Method and system for detecting running state of excavator-anchor integrated machine based on multi-sensor fusion

Abstract

The application discloses a multi-sensor fusion-based tunneling-anchor integrated machine operation state detection method and system, relates to the technical field of mining equipment, and comprises the following steps: real-time monitoring of the tunneling-anchor integrated machine through a multi-sensor terminal to obtain a hydraulic, mechanical and electrical monitoring sequence with a work node identifier; constructing a credible state grid space according to work control decisions; obtaining a ternary abnormality detection result according to node identifier-guided abnormality detection; and after combining with historical fault deduction, introducing a tunneling-anchor machine coupling fault prediction model for prediction to obtain a detection atlas. The application solves the technical problems of isolated multi-sensor data, low abnormality detection precision and the inability to realize multi-dimensional fault trend deduction during the operation of the existing tunneling-anchor integrated machine, achieves multi-source data fusion analysis of the tunneling-anchor integrated machine, improves the abnormality detection accuracy of the tunneling-anchor integrated machine, and realizes the technical effect of accurate fault prediction of the tunneling-anchor integrated machine.

Description

Technical Field

[0001] This invention relates to the field of mining equipment technology, and in particular to a method and system for detecting the operating status of an integrated tunneling and anchoring machine based on multi-sensor fusion. Background Technology

[0002] In mining operations, the stable operation of roadheader-anchor (BAR) machines is crucial for mining efficiency and operational safety, and accurate detection of their operating status is key to ensuring production. Current technologies for detecting the operating status of BAR machines often rely on single or a few sensors for data collection, followed by simple analysis to determine the equipment's condition. While these methods are effective in scenarios with relatively stable equipment conditions, their limitations become apparent as mining operations become more complex. Due to the complexity of the mining environment and the strong coupling between various systems within the BAR machine, such as hydraulic, mechanical, and electrical systems, traditional methods cannot comprehensively capture multi-dimensional status information. This results in incomplete and inaccurate detection data, failing to meet the needs for precise assessment of the BAR machine's operating status and fault early warning. Summary of the Invention

[0003] This application provides a method and system for detecting the operating status of a roadheader and anchorer based on multi-sensor fusion, which solves the technical problems of isolated multi-sensor data, low anomaly detection accuracy, and inability to perform multi-dimensional fault trend inference during the operation of existing roadheader and anchorer machines.

[0004] The first aspect of this application provides a method for detecting the operating status of a roadheader / anchor operator based on multi-sensor fusion. The method includes: real-time monitoring of the roadheader / anchor operator using multiple sensor terminals to obtain a monitoring sequence of the roadheader / anchor operator with operation node identifiers, the monitoring sequence including a hydraulic monitoring sequence, a mechanical monitoring sequence, and an electrical monitoring sequence; performing reliable state feature fitting and analysis on the roadheader / anchor operator based on roadheader / anchor operation control decisions to construct a reliable state grid space; guiding the reliable state grid space to perform anomaly detection on the monitoring sequence of the roadheader / anchor operator based on the operation node identifiers to obtain a ternary anomaly detection result; performing multi-dimensional fault trend deduction on the ternary anomaly detection result based on the historical fault event set of the roadheader / anchor operator to obtain a first fault deduction result, a second fault deduction result, and a third fault deduction result; and introducing a roadheader / anchor operator coupled fault prediction model to perform coupled fault prediction on the first fault deduction result, the second fault deduction result, and the third fault deduction result to obtain a roadheader / anchor operator detection map.

[0005] In a possible implementation, a 3D model of the integrated roadheader and anchorer is performed to obtain a roadheader and anchorer model; multiple simulation controls are performed on the roadheader and anchorer model based on the roadheader and anchorer operation control decisions to obtain multiple fitted state datasets; hydraulic feature reliability analysis is performed based on the multiple fitted state datasets to construct a hydraulic reliability mesh space; mechanical feature reliability analysis is performed based on the multiple fitted state datasets to construct a mechanical reliability mesh space; electrical feature reliability analysis is performed based on the multiple fitted state datasets to construct an electrical reliability mesh space; and the reliability state mesh space is generated based on the hydraulic reliability mesh space, the mechanical reliability mesh space, and the electrical reliability mesh space.

[0006] In a possible implementation, hydraulic features are extracted from the multiple fitted state datasets to obtain multiple hydraulic response fitting sets; temporal synchronization is performed on the multiple hydraulic response fitting sets to obtain each time-series hydraulic response set; multidimensional central tendency analysis is performed on each time-series hydraulic response set to construct a reliable sequence of hydraulic features for each time-series; and gridding is performed on each reliable sequence of hydraulic features to generate the hydraulic reliable grid space.

[0007] In a possible implementation, the trusted state grid space is mapped to nodes based on the operation node identifier to obtain a matching hydraulic grid space, a matching mechanical grid space, and a matching electrical grid space; anomaly detection is performed on the hydraulic monitoring sequence based on the matching hydraulic grid space to obtain a hydraulic anomaly detection result; anomaly detection is performed on the mechanical monitoring sequence based on the matching mechanical grid space to obtain a mechanical anomaly detection result; anomaly detection is performed on the electrical monitoring sequence based on the matching electrical grid space to obtain an electrical anomaly detection result; and the ternary anomaly detection result is generated by combining the hydraulic anomaly detection result and the mechanical anomaly detection result.

[0008] In possible implementations, the historical fault event set is classified to obtain a hydraulic fault event set, a mechanical fault event set, and an electrical fault event set; based on a fault triggering sensitivity threshold, hydraulic fault trend inference is performed on the hydraulic anomaly detection results according to the hydraulic fault event set to generate a first fault inference result; based on the fault triggering sensitivity threshold, mechanical fault trend inference is performed on the mechanical anomaly detection results according to the mechanical fault event set to generate a second fault inference result; based on the fault triggering sensitivity threshold, electrical fault trend inference is performed on the electrical anomaly detection results according to the electrical fault event set to generate a third fault inference result.

[0009] In a possible implementation, fault type identification is performed based on the hydraulic fault event set to obtain top-level fault factors; direct factor tracing is performed on the top-level fault factors based on the hydraulic fault event set to obtain direct fault factors; bottom-level factor tracing is performed on the direct fault factors based on the hydraulic fault event set to obtain bottom-level fault factors; Boolean logic association is performed on the top-level fault factors, the direct fault factors, and the bottom-level fault factors to obtain a hydraulic fault deduction channel; the hydraulic anomaly detection result is input into the hydraulic fault deduction channel to obtain multiple hydraulic fault deduction paths; trigger sensitivity optimization is performed on the multiple hydraulic fault deduction paths based on the fault trigger sensitivity threshold to obtain the first fault deduction result.

[0010] In a possible implementation, triggering factors are identified based on the multiple hydraulic fault deduction paths to obtain multiple hydraulic fault factors; based on the hydraulic anomaly detection results, the multiple hydraulic fault factors are mapped and compared to obtain multiple fault triggering sensitivity coefficients; based on the multiple fault triggering sensitivity coefficients, the multiple hydraulic fault deduction paths are optimized and filtered according to the fault triggering sensitivity threshold to generate the first fault deduction result.

[0011] In a possible implementation, a roadheader monitoring dataset is obtained based on the multi-sensor terminal; the roadheader monitoring dataset is cleaned and classified to generate the roadheader monitoring sequence.

[0012] In one possible implementation, a roadheader warning signal is generated based on the roadheader detection map.

[0013] A second aspect of this application provides a multi-sensor fusion-based system for detecting the operating status of a roadheader / anchor operator. The system includes: a roadheader / anchor operator monitoring sequence acquisition module, used to monitor the roadheader / anchor operator in real time via multiple sensor terminals to obtain a roadheader / anchor operator monitoring sequence with operation node identifiers, the monitoring sequence including hydraulic monitoring sequences, mechanical monitoring sequences, and electrical monitoring sequences; a reliable state grid space construction module, used to perform reliable state feature fitting and analysis on the roadheader / anchor operator based on roadheader / anchor operator control decisions to construct a reliable state grid space; and a ternary anomaly detection result acquisition module, used to obtain the operating status of the roadheader / anchor operator based on the operation node identifiers. The industry node identifier guides the trusted state grid space to perform anomaly detection on the roadheader monitoring sequence, obtaining a ternary anomaly detection result; the fault inference result acquisition module is used to perform multi-dimensional fault trend inference on the ternary anomaly detection result based on the historical fault event set of the roadheader, obtaining a first fault inference result, a second fault inference result, and a third fault inference result; the roadheader detection map acquisition module is used to introduce a roadheader coupled fault prediction model to perform coupled fault prediction on the first fault inference result, the second fault inference result, and the third fault inference result, obtaining a roadheader detection map.

[0014] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0015] This application utilizes multiple sensor terminals to monitor the integrated roadheader and anchor machine in real time, obtaining hydraulic, mechanical, and electrical monitoring sequences with work node identifiers. Based on roadheader and anchor operation control decisions, a reliable state grid space is constructed. Guided by the work node identifiers, anomaly detection is performed on the monitoring sequences to obtain ternary anomaly detection results. Combined with historical fault event sets, multi-dimensional fault trend extrapolation is performed, and a coupled fault prediction model is introduced for prediction, resulting in a roadheader and anchor machine detection map. This achieves accurate detection of the integrated roadheader and anchor machine's operating status, making the equipment operating status detection results in mining scenarios more comprehensive and reliable. It achieves the technical effect of multi-source data fusion analysis of the integrated roadheader and anchor machine, improving the accuracy of anomaly detection, and realizing accurate fault prediction of the integrated roadheader and anchor machine. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating the method for detecting the operating status of an integrated tunneling and anchoring machine based on multi-sensor fusion, as provided in the embodiments of this application.

[0018] Figure 2 This is a schematic diagram of the operating status detection system for an integrated tunneling and anchoring machine based on multi-sensor fusion provided in an embodiment of this application.

[0019] Figure labeling: Module 1 for acquiring monitoring sequence of roadheader and anchor machine, Module 2 for constructing reliable state grid space, Module 3 for acquiring ternary anomaly detection results, Module 4 for acquiring fault inference results, and Module 5 for acquiring roadheader and anchor machine detection map. Detailed Implementation

[0020] This application provides a method and system for detecting the operating status of a roadheader and anchorer based on multi-sensor fusion, which solves the technical problems of isolated multi-sensor data, low anomaly detection accuracy, and inability to perform multi-dimensional fault trend inference during the operation of existing roadheader and anchorer machines.

[0021] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0022] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0023] Example 1, as Figure 1 As shown, a method for detecting the operating status of an integrated tunneling and anchoring machine based on multi-sensor fusion is provided, wherein the method includes:

[0024] Step A100: Real-time monitoring of the tunneling and anchoring machine is performed through multiple sensor terminals to obtain a monitoring sequence of the tunneling and anchoring machine with operation node identifiers. The monitoring sequence of the tunneling and anchoring machine includes a hydraulic monitoring sequence, a mechanical monitoring sequence, and an electrical monitoring sequence.

[0025] In this embodiment of the application, the tunneling and anchoring machine is a key piece of equipment used for tunneling and anchoring operations in mining scenarios. Its operating status involves multiple systems such as hydraulics, mechanics, and electrical systems, and its operating status needs to be detected through technologies such as multi-sensor fusion to ensure its stable operation and operational safety in mining operations.

[0026] Specifically, in mining scenarios, to comprehensively capture the operating status of the tunneling and anchoring machine, multiple sensor terminals need to be deployed at key parts of the equipment. These multiple sensor terminals can selectively collect real-time data on parameters such as pressure and flow of the hydraulic system, vibration and displacement of the mechanical structure, and current and voltage of the electrical system. Detailed steps are explained in A110-A120.

[0027] During the data collection process, each sensor terminal will synchronously record the operation node identifier corresponding to the monitored location. This identifier is associated with the specific operation stage or part of the tunneling and anchoring machine, ensuring that each set of monitoring data can accurately correspond to the actual operating scenario of the equipment.

[0028] Subsequently, the raw monitoring datasets acquired by the multi-sensor terminals are systematically processed. First, invalid or interfering data is removed by cleaning. Then, the datasets are classified and organized according to three system categories: hydraulic, mechanical, and electrical. Finally, hydraulic monitoring sequences, mechanical monitoring sequences, and electrical monitoring sequences with operation node identifiers are formed.

[0029] By deploying multiple sensor terminals, collecting real-time data, identifying and classifying associated work nodes, a comprehensive and scenario-specific monitoring sequence for the tunneling and anchoring machine was obtained, providing a reliable data foundation for subsequent status monitoring.

[0030] Step A200: Based on the tunneling and anchoring operation control decision, perform reliable state feature fitting and analysis on the integrated tunneling and anchoring machine to construct a reliable state grid space.

[0031] Optionally, a 3D model of the tunneling and anchoring machine is obtained. Based on the tunneling and anchoring operation control decision, the model is simulated and controlled multiple times to obtain multiple fitted state datasets. Then, the hydraulic, mechanical, and electrical features are analyzed separately to construct their respective reliable state mesh spaces. Finally, they are integrated to generate a reliable state mesh space. The specific steps are explained in detail in A210-A260.

[0032] Step A300: Guide the trusted state grid space to perform anomaly detection on the tunneling and anchoring machine monitoring sequence according to the operation node identifier, and obtain the ternary anomaly detection result.

[0033] In one embodiment of this application, node mapping is performed on the trusted state grid space according to the operation node identifier to obtain three matching grid spaces. Anomaly detection is performed on the hydraulic, mechanical and electrical monitoring sequences respectively to obtain corresponding results. Combined with the generated ternary anomaly detection results, the specific steps are described in detail in A310-A340.

[0034] Step A400: Based on the historical fault event set of the tunneling and anchoring machine, perform multi-dimensional fault trend deduction on the three-dimensional anomaly detection results to obtain the first fault deduction result, the second fault deduction result, and the third fault deduction result.

[0035] Specifically, the historical fault event set is classified into three fault event sets. Based on the fault trigger sensitivity threshold, the trend of hydraulic, mechanical and electrical anomaly detection results is extrapolated to generate the first, second and third fault extrapolation results. The specific steps are explained in detail in A410-A440.

[0036] Step A500: Introduce the roadheader-anchor machine coupled fault prediction model to perform coupled fault prediction on the first fault prediction result, the second fault prediction result and the third fault prediction result, and obtain the roadheader-anchor machine detection map.

[0037] Specifically, firstly, a roadheader-anchor-machine coupled fault prediction model is constructed, defining the input and output elements. The input consists of the first, second, and third fault deduction results, from which key features are extracted, including fault types in each system (e.g., hydraulic cylinder leakage, mechanical gear wear), the correlation strength of fault paths, and the temporal relationship of fault occurrence. These features collectively constitute the model's input vector. The output consists of possible coupled fault types and their corresponding probabilities of occurrence, providing core information for the subsequent generation of roadheader-anchor-machine detection maps.

[0038] The roadheader-anchor machine coupling fault prediction model selects a machine learning architecture, such as a graph neural network, that can capture the interrelationships between multiple systems. Its structure includes an input layer, a graph convolutional layer, a fully connected layer, and an output layer, which are used to receive extracted feature vectors; strengthen the influence of relevant features by learning the correlation weights between faults in each system; integrate the processed features; and output the prediction result of the coupling fault. Simultaneously, an attention mechanism is introduced to assign higher weights to fault features with high correlation, such as the strong correlation between hydraulic and mechanical faults due to power transmission, thereby improving the predictive accuracy.

[0039] When training the anchor blocker coupling fault prediction model, historical coupling fault records are used as the base data. The data is first preprocessed, converting fault types into a computable coded form and standardizing features such as correlation strength to ensure the data format meets the model input requirements. The training process is conducted in stages: initial training is performed using a portion of the data, adjusting network parameters through a loss function to allow the model to initially learn fault correlation patterns; then, the model is validated using another portion of the data, optimizing the network structure, such as adjusting the number of neurons in the graph convolutional layers to avoid overfitting; finally, the model is tested with the remaining data to ensure it has stable predictive ability for unseen coupling faults.

[0040] When applying the roadheader-anchor-jacking machine (BAM) coupled fault prediction model, the feature vectors of the first, second, and third fault deduction results are input into the trained model. The model calculates the correlation weights between faults in each system through graph convolutional layers, such as the correlation weights between hydraulic and mechanical faults due to power transmission. After processing by fully connected layers, the model outputs prediction results for various coupled faults, including fault type, involved systems, and probability of occurrence. These results are integrated into a roadheader-anchor-jacking machine (BAM) detection map, which clearly presents the correlation paths and risk levels of each coupled fault.

[0041] By constructing and training a roadheader-anchor machine coupled fault prediction model, correlation analysis and coupled prediction are performed on the fault inference results of the three systems. The resulting roadheader-anchor machine detection map can comprehensively reflect the fault linkage relationship between multiple systems, providing accurate and systematic decision-making basis for coupled fault early warning and targeted maintenance of the roadheader-anchor machine.

[0042] Furthermore, step A200 in the method provided in this application embodiment includes:

[0043] A210: Perform 3D modeling based on the aforementioned tunneling and anchoring machine to obtain the tunneling and anchoring machine model.

[0044] A220: Based on the aforementioned tunneling and anchoring operation control decision, perform multiple simulation controls on the tunneling and anchoring machine model to obtain multiple fitted state datasets.

[0045] A230: Perform reliable analysis of hydraulic features based on the multiple fitted state datasets to construct a reliable hydraulic grid space.

[0046] A240: Perform reliable analysis of mechanical features based on the multiple fitted state datasets to construct a reliable mechanical grid space.

[0047] A250: Perform electrical feature reliability analysis based on the multiple fitted state datasets to construct an electrical reliability grid space.

[0048] A260: Generate the trusted state grid space based on the hydraulic trusted grid space, the mechanical trusted grid space, and the electrical trusted grid space.

[0049] Specifically, to construct a reliable state grid space, a 3D model of the tunneling and anchoring machine is first performed. A laser scanning device is used to scan the overall structure, core component dimensions, and relative positions of the machine, acquiring point cloud data with an accuracy of 0.1mm. Then, 3D modeling software is used to process the data, reconstructing a 3D model containing key components such as the tunneling arm, anchor drilling mechanism, hydraulic pipelines, and motors. The geometric parameter error between the model and the actual object is controlled within ±1mm, ensuring that the model accurately reflects the physical structure and motion characteristics of the equipment.

[0050] Based on the constructed 3D model of the roadheader, multiple simulation controls were conducted in conjunction with roadheader operation control decisions. These decisions covered parameter settings for different operational scenarios. Examples included setting the roadheader speed to 0.8 m / min when the coal seam hardness was f=3-5, setting the cylinder thrust to 20 kN during anchoring operations, and setting the hydraulic system working pressure to 20 MPa. Specific decisions were determined by those skilled in the art based on the actual operational scenario. For these decisions, different working conditions were simulated in the simulation system, with over 200 simulation runs, each lasting one hour. Data such as hydraulic system flow rate, vibration frequency of mechanical components, and motor operating current were recorded at each moment, ultimately forming multiple fitted state datasets. Each dataset contained over 10,000 time-series data points, covering key operational characteristics of the hydraulic, mechanical, and electrical systems.

[0051] The simulation system is built around a 3D model of a roadheader and anchor machine, integrating parameter settings for different working conditions in roadheader and anchor operation control decisions, such as roadheading speed under different coal seam hardness, cylinder thrust and hydraulic system working pressure during anchoring operations. The system uses programming to drive and control the 3D model, enabling it to simulate the operation of the equipment in various working scenarios. At the same time, the system has a built-in data acquisition module that can record key parameters such as hydraulic system flow, mechanical component vibration frequency, and motor operating current in real time during the simulation, supporting simulation operation under different working conditions. This generates a fitted state dataset containing a large amount of time-series data, covering the key operating characteristics of each system of the equipment.

[0052] Next, hydraulic features are extracted from multiple fitted state datasets to obtain multiple hydraulic response fitted sets. Through time-series synchronous correlation, each time-series hydraulic response set is obtained. Then, a reliable sequence of each time-series hydraulic feature is constructed through multi-dimensional central tendency analysis. Finally, a reliable hydraulic grid space is generated through gridding processing. The specific steps are explained in detail in A231-A234.

[0053] Then, mechanical features are extracted from multiple fitted state datasets to obtain multiple mechanical response fitting sets. These fitted state datasets contain key operating parameters of the mechanical system. Feature parameters such as the vibration acceleration of the tunneling arm, the displacement of the anchoring mechanism, and the rotational speed of the transmission components are selected from these datasets, and redundant information is removed using feature extraction algorithms. For example, the mechanical response fitting set during tunneling operations includes time-series data for vibration acceleration of 0.5-3g and rotational speed of 100-300 r / min, while the set for anchoring operations includes time-series data for displacement of 50-200 mm. Each fitting set precisely corresponds to the mechanical system response characteristics under specific working conditions.

[0054] Multiple mechanical response fitting sets are time-series synchronized and correlated. Using the trigger time of the tunneling and anchoring machine's operation as the time reference, a timestamp alignment algorithm unifies the time-series data of each fitting set to the same time axis, with a sampling interval of 50ms. After synchronization, the mechanical parameters at each time node can correspond to each other, forming various time-series mechanical response sets. For example, the data at time 20 seconds simultaneously includes vibration acceleration, displacement, and rotational speed data, achieving time-series consistency of mechanical data under different working conditions.

[0055] Multidimensional central tendency analysis is performed on the sets of mechanical responses over various time periods to construct a reliable sequence of mechanical characteristics for each time period. For the multidimensional mechanical parameters at each time period node, the mean and standard deviation are calculated under multiple fitted states to determine the normal fluctuation range. For example, if the mean vibration acceleration at a certain time period node is 1.5g and the standard deviation is 0.3g, the confidence interval is set to 0.6-2.4g. Outliers outside the range are removed, and parameters that conform to the central tendency are retained to form a reliable sequence of mechanical characteristics for that time period, ensuring that the sequence can represent the normal operating state of the mechanical system.

[0056] The reliable sequence of mechanical characteristics at each time series is processed into a grid. The time axis is used as the X-axis, with each minute as a grid cell, and the reliable interval of the mechanical parameters is used as the Y-axis. For example, the vibration acceleration is divided into intervals of 0.5g. The reliable sequence data of each time series node is mapped to the grid cell, and the frequency and probability of the parameters appearing within the range are recorded to generate a reliable mechanical grid space, which intuitively presents the distribution of the normal operating state of the mechanical system at different times and within different parameter ranges.

[0057] Subsequently, electrical features are extracted from multiple fitted state datasets to obtain multiple electrical response fitting sets. These datasets contain key parameters of the electrical system. Feature parameters such as drive motor current, control loop voltage, and power factor are selected and processed using feature extraction algorithms to form electrical response fitting sets corresponding to different operating conditions. For example, the electrical response fitting set for heavy-load operation includes time-series data for current 80-150A and voltage 380V±10%, while the set for light-load operation includes time-series data for current 60-100A, accurately reflecting the response characteristics of the electrical system under different operating conditions.

[0058] Multiple electrical response fitting sets are time-series synchronized and correlated. The trigger time of the operation is used as the time base, and the time axis is unified through a timestamp alignment algorithm. The sampling interval is set to 50ms, so that the time-series data of each fitting set correspond at the same time node, forming each time-series electrical response set. For example, at the 30th second, the set simultaneously contains the current, voltage, and power factor data at that time, achieving consistency of electrical data in time sequence.

[0059] Multidimensional central tendency analysis is performed on the sets of electrical responses for each time series to construct a reliable sequence of electrical characteristics for each time series. For the multidimensional electrical parameters of each time series node, the mean and standard deviation are calculated under multiple fitted states to determine the normal fluctuation range. For example, if the mean current of a certain time series node is 110A and the standard deviation is 10A, the reliability interval is set to 80-140A. After removing outliers, parameters that conform to the central tendency are retained to form a reliable sequence of electrical characteristics for that time series, ensuring that the sequence can represent the normal operating state of the electrical system.

[0060] The reliable electrical characteristic sequences of each time series are processed into a grid. The time axis is used as the X-axis, with each minute as a grid cell and the reliable range of electrical parameters as the Y-axis. For example, the current is divided into 20A intervals. The reliable sequence data of each time series node is mapped to the grid cell, and the frequency and probability of the parameter in the range are recorded to generate an electrical reliable grid space, which intuitively presents the distribution of the normal operation status of the electrical system at different times and within different parameter ranges.

[0061] Finally, the hydraulic, mechanical, and electrical reliable grid spaces are aligned in the time dimension to establish the correlation mapping of each system parameter at the same time node. For example, the correspondence between the hydraulic pressure value and the mechanical vibration acceleration and electrical current value at a certain time node is established. Finally, they are integrated into a reliable state grid space covering the hydraulic, mechanical, and electrical systems to realize a reliable representation of the overall operating status of the tunneling and anchoring machine.

[0062] By performing reliable analysis on mechanical and electrical characteristics to construct their respective reliable mesh spaces, and integrating them with the hydraulic reliable mesh space, a reliable state mesh space that comprehensively reflects the normal operating status of each system of the tunneling and anchoring machine is formed, providing a unified and reliable reference benchmark for subsequent anomaly detection based on this space.

[0063] Furthermore, step A230 in the method provided in this application embodiment includes:

[0064] A231: Perform hydraulic feature extraction based on the multiple fitted state datasets to obtain multiple hydraulic response fitted sets.

[0065] A232: Based on the multiple hydraulic response fitting sets, perform time-series synchronization association to obtain each time-series hydraulic response set.

[0066] A233: Perform multidimensional central tendency analysis based on the aforementioned time-series hydraulic response sets to construct a reliable sequence of hydraulic characteristics for each time series.

[0067] A234: The hydraulic reliable grid space is generated by performing gridding processing based on the reliable sequence of each time-series hydraulic feature.

[0068] In this embodiment, the hydraulic response fitting set is a set obtained by extracting hydraulic features from multiple fitted state datasets. It corresponds to different working conditions and includes response data of key parameters such as pressure and flow rate of the hydraulic system, which can reflect the response characteristics of the hydraulic system under the corresponding working conditions.

[0069] Optionally, when constructing the hydraulic reliable mesh space, hydraulic features are first extracted from multiple fitted state datasets. These datasets contain various operating parameters of the hydraulic system. Key hydraulic feature parameters such as pressure, flow rate, and cylinder extension / retraction are selected. Redundant information is removed using principal component analysis feature extraction algorithms, resulting in multiple hydraulic response fitting sets. Each fitting set corresponds to a typical operating condition. For example, the hydraulic response fitting set for tunneling includes time-series data for pressures of 18-22 MPa and flow rates of 80-100 L / min, while the set for anchoring includes time-series data for pressures of 20-24 MPa and flow rates of 60-80 L / min, ensuring that each fitting set accurately reflects the hydraulic system response characteristics under the corresponding operating condition.

[0070] Next, multiple hydraulic response fitting sets are synchronized and correlated in time. Since the sampling start time and interval may differ between different fitting sets, the trigger time of the tunneling and anchoring machine's operation is used as the time reference. A timestamp alignment algorithm is used to unify the time-series data of each fitting set onto the same time axis, with a uniform sampling interval of 50ms. After synchronization, the hydraulic parameters at each time node can correspond to each other, forming various time-series hydraulic response sets. For example, at the 10th second, the set simultaneously includes the pressure value, flow rate value, and cylinder extension / retraction amount at that moment, achieving time-series consistency of hydraulic data under different operating conditions.

[0071] Then, based on the hydraulic response sets of each time series, multidimensional central tendency analysis is performed to construct a reliable sequence of hydraulic characteristics for each time series. For the multidimensional hydraulic parameters of each time series node, namely pressure, flow rate, and cylinder extension / retraction, the mean, median, and standard deviation are calculated under multiple fitted states to determine the normal fluctuation range of the parameters. For example, if the mean pressure of a certain time series node is 20 MPa and the standard deviation is 1 MPa, then 17-23 MPa is taken as the reliable pressure interval for that node. Outliers outside this range are removed, and parameters that conform to the central tendency are retained to form a reliable sequence of hydraulic characteristics for that time series, ensuring that the data in the sequence can represent the normal operating state of the hydraulic system.

[0072] Finally, the reliable sequences of hydraulic characteristics at each time series are gridded. Using the time axis as the X-axis and the reliable intervals of hydraulic parameters as the Y-axis, the reliable sequence data of each time series node is mapped to a grid cell. Each grid cell corresponds to a specific time interval and parameter range; for example, one grid cell is defined for every 1 minute on the X-axis, and each pressure interval is defined for every 2 MPa on the Y-axis. The frequency and probability of parameters occurring within this range are recorded within the grid, forming a reliable hydraulic grid space covering the entire operating cycle. This visually presents the distribution of the normal operating state of the hydraulic system at different times and within different parameter ranges.

[0073] Through the above steps, from hydraulic feature extraction to meshing, a mesh space that can accurately reflect the reliable state of the hydraulic system is constructed, providing a reliable reference benchmark for subsequent hydraulic anomaly detection based on this space.

[0074] Furthermore, step A300 in the method provided in this application embodiment includes:

[0075] A310: Map nodes in the trusted state grid space according to the operation node identifier to obtain a matching hydraulic grid space, a matching mechanical grid space, and a matching electrical grid space.

[0076] A320: Perform anomaly detection on the hydraulic monitoring sequence based on the matched hydraulic grid space to obtain hydraulic anomaly detection results.

[0077] A330: Perform anomaly detection on the mechanical monitoring sequence based on the matched mechanical grid space to obtain mechanical anomaly detection results.

[0078] A340: Perform anomaly detection on the electrical monitoring sequence according to the matching electrical grid space to obtain electrical anomaly detection results, and generate the ternary anomaly detection results by combining the hydraulic anomaly detection results and the mechanical anomaly detection results.

[0079] Specifically, when mapping nodes in the trusted state grid space based on the work node identifier, the work node where the current tunneling and anchoring machine is located is first determined, such as the tunneling node, anchoring node, or transfer node. Each node corresponds to a specific range of equipment operating parameters, as shown in Table 1.

[0080] Table 1: Equipment Operating Parameter Ranges for Each Operation Node of the Integrated Tunneling and Anchoring Machine

[0081]

[0082] Based on these node characteristics, subspaces matching the parameter range of the current node are selected from the reliable state grid space. These subspaces are: the hydraulic grid space, which contains the reliable range of hydraulic parameters under the node; the mechanical grid space, which contains the reliable range of mechanical parameters under the node; and the electrical grid space, which contains the reliable range of electrical parameters under the node. This achieves a precise correspondence between the grid space and the work node.

[0083] Next, anomaly detection is performed on the hydraulic monitoring sequence based on the matched hydraulic grid space. The real-time collected hydraulic monitoring data (such as pressure, flow rate, cylinder extension / retraction speed, etc.) are compared one by one with the reliable parameter ranges in the matched hydraulic grid space. For example, the reliable pressure range of the tunneling node in the matched hydraulic grid space is 18-22 MPa. If the pressure is 24.5 MPa at a certain moment and exceeds this range for three consecutive sampling cycles (50 ms each), it is determined to be a hydraulic anomaly and recorded as pressure exceeding the limit in the hydraulic anomaly detection results.

[0084] When performing anomaly detection on the mechanical monitoring sequence based on the matched mechanical grid space, real-time collected mechanical parameters, such as the vibration acceleration of the tunneling arm, the rotational speed of the anchoring mechanism, and the temperature of the transmission gears, are compared with the confidence thresholds in the matched mechanical grid space. Assuming the confidence range of the vibration acceleration of the anchoring node in the matched mechanical grid space is 0.5-1.5g, if a vibration acceleration of 2.3g is detected in real time and persists for two sampling cycles, it is determined to be a mechanical anomaly, and the anchoring mechanism vibration anomaly is marked in the mechanical anomaly detection results.

[0085] When performing anomaly detection on electrical monitoring sequences based on the matched electrical grid space, the real-time collected electrical parameters (such as motor operating current, control circuit voltage, power factor, etc.) are compared with the standard range in the matched electrical grid space. For example, if the reliable range of motor current at the transfer node in the matched electrical grid space is 80-110A, and a sudden increase in current to 140A is detected in real time and does not recover within one sampling period, it is determined to be an electrical anomaly, and the transfer motor overcurrent is recorded in the electrical anomaly detection results. Subsequently, the detection results of hydraulic, mechanical, and electrical anomalies are integrated to clarify the anomaly type, occurrence time, and parameter deviation value of each system, generating a ternary anomaly detection result containing information on the three types of anomalies.

[0086] By using grid space matching and subsystem anomaly detection guided by work node identifiers, the abnormal states of the hydraulic, mechanical, and electrical systems of the tunneling and anchoring machine were accurately identified and integrated, providing accurate basic data for subsequent fault trend inference based on the three-dimensional anomaly detection results.

[0087] Furthermore, step A400 in the method provided in this application embodiment includes:

[0088] A410: Classify the historical fault event set to obtain hydraulic fault event set, mechanical fault event set and electrical fault event set.

[0089] A420: Based on the fault trigger sensitivity threshold, perform hydraulic fault trend inference on the hydraulic anomaly detection results according to the hydraulic fault event set, and generate the first fault inference result.

[0090] A430: Based on the fault triggering sensitivity threshold, perform mechanical fault trend deduction on the mechanical anomaly detection results according to the mechanical fault event set, and generate the second fault deduction result.

[0091] A440: Based on the fault triggering sensitivity threshold, the electrical fault trend is extrapolated from the electrical anomaly detection results according to the electrical fault event set, and the third fault extrapolation result is generated.

[0092] In this embodiment, the fault trigger sensitivity threshold is a critical value used to measure the probability of a fault occurring. In the fault trend prediction of hydraulic, mechanical, and electrical systems, the more likely fault prediction path is selected by comparing it with the fault trigger sensitivity coefficient, so as to optimize the fault prediction results.

[0093] Specifically, before classifying the historical fault event set of the tunneling and anchoring machine, a sufficient number of historical fault records are first collected. These records cover various faults that have occurred since the equipment was put into use, including information such as the time of occurrence, the symptoms, the components involved, the maintenance plan, and the cause of the fault. For example, fault events from the past 5 years are collected to form a corresponding historical fault event set. Each record describes in detail specific details such as a sudden drop in hydraulic system pressure during tunneling operations, which was found to be due to aging and damage to the cylinder seals, or jamming of the anchor drill mechanism during anchoring operations, caused by excessive wear of the transmission gears.

[0094] During the classification process, clear standards were established based on the system type to which the fault belonged: hydraulic fault events refer to faults related to the hydraulic system, such as cylinder leakage, insufficient hydraulic pump pressure, and hydraulic pipeline blockage; mechanical fault events target faults in mechanical structures and transmission components, such as gear wear, bearing damage, and tunneling arm deformation; electrical fault events involve electrical control systems and motors, such as motor overload, short circuits, and sensor malfunctions. All historical fault records were meticulously reviewed according to these standards. For example, hydraulic valve jamming leading to sluggish action was categorized into hydraulic fault events, conveyor belt roller bearing burnout into mechanical fault events, and unstable control circuit voltage causing shutdowns into electrical fault events.

[0095] After classification, each set needs to be validated by cross-checking the component attribution and system association in the fault repair records to ensure accurate classification. The final result is a set of hydraulic fault events, a set of mechanical fault events, and a set of electrical fault events, where each fault event has clear system attributes and characteristic descriptions.

[0096] Next, based on the fault triggering sensitivity threshold, the hydraulic fault inference channel is obtained by identifying the top-level fault factor, tracing the direct and bottom-level factors, and performing Boolean logic association on the hydraulic fault event set. After inputting the hydraulic anomaly detection result to obtain multiple paths, the first fault inference result is generated through optimization. The specific steps are explained in detail in A421-A426.

[0097] Then, based on the set of mechanical failure events, mechanical failure trend inference is performed. First, the failure events within the set are identified by type. For example, through statistical analysis, four main failure types are extracted from the set of mechanical failure events: gear wear, bearing damage, tunneling arm deformation, and anchor drill mechanism jamming. These serve as top-level factors for mechanical failures, with each top-level factor corresponding to dozens of specific event records.

[0098] For each mechanical failure top-level factor, direct factors are traced. For example, for the top-level factor of gear wear, three direct factors are found from the corresponding records: insufficient lubrication, excessive load, and insufficient gear material strength. For bearing damage, direct factors such as installation deviation and excessive axial force are traced. Each top-level factor corresponds to an average of several direct factors.

[0099] Further investigation was conducted into the underlying factors of each direct factor. For example, the underlying factors for insufficient lubrication include insufficient lubricating oil, blocked lubrication pipelines, and excessively long lubrication cycles. The underlying factors for excessively heavy loads include a sudden increase in coal seam hardness and unreasonable setting of operating parameters. Each direct factor corresponds to several underlying factors.

[0100] Boolean logic is used to connect the top-level factors, direct factors, and bottom-level factors of mechanical failures to construct a mechanical failure deduction channel. For example, the trigger logic for gear wear of the top-level factor is: it occurs when the direct factor is not lubricated enough and (the bottom-level factor has insufficient lubricating oil or the bottom-level factor's lubrication pipeline is blocked), or when the direct factor is continuously overloaded and the bottom-level factor's coal seam hardness suddenly increases, forming a complete causal chain channel.

[0101] The mechanical anomaly detection results are input into the mechanical fault deduction channel. The anomaly characteristics are matched with the channel factors to obtain multiple mechanical fault deduction paths, such as sudden increase in coal seam hardness → excessive load → gear wear, blockage of lubrication pipelines → insufficient lubrication → gear wear.

[0102] Triggering factors were identified for multiple mechanical fault simulation paths, including sudden increases in coal seam hardness and blockages in lubrication pipelines. Based on the mechanical anomaly detection results, a mapping comparison was performed with each triggering factor to calculate the fault triggering sensitivity coefficient: if the current gear vibration exceeding the standard has a similarity of 0.7 to historical cases of wear caused by sudden increases in coal seam hardness and a similarity of 0.6 to cases of blocked lubrication pipelines, the corresponding coefficient was obtained. A fault triggering sensitivity threshold of 0.5 was set, and paths with coefficients ≥ 0.5 were selected to generate a second fault simulation result.

[0103] Subsequently, the trend extrapolation process for electrical faults follows the same principle as for mechanical faults. From the set of electrical fault events, top-level factors such as motor overload, circuit short circuit, and sensor malfunction are identified. For motor overload, the direct factors are traced back to excessive load and cooling system failure. Then, the underlying factors are traced back to excessive load corresponding to a sudden increase in operating resistance and incorrect electrical parameter settings. Cooling system failure corresponds to a damaged cooling fan and blocked ventilation holes.

[0104] An electrical fault deduction path is constructed using Boolean logic associations. For example, motor overload (top-level factor) is triggered when the load is too large (direct factor) and the operating resistance suddenly increases (bottom-level factor) or the electrical parameters are set incorrectly (bottom-level factor). The electrical anomaly detection results are input into the channel to obtain multiple electrical fault deduction paths, such as sudden increase in operating resistance → excessive load → motor overload, cooling fan failure → cooling system failure → motor overload.

[0105] Identify the triggering factors for each path, such as a sudden increase in operating resistance or a malfunctioning cooling fan. Map and compare these factors with electrical anomaly detection results to calculate a sensitivity coefficient. For example, the similarity between the current overcurrent and the case of a sudden increase in operating resistance is 0.65, and the similarity with the case of a malfunctioning cooling fan is 0.55. Based on a threshold of 0.5, retain the paths that meet the criteria to generate a third fault deduction result.

[0106] Using the same hierarchical factor tracing, logical association, and path screening methods as hydraulic fault trend extrapolation, fault trend extrapolation was completed for mechanical and electrical systems respectively. The generated second and third fault extrapolation results accurately retained high-probability fault paths that were highly correlated with the anomaly detection results, providing systematic and reliable basic data for subsequent coupled fault prediction.

[0107] Furthermore, step A420 in the method provided in this application embodiment includes:

[0108] A421: Based on the set of hydraulic fault events, identify the fault type and obtain the top-level factor for each fault.

[0109] A422: Based on the set of hydraulic failure events, perform direct factor tracing on the top-level factors of each failure to obtain the direct factors of each failure.

[0110] A423: Based on the set of hydraulic failure events, trace the underlying factors of each failure direct factor to obtain the underlying factors of each failure.

[0111] A424: Based on the Boolean logic association of each fault top-level factor, each fault direct factor, and each fault bottom-level factor, a hydraulic fault deduction channel is obtained.

[0112] A425: Input the hydraulic anomaly detection results into the hydraulic fault deduction channel to obtain multiple hydraulic fault deduction paths.

[0113] A426: Based on the fault trigger sensitivity threshold, trigger sensitivity optimization is performed on the multiple hydraulic fault deduction paths to obtain the first fault deduction result.

[0114] Specifically, when identifying fault types based on a set of hydraulic fault events, the fault events within the set are first summarized, and representative fault categories are extracted as top-level factors. Taking a set of hydraulic fault events as an example, by statistically analyzing the frequency of occurrence and common characteristics of each fault phenomenon, five main fault types are identified: cylinder leakage, insufficient hydraulic pump pressure, hydraulic valve sticking, pipeline blockage, and excessively high oil temperature. These are determined as the top-level factors for each fault, with each top-level factor corresponding to dozens of specific fault event records.

[0115] Next, when tracing the direct causes of each top-level fault factor, the direct causes leading to the top-level fault are extracted from the set of hydraulic fault events. For example, for the top-level factor of cylinder leakage, dozens of corresponding fault records are analyzed to trace three factors that directly caused the leakage: aging and damage to the seals, scratches on the cylinder barrel, and misalignment of the seals during installation. These are taken as the direct causes of the top-level fault. For the top-level factor of insufficient hydraulic pump pressure, direct causes such as wear of internal pump parts, air leakage in the oil suction line, and mismatch in oil viscosity are traced. Each top-level factor corresponds to an average of several direct causes.

[0116] Then, the underlying factors of each direct factor of the failure are traced to further uncover the root causes of the direct factors. Taking the aging and damage of the seal as an example, the analysis of relevant failure records reveals three underlying factors: insufficient temperature resistance of the seal material, continuous operating temperature exceeding 35°C, and excessive impurities in the oil. For the wear of internal pump parts, the underlying factors such as insufficient hardness of the parts material, poor lubrication, and long-term overload operation are traced. Each direct factor usually corresponds to several underlying factors.

[0117] Subsequently, when performing Boolean logic associations based on the top-level factors, direct factors, and bottom-level factors of each fault, the causal relationship between the three is first clarified. For example, the logic for the occurrence of the top-level factor, hydraulic cylinder leakage, is as follows: when the direct factor, the seal, ages and breaks (and the bottom-level factor, the seal material has insufficient temperature resistance, or the bottom-level factor's operating environment temperature continuously exceeds 35°C, or the bottom-level factor's oil contains excessive impurities), or when the direct factor, the hydraulic cylinder barrel, is scratched and foreign objects enter during the bottom-level factor's assembly, it is triggered. Through such logical associations, a complete hydraulic fault deduction channel is constructed, and each channel contains a causal chain from the bottom-level factor to the top-level factor.

[0118] The hydraulic anomaly detection results are then input into the hydraulic fault deduction channel, matching the anomaly characteristics with the factors in the channel. For example, if the hydraulic anomaly detection result is a sudden drop in pressure and an oil temperature rise to 40°C in the cylinder area during the tunneling stage, this anomaly will trigger the logic chain related to seal aging and damage in the cylinder leakage deduction channel, generating multiple hydraulic fault deduction paths such as insufficient temperature resistance of seal material → seal aging and damage → cylinder leakage, or excessively high working environment temperature → seal aging and damage → cylinder leakage. Each path corresponds to a possible fault development process.

[0119] Finally, hydraulic fault factors are obtained by identifying triggering factors from multiple hydraulic fault deduction paths. The fault triggering sensitivity coefficient is obtained by comparing the hydraulic anomaly detection results. Then, the path is screened based on the coefficient and the fault triggering sensitivity threshold to generate the first fault deduction result. The specific steps are explained in detail in A426-1-A426-3.

[0120] By tracing and logically associating factors in a set of hydraulic failure events, a hydraulic failure deduction channel is constructed. Combined with the results of hydraulic anomaly detection, multiple failure deduction paths are generated, realizing a systematic deduction from abnormal phenomena to the root cause of failure. This lays the foundation for subsequent path optimization based on fault trigger sensitivity thresholds.

[0121] Furthermore, step A426 in the method provided in this application embodiment includes:

[0122] A426-1: Based on the multiple hydraulic fault deduction paths, triggering factors are identified to obtain multiple hydraulic fault factors.

[0123] A426-2: Based on the hydraulic anomaly detection results, the multiple hydraulic fault factors are mapped and compared to obtain multiple fault trigger sensitivity coefficients.

[0124] A426-3: Based on the multiple fault triggering sensitivity coefficients, the multiple hydraulic fault deduction paths are optimized and filtered according to the fault triggering sensitivity thresholds to generate the first fault deduction result.

[0125] In one embodiment, when identifying triggering factors based on multiple hydraulic fault deduction paths, key factors directly causing the fault are extracted from each path as hydraulic fault factors. For example, if there are three hydraulic fault deduction paths: insufficient temperature resistance of the seal material → aging and damage of the seal → cylinder leakage, excessively high operating temperature → aging and damage of the seal → cylinder leakage, excessive impurities in the oil → seal wear → cylinder leakage, then three hydraulic fault factors are identified: insufficient temperature resistance of the seal material, excessively high operating temperature, and excessive impurities in the oil. Each factor corresponds to the initial triggering condition that causes the fault in the path.

[0126] Next, based on the hydraulic anomaly detection results, a similarity comparison was performed on multiple hydraulic fault factors. First, the specific parameters of the anomaly detection results were clarified, and then quantitative calculations were used to measure their similarity to each factor. Assume the hydraulic anomaly detection results are: a sustained oil temperature of 38℃ in the cylinder area during the tunneling stage (exceeding the normal range of 35℃), a seal wear degree of 0.6 (normal threshold ≤ 0.5), and an oil impurity content of 0.1 mg / L (normal threshold ≤ 0.2 mg / L). The similarity calculation used normalization: for the factor of excessively high ambient temperature, historical data shows that the corresponding oil temperature exceedance value for this factor was 3-8℃, and the current exceedance is 3℃, so the similarity is 3 / 8 = 0.375; for the factor of insufficient temperature resistance of the seal material, historically this factor was often accompanied by seal wear exceeding the threshold by 0.1-0.3, and the current exceedance is 0.1, so the similarity is 0.1 / 0.3 ≈ 0.333; for the factor of excessive oil impurities, the current impurity content is not excessive, so the similarity is 0. The final three fault trigger sensitivity coefficients were 0.375, 0.333, and 0.

[0127] Finally, based on multiple fault trigger sensitivity coefficients, when optimizing and screening paths according to fault trigger sensitivity thresholds, a reasonable threshold (such as 0.3) needs to be set. By comparing each coefficient with the threshold, paths corresponding to excessively high operating ambient temperature (0.375≥0.3) and insufficient temperature resistance of sealing material (0.333≥0.3) are retained, while paths corresponding to excessive impurities in the oil (0<0.3) are eliminated. Finally, a first fault deduction result containing two high-probability paths is generated.

[0128] By identifying triggering factors, quantitatively comparing similarities, and filtering thresholds, the fault deduction path that is highly correlated with the hydraulic anomaly detection results is accurately preserved, effectively improving the pertinence and reliability of the first fault deduction result.

[0129] Furthermore, step A100 in the method provided in this application embodiment includes:

[0130] A110: Obtain the anchor-breaking machine monitoring dataset based on the multi-sensor terminal.

[0131] A120: Clean and classify the blocker monitoring dataset to generate the blocker monitoring sequence.

[0132] Optionally, in mining operations, real-time monitoring of the roadheader / anchor machine requires data acquisition using multiple sensor terminals. These terminals typically include pressure sensors deployed in the machine's hydraulic lines, vibration sensors in the mechanical joints, and current sensors in the electrical control cabinet. These sensors collect data every 100 milliseconds at a preset frequency, forming a roadheader / anchor machine monitoring dataset. This dataset encompasses various types of raw information, including hydraulic system pressure values, mechanical structure vibration acceleration, and electrical system current values.

[0133] After obtaining the monitoring dataset for the tunneling and anchoring machine, it needs to be cleaned and classified. During the cleaning process, outliers caused by momentary sensor malfunctions, such as invalid data where hydraulic pressure suddenly jumps to 50 MPa, need to be removed. A small number of missing values ​​due to signal interruptions need to be filled in using the average of five adjacent data points to ensure data integrity and validity. In the classification stage, based on the system attributes corresponding to the data, the cleaned data is divided into three categories: pressure and flow data related to hydraulic pumps and cylinders are classified into the hydraulic monitoring sequence; vibration and displacement data related to the tunneling arm and anchoring mechanism are classified into the mechanical monitoring sequence; and current and voltage data related to the drive motor and control system are classified into the electrical monitoring sequence.

[0134] Through the above steps, raw data is collected from multiple sensor terminals, and after cleaning to remove invalid information and classifying the data to clarify its ownership, a well-structured hydraulic, mechanical, and electrical monitoring sequence is finally generated. This provides accurate and clearly classified basic data for subsequent anomaly detection based on a trusted state grid space, ensuring the smooth progress of the entire detection process.

[0135] Furthermore, step A500 in the method provided in this application embodiment includes:

[0136] A510: Generate a roadheader warning signal based on the roadheader detection map.

[0137] In one embodiment, when generating an early warning signal based on the detection map of the tunneling and anchoring machine, the core information in the map must first be analyzed. The detection map contains various coupled fault types, their corresponding probabilities of occurrence, and associated system information, such as coupled faults of hydraulic leakage-mechanical load imbalance