learning method based on distance metrics designed to divide a dataset into a predetermined number of categories. Through an iterative optimization process, K-means clustering minimizes the sum of squared distances between each data point and the center of its cluster.

[0061] The K clusters are several datasets formed by grouping multiple suspected wedge-shaped accumulation points at the initial relocation angle using the K-means clustering algorithm based on similarities in features such as spatial distribution, load intensity, and failure probability. Each of the K clusters represents a class of suspected accumulation regions with similar physical characteristics; for example, suspected points that are geographically close and have similar impact load indices are grouped into one cluster.

[0062] In some embodiments, the value of K can be determined by a preset relationship table between the value of K and the spatial distribution density and load characteristic difference of the initial suspected wedge-shaped accumulation points. The higher the spatial distribution density and the greater the load characteristic difference of the initial suspected wedge-shaped accumulation points, the larger the value of K. The preset relationship table between the value of K and the spatial distribution density and load characteristic difference of the initial suspected wedge-shaped accumulation points is artificially constructed in advance.

[0063] K clusters can be used to classify, summarize, and regionalize a large amount of suspected point information.

[0064] The information on multiple suspected wedge-shaped accumulation points at the initial transfer angle contains a large number of scattered spatial points with different attributes, representing potential fault risks within the transfer machine. Since the number of suspected points may be large and their distribution uneven, directly adjusting the angle to verify each point would result in excessive computational overhead and difficulty in capturing regional evolutionary patterns. By using this suspected point information as clustering input, the spatial correlation and physical attribute similarity between points can be utilized to transform isolated suspected signals into statistically significant groups, enabling a more accurate description of the macroscopic distribution pattern of wedge-shaped accumulation within the transfer machine.

[0065] The process of clustering multiple suspected wedge-shaped accumulation points using the K-means clustering algorithm is as follows: First, the information of multiple suspected wedge-shaped accumulation points at the initial relocation angle is mapped to a multi-dimensional feature space. The coordinate axes of the dimension consist of parameters such as position coordinates, estimated thickness, and impact intensity. The algorithm randomly selects K points as initial cluster centers and calculates the Euclidean distance from each suspected wedge-shaped accumulation point to these centers. According to the principle of minimum distance, each suspected point is assigned to the center closest to its physical characteristics, thus forming a preliminary cluster partition. Subsequently, the average coordinates of all points within each cluster in the multi-dimensional space are calculated, and the position of the cluster center is updated using this average value. This iterative process continues until the position change of the cluster center is lower than a set threshold, or a preset number of iterations is reached, to ensure that the homogeneity of members within the cluster is maximized.

[0066] Step S32: Determine multiple verification test retransmission angles for each of the K clusters.

[0067] In some embodiments, a verification angle determination model can be used to determine multiple verification test recursion angles for each cluster. The verification angle determination model is a deep neural network model. The input to the verification angle determination model is the K clusters, and the output of the verification angle determination model is multiple verification test recursion angles for each cluster.

[0068] Deep neural network models include deep neural networks (DNNs), which are non-linear information processing architectures composed of multiple hidden layers. Deep neural networks can fit complex patterns by mimicking the connection methods of biological neurons. They can adjust network weights using the backpropagation algorithm and extract deep patterns from abstract features. A deep neural network consists of an input layer, multiple fully connected layers, activation function layers, and an output layer.

[0069] The multiple verification test relocation angles for each cluster are determined by the verification angle determination model. For the suspected accumulation area corresponding to a single cluster, a set of relocation angles are used to make the suspected wedge-shaped accumulation points that were not visible under the initial relocation angle of the cluster visible by adjusting the angle.

[0070] By determining multiple verification test relocation angles for each cluster, multiple sets of comparative observation conditions under different operating conditions can be constructed for the suspected accumulation area corresponding to each cluster. By setting gradient and differentiated relocation angles, suspected wedge-shaped accumulation points that were not apparent under the initial relocation angle can be made visible within the same cluster area, thereby enhancing the identifiability of suspected wedge-shaped accumulation anomaly signals.

[0071] K clusters categorize scattered suspected wedge-shaped accumulation points by feature classification. Points within each cluster have similar spatial locations and physical characteristics, thus fully characterizing the accumulation risk attributes of the corresponding area. The feature data of a single cluster can centrally reflect the material accumulation characteristics of the corresponding area, providing a precise regional feature basis for verifying the setting of the transfer angle.

[0072] Deep neural networks, through their multi-layered nonlinear transformation structure, can perform in-depth analysis of the physical parameters contained in K clusters. The first layer of neurons in the deep neural network, by analyzing the coordinates of the spatial centroid, can accurately identify the sensitivity of suspected accumulation areas to changes in the transfer angle. For example, it can determine whether the accumulation area is located on the flank of a chute, which is easily detached due to gravity, or at the bottom of a chute where friction is greater. The hidden layers in the middle utilize nonlinear activation functions to establish a complex dynamic simulation environment, simulating the dynamic changes in the material's own gravity component and the antagonistic balance between material adhesion forces under different transfer angle adjustments. During this process, the model can combine the total impact energy density in the K clusters to calculate the critical angle point that can break the current adhesion balance and cause a significant deviation in the material's flow trajectory. Subsequently, the model generates a set of gradient-based verification test transfer angles for each of the K clusters in the output layer based on these calculated dynamic critical data. Using this set of gradient angles, suspected wedge-shaped accumulation points not apparent at the initial transfer angle within the cluster area can be effectively revealed, and the dynamic feedback differences of suspected accumulation areas at different angles can be maximized. For example, for clusters located above the chute side and with high failure probability distribution scores, the model generates a set of verification test transfer angles with a large span to observe whether material falls off or the trajectory reconstructs when the slope changes significantly. For clusters located at the bottom with stable loads, the model generates a set of fine-tuned verification test transfer angles to detect data disturbances caused by extremely subtle trajectory offsets. Ultimately, through this targeted calculation, the deep neural network can determine the precise verification test transfer angle for each cluster.

[0073] In some embodiments, determining multiple verification test retransmission angles for each of the K clusters includes steps S321-S323:

[0074] Step S321: Based on the K clusters, determine the estimated value of the adhesion force of the deposit corresponding to each cluster and the surface friction coefficient score of the region where each cluster is located.

[0075] In some embodiments, deep neural networks can be used to determine the estimated adhesion force of the deposits for each cluster and the surface friction coefficient score of the region where each cluster is located.

[0076] The estimated adhesion force of each cluster refers to the estimated force of adhesion between the material and the chute wall within each suspected cluster area.

[0077] The estimated adhesion force of a material deposit can be used to characterize how easily a material adheres to a wall surface. The higher the estimated adhesion force, the less likely the material is to fall off.

[0078] The surface friction coefficient score for each cluster's region refers to a quantitative score of the roughness of the chute wall corresponding to each cluster.

[0079] The surface friction coefficient score can be used to characterize the frictional resistance encountered by a material when it slides in that area. The higher the surface friction coefficient score, the more difficult it is for the material to slide and the easier it is to accumulate.

[0080] Deep neural networks possess the ability to perform nonlinear fitting, correlation mining, and trend inference on multidimensional physical data. They can uncover hidden physical correlations from information such as the three-dimensional spatial coordinates of suspected accumulation points, the estimated thickness of suspected accumulated material, the intensity of asymmetric impact loads, and the probability of wedge-shaped accumulation failure within K clusters. By jointly analyzing the spatial distribution of suspected accumulation points, material thickness, and impact loads, deep neural networks can establish a nonlinear mapping relationship between material accumulation morphology and adhesion forces, thereby inferring and outputting the estimated adhesion force of the accumulated material for each cluster. Furthermore, by combining the spatial location and load distribution characteristics of clusters, deep neural networks can simultaneously learn the intrinsic correlation between the wall state and frictional resistance in different regions of the chute, thus quantifying the surface friction coefficient score for each cluster's region.

[0081] Step S322: Based on the estimated adhesion force of the deposits corresponding to each cluster and the surface friction coefficient score of the region where each cluster is located, determine the wedge stacking risk score of each cluster, the wedge stacking transfer angle trigger threshold of each cluster, and the angle sensitivity coefficient of each cluster.

[0082] In some embodiments, a deep neural network can be used to determine the wedge stacking risk score for each cluster, the wedge stacking relocation angle trigger threshold for each cluster, and the angle sensitivity coefficient for each cluster.

[0083] The wedge stacking risk score for each cluster refers to the quantitative score of the probability of wedge stacking failure in each cluster, determined by a deep neural network. The higher the score, the more likely the region is to form wedge stacking.

[0084] The wedge-shaped stacking transfer angle trigger threshold for each cluster refers to the critical transfer angle value that can break the material balance in the cluster area, cause the stacking to appear, or cause it to begin to fall off.

[0085] The angle sensitivity coefficient for each cluster is a quantification of how sensitive the packing state of that cluster region changes with the transfer angle. A higher angle sensitivity coefficient indicates that even a slight change in angle will significantly affect the packing state.

[0086] Deep neural networks can weight and fuse the predicted adhesion force and surface friction coefficient scores of the material accumulation, quantify the risk, and establish a mapping relationship between material retention tendency and failure probability, thereby outputting a wedge-shaped accumulation risk score for each cluster. By simulating the dynamic balance between the gravitational component of the material and the adhesion and friction forces at different transfer angles, the model can fit the inflection point value of the material transitioning from a stable state to a flowing or detaching state, thus accurately calculating the triggering critical value of the wedge-shaped accumulation transfer angle for each cluster. Furthermore, deep neural networks can uncover the response amplitude patterns of adhesion force and friction coefficient with angle changes and quantify the degree of influence of angle changes on the accumulation state, thereby determining the angle sensitivity coefficient of each cluster.

[0087] Step S323: Based on the wedge stacking risk score of each cluster, the wedge stacking relocation angle trigger threshold of each cluster, and the angle sensitivity coefficient of each cluster, determine multiple verification test relocation angles for each cluster.

[0088] In some embodiments, a deep neural network can be used to determine multiple verification test recursion angles for each cluster.

[0089] Deep neural networks possess powerful capabilities in multi-dimensional decision-making, gradient sequence generation, and constraint optimization. Based on the wedge-shaped stacking risk score, wedge-shaped stacking relocation angle trigger threshold, and angle sensitivity coefficient for each cluster, they learn intelligent decision-making patterns related to fault risk level, angle critical conditions, and angle change amplitude. This allows for the generation of multiple scientifically sound and suitable verification test relocation angles for each cluster. The deep neural network can prioritize the detection of each cluster based on the wedge-shaped stacking risk score, using a wider angle range for high-risk clusters and a finer angle range for low-risk clusters to ensure the angle settings match the risk level. The model can expand the gradient around the wedge-shaped stacking relocation angle trigger threshold and extend it outwards from the threshold to ensure the generated angles effectively trigger changes in the stacking state, thus fully revealing hidden suspected points. Furthermore, the model can adaptively adjust the angle interval based on the angle sensitivity coefficient, using smaller intervals for highly sensitive clusters and larger intervals for low-sensitive clusters, thereby improving overall efficiency while maintaining detection accuracy. Finally, the model can output multiple verification test relocation angles for each cluster that meet both physical rationality and engineering practicality requirements.

[0090] Step S33: Based on the K clusters and the multiple verification test relocation angles of each cluster, determine multiple test relocation angles.

[0091] In some embodiments, a test angle determination model can be used to determine multiple test recursion angles. The test angle determination model is a deep neural network model. The input to the test angle determination model is the K clusters and multiple verification test recursion angles for each cluster, and the output of the test angle determination model is the multiple test recursion angles.

[0092] Multiple test transfer angles are determined by the test angle determination model. They are used to cover each cluster of suspected accumulation areas by adjusting the angles, so as to effectively detect suspected wedge-shaped accumulation points that are not visible at the initial transfer angle throughout the entire range of the coal mine transfer machine.

[0093] Multiple test transfer angles can take into account both the overall working condition adaptability of the coal mine transfer machine and the coverage of suspected areas in each cluster. It can effectively detect suspected wedge-shaped accumulation points that are not apparent under the initial transfer angle in each cluster area, while also being compatible with the differences in accumulation characteristics of different cluster areas.

[0094] K clusters can represent the spatial distribution characteristics and physical properties of different suspected accumulation regions. Multiple verification test reloading angles for each cluster are used to provide the angle variation range of the corresponding region. By using K clusters and multiple verification test reloading angles for each cluster, the model can comprehensively analyze the sensitivity of each cluster to angle changes, and thus determine the test reloading angles applicable to the whole machine.

[0095] Deep neural networks, by constructing a global optimization function, can collaboratively analyze the physical characteristics of K input clusters and the corresponding verification test relocation angles for each cluster. The model uses hidden layers to calculate the degree of mutual influence and response differences among different clusters at the same relocation angle; for example, it determines whether a certain angle, while triggering the stacking characteristics of the first cluster, will enhance the adhesion characteristics of the second cluster. The model can utilize a weighting mechanism to weight the failure probabilities of each cluster, prioritizing the testing needs of high-probability failure areas. Through this cross-regional correlation analysis, deep neural networks can identify representative test relocation angles within a complex parameter space and extract multiple test relocation angles from the verification combinations that effectively expose suspected wedge-shaped stacking points in each region while maintaining the overall continuous operation of the equipment, ensuring that each selected angle fully reflects the evolutionary characteristics of wedge-shaped stacking.

[0096] Step S4: Based on the multiple test transfer angles, control the transfer machine adjustment device to adjust the transfer angle between the transfer machine and the belt conveyor, and obtain multi-source operating data of the coal mine transfer machine under each test transfer angle.

[0097] Multi-source operating data of a coal mine transfer machine at each test transfer angle refers to the set of digital information reflecting the operating status of the equipment, which is collected in real time by a sensor array deployed at key parts of the equipment when the coal mine transfer machine is in the corresponding test transfer angle state.

[0098] Multi-source operational data for the coal mine transfer machine at each test transfer angle includes ultrasonic testing data, transfer machine operation video, material moisture information, vibration sensor data, temperature sensor data, and current sensor data.

[0099] Step S5: Based on the multi-source operating data of the coal mine transfer machine at each test transfer angle, determine the information of multiple suspected wedge-shaped accumulation points at each test transfer angle.

[0100] In some embodiments, a second suspected point determination model can be used to determine multiple suspected wedge-shaped accumulation points at each test transfer angle. The second suspected point determination model is a Transformer model. The input to the second suspected point determination model is multi-source operating data of the coal mine transfer machine at each test transfer angle, and the output of the second suspected point determination model is multiple suspected wedge-shaped accumulation points at each test transfer angle.

[0101] Information on multiple suspected wedge-shaped accumulation points under each test transfer angle was determined through analysis using a second suspected point determination model. This model identifies the abnormal locations and characteristics of suspected asymmetric material accumulation within the transfer machine chute under each corresponding test transfer angle. Information on each suspected wedge-shaped accumulation point under each test transfer angle includes the three-dimensional spatial coordinates of the suspected accumulation point, the estimated thickness of the suspected accumulated material, the intensity of the asymmetric impact load at the suspected accumulation point, and the probability that the suspected accumulation point represents a wedge-shaped accumulation fault.

[0102] The Transformer model, through a multi-head self-attention mechanism, analyzes multi-source operating data of coal mine transfer machines at each test transfer angle. First, it normalizes and maps heterogeneous data such as ultrasonic detection data, current sensor data, vibration sensor data, temperature sensor data, material humidity information, and transfer machine operating videos to a unified feature space. During the encoding stage, the self-attention mechanism captures the correlation features between time-series data and identifies abnormal behaviors in the operating data. Multiple attention heads extract fault features at different scales, and after multi-layer feature fusion, hidden wedge-shaped accumulation anomalies are located. In the output stage, the Transformer model, based on the correlation between material flow and equipment response, traverses and evaluates the adhesion and accumulation trends at various points within the chute, ultimately identifying multiple suspected wedge-shaped accumulation points at each test transfer angle.

[0103] Step S6: Determine the severity of the coal mine wedge accumulation fault based on the information of multiple suspected wedge accumulation points under the initial transfer angle and the information of multiple suspected wedge accumulation points under each test transfer angle.

[0104] In some embodiments, Figure 5 This invention provides a flowchart illustrating the determination of the severity of a wedge-shaped coal mine accumulation fault, comprising steps S61-S62:

[0105] Step S61: Construct a wedge-shaped stacking analysis map. The wedge-shaped stacking analysis map includes multiple relocation angle nodes and edges between the multiple relocation angle nodes. The multiple relocation angle nodes include initial relocation angle nodes and test relocation angle nodes. The node features of the initial relocation angle nodes include information on multiple suspected wedge-shaped stacking points under the initial relocation angle. The node features of the test relocation angle nodes include information on multiple suspected wedge-shaped stacking points under the test relocation angle. The edges between the multiple relocation angle nodes are the differences in relocation angles.

[0106] The wedge-shaped stacking analysis map is a data structure used to characterize the fault features and their correlations under different relocation angles. The wedge-shaped stacking analysis map includes multiple relocation angle nodes and the edges between these nodes.

[0107] The identification and severity assessment of wedge-shaped pile-up faults rely on the changing patterns of suspected pile-up point information under different transfer angles, and there are complex correlations among the fault characteristics at these different angles. Wedge-shaped pile-up analysis maps can effectively represent these correlations and reflect the changing patterns of fault characteristics under different transfer angles.

[0108] Step S62: Based on the wedge-shaped accumulation analysis model, the wedge-shaped accumulation analysis map is processed to obtain the severity of the wedge-shaped accumulation fault in the coal mine.

[0109] In some embodiments, the wedge-shaped accumulation analysis model is a graph neural network model. The input of the wedge-shaped accumulation analysis model is the wedge-shaped accumulation analysis map, and the output of the wedge-shaped accumulation analysis model is the severity of the wedge-shaped accumulation fault in the coal mine.

[0110] Graph Neural Network (GNN) models consist of Graph Neural Networks (GNNs) and fully connected layers. GNNs are deep learning models that learn features from graph data. Through message passing mechanisms, GNNs can aggregate and update features of nodes and edges. GNNs can effectively capture the relationships between nodes in a graph structure and uncover hidden patterns in node and edge features. GNNs exhibit excellent performance in tasks such as classification, regression, and anomaly assessment of graph data.

[0111] The severity of wedge-shaped accumulation failure in coal mines is an indicator of the risk level of wedge-shaped accumulation failure of coal and mineral materials in the chute of a coal mine transfer machine, determined by a wedge-shaped accumulation analysis model.

[0112] The severity of wedge-shaped accumulation faults in coal mines can be differentiated by numerical classification, which distinguishes the scale of material accumulation in the chute, the intensity of impact load on the equipment, and the risk level of equipment wear and deformation.

[0113] Wedge-shaped accumulation analysis maps can characterize the spatial state of material accumulation at different transfer angles and intuitively present the correlation between the features of each transfer angle node and its suspected wedge-shaped accumulation points. This structured representation facilitates graph neural networks' understanding and processing of complex wedge-shaped accumulation evolution relationships. Through node features and edge information, graph neural networks can better capture the load similarity and state differences between suspected accumulation points at different angles. For example, the asymmetric impact load intensity and failure probability in the node features provide the true stress and accumulation state of the chute wall at that angle, while the angle difference on the edges reflects the continuity and evolution trend of these states as the transfer angle changes.

[0114] The construction of a wedge-shaped stacking analysis map effectively organizes scattered relocation angle data and their fault correlations, facilitating feature aggregation calculations by graph neural networks. This organization reduces feature redundancy and improves the model's inference efficiency and discrimination accuracy when assessing fault severity. Because the wedge-shaped stacking analysis map provides structured node and edge information, graph neural networks can better handle multi-dimensional operating parameters, thus avoiding the state fragmentation and spatial correlation loss problems caused by independent angle data in traditional methods. This structural advantage makes graph neural networks more advantageous in handling complex working condition evolution relationships such as dynamic wedge stacking.

[0115] By comprehensively analyzing the stacking characteristics and angular relationships of nodes in the wedge stacking analysis graph, the graph neural network can extract comprehensive features that fully reflect the stacking scale, load intensity and equipment wear risk in the state space formed by multiple test angles, and finally complete the accurate assessment of the severity of wedge stacking failure in coal mines.

[0116] Graph neural networks (Graph Neural Networks) can perform message passing operations on wedge-shaped stacking analysis maps through graph convolutional layers. During each convolutional pass, each relocation angle node collects information from its neighbors and updates its features by combining the relocation angle differences of connecting edges. Specifically, the Graph Neural Network can calculate the drift distance and energy attenuation ratio of suspected wedge-shaped stacking points at adjacent angles. If, as the relocation angle flattens, the suspected point shows a significant reduction in area or load, the model will confirm that the path conforms to the physical characteristics of wedge-shaped stacking through a weight learning mechanism. Utilizing its deep structure, the Graph Neural Network can achieve global information perception after multi-layer propagation and aggregate fragmented features from different test angles into a continuous fault evolution sequence. Subsequently, the Graph Neural Network can compress the full-map features through a global pooling layer, extracting comprehensive characterization information reflecting stacking stability and trajectory deflection intensity. The model inputs these features into the regressor of the output layer, and then, combined with the wear discrimination criteria learned by the model, can calculate the cumulative stress impact of asymmetric material loads on the equipment structure. By analyzing the distribution of this cumulative effect under different working conditions, the graph neural network ultimately outputs the severity of the coal mine wedge-shaped accumulation fault.

[0117] Based on the same inventive concept Figure 6 This is a schematic diagram of a coal mine inspection system for a coal mine transfer machine based on ultrasonic detection, provided as an embodiment of the present invention. The coal mine inspection system for the coal mine transfer machine based on ultrasonic detection includes:

[0118] Data acquisition module 71 is used to acquire multi-source operating data of the coal mine transfer machine under the initial transfer angle;

[0119] The first suspected point determination module 72 is used to determine information on multiple suspected wedge-shaped accumulation points under the initial transfer angle based on the multi-source operating data of the coal mine transfer machine under the initial transfer angle.

[0120] The test angle determination module 73 is used to determine multiple test re-transfer angles based on information about multiple suspected wedge-shaped accumulation points under the initial re-transfer angle.

[0121] Angle adjustment and data acquisition module 74 is used to control the transfer machine adjustment device to adjust the transfer angle of the transfer machine and the belt conveyor based on the multiple test transfer angles, and to acquire multi-source operating data of the coal mine transfer machine under each test transfer angle;

[0122] The second suspected point determination module 75 is used to determine multiple suspected wedge-shaped accumulation point information under each test transfer angle based on the multi-source operating data of the coal mine transfer machine under each test transfer angle.

[0123] The fault severity determination module 76 is used to determine the severity of the coal mine wedge accumulation fault based on the information of multiple suspected wedge accumulation points under the initial transfer angle and the information of multiple suspected wedge accumulation points under each test transfer angle.

[0124] It should be noted that, in order to simplify the descriptions disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments of this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0125] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A coal mine inspection method for coal mine transfer machines based on ultrasonic testing, characterized in that, include: Acquire multi-source operating data of the coal mine transfer machine at the initial transfer angle, wherein the multi-source operating data of the coal mine transfer machine at the initial transfer angle includes ultrasonic detection data of the coal mine transfer machine; Based on the multi-source operating data of the coal mine transfer machine under the initial transfer angle, the first suspected point determination model is used to determine the information of multiple suspected wedge-shaped accumulation points under the initial transfer angle. The first suspected point determination model is a Transformer model. The input of the first suspected point determination model is the multi-source operating data of the coal mine transfer machine under the initial transfer angle, and the output of the first suspected point determination model is the information of multiple suspected wedge-shaped accumulation points under the initial transfer angle. Multiple test relocation angles are determined based on information from multiple suspected wedge-shaped accumulation points under the initial relocation angle. Based on the multiple test transfer angles, the transfer machine adjustment device is controlled to adjust the transfer angle between the transfer machine and the belt conveyor, and multi-source operating data of the coal mine transfer machine under each test transfer angle are obtained; Based on the multi-source operating data of the coal mine transfer machine under each test transfer angle, the second suspected point determination model is used to determine the information of multiple suspected wedge-shaped accumulation points under each test transfer angle. The second suspected point determination model is a Transformer model. The input of the second suspected point determination model is the multi-source operating data of the coal mine transfer machine under each test transfer angle, and the output of the second suspected point determination model is the information of multiple suspected wedge-shaped accumulation points under each test transfer angle. The severity of the wedge-shaped accumulation fault in the coal mine is determined based on the information of multiple suspected wedge-shaped accumulation points under the initial transfer angle and the information of multiple suspected wedge-shaped accumulation points under each test transfer angle.

2. The coal mine inspection method for coal mine transfer machines based on ultrasonic detection as described in claim 1, characterized in that, The determination of multiple test retracing angles based on the information of the multiple suspected wedge-shaped accumulation points includes: K clusters were obtained by clustering based on the information of the multiple suspected wedge-shaped accumulation points; Based on the K clusters, multiple verification test re-transfer angles are determined for each cluster; Based on the K clusters and the multiple verification test relocation angles for each cluster, multiple test relocation angles are determined.

3. The coal mine inspection method for coal mine transfer machines based on ultrasonic detection as described in claim 1, characterized in that, The determination of the severity of the coal mine wedge accumulation fault based on the information of multiple suspected wedge accumulation points under the initial transfer angle and the information of multiple suspected wedge accumulation points under each test transfer angle includes: A wedge-shaped stacking analysis map is constructed, which includes multiple relocation angle nodes and edges between the multiple relocation angle nodes. The multiple relocation angle nodes include initial relocation angle nodes and test relocation angle nodes. The node features of the initial relocation angle nodes include information on multiple suspected wedge-shaped stacking points under the initial relocation angle. The node features of the test relocation angle nodes include information on multiple suspected wedge-shaped stacking points under the test relocation angle. The edges between the multiple relocation angle nodes are the differences in relocation angles. The severity of wedge-shaped accumulation faults in coal mines is obtained by processing the wedge-shaped accumulation analysis map based on the wedge-shaped accumulation analysis model.

4. The coal mine inspection method for coal mine transfer machines based on ultrasonic detection as described in claim 1, characterized in that, The multi-source operating data of the coal mine transfer machine at the initial transfer angle also includes transfer machine operating video, material humidity information, vibration sensor data, temperature sensor data, and current sensor data.

5. A coal mine inspection system for a coal mine transfer machine based on ultrasonic detection, used to execute the coal mine inspection method for a coal mine transfer machine based on ultrasonic detection as described in any one of claims 1-4, characterized in that, include: The data acquisition module is used to acquire multi-source operating data of the coal mine transfer machine at the initial transfer angle, including ultrasonic detection data of the coal mine transfer machine. The first suspected point determination module is used to determine multiple suspected wedge-shaped accumulation points under the initial transfer angle based on the multi-source operating data of the coal mine transfer machine under the initial transfer angle using the first suspected point determination model. The first suspected point determination model is a Transformer model. The input of the first suspected point determination model is the multi-source operating data of the coal mine transfer machine under the initial transfer angle, and the output of the first suspected point determination model is the information of multiple suspected wedge-shaped accumulation points under the initial transfer angle. The test angle determination module is used to determine multiple test relocation angles based on information from multiple suspected wedge-shaped accumulation points under the initial relocation angle. Angle adjustment and data acquisition module is used to control the transfer machine adjustment device to adjust the transfer angle of the transfer machine and the belt conveyor based on the multiple test transfer angles, and to acquire multi-source operating data of the coal mine transfer machine under each test transfer angle; The second suspected point determination module is used to determine multiple suspected wedge-shaped accumulation points under each test transfer angle based on the multi-source operating data of the coal mine transfer machine under each test transfer angle using the second suspected point determination model. The second suspected point determination model is a Transformer model. The input of the second suspected point determination model is the multi-source operating data of the coal mine transfer machine under each test transfer angle. The output of the second suspected point determination model is multiple suspected wedge-shaped accumulation points under each test transfer angle. The fault severity determination module is used to determine the severity of the coal mine wedge accumulation fault based on the information of multiple suspected wedge accumulation points under the initial transfer angle and the information of multiple suspected wedge accumulation points under each test transfer angle.

6. The coal mine detection system for a coal mine transfer machine based on ultrasonic detection as described in claim 5, characterized in that, The test angle determination module is also used for: K clusters were obtained by clustering based on the information of the multiple suspected wedge-shaped accumulation points; Based on the K clusters, multiple verification test re-transfer angles are determined for each cluster; Based on the K clusters and the multiple verification test relocation angles for each cluster, multiple test relocation angles are determined.

7. The coal mine detection system for a coal mine transfer machine based on ultrasonic detection as described in claim 5, characterized in that, The fault severity determination module is also used for: A wedge-shaped stacking analysis map is constructed, which includes multiple relocation angle nodes and edges between the multiple relocation angle nodes. The multiple relocation angle nodes include initial relocation angle nodes and test relocation angle nodes. The node features of the initial relocation angle nodes include information on multiple suspected wedge-shaped stacking points under the initial relocation angle. The node features of the test relocation angle nodes include information on multiple suspected wedge-shaped stacking points under the test relocation angle. The edges between the multiple relocation angle nodes are the differences in relocation angles. The severity of wedge-shaped accumulation faults in coal mines is obtained by processing the wedge-shaped accumulation analysis map based on the wedge-shaped accumulation analysis model.

8. The coal mine detection system for a coal mine transfer machine based on ultrasonic detection as described in claim 5, characterized in that, The multi-source operating data of the coal mine transfer machine at the initial transfer angle also includes transfer machine operating video, material humidity information, vibration sensor data, temperature sensor data, and current sensor data.

9. An electronic device, characterized in that, include: processor; Memory; And a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the coal mine detection method for a coal mine transfer machine based on ultrasonic detection as described in any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the coal mine detection method for a coal mine transfer machine based on ultrasonic detection as described in any one of claims 1 to 4.

Images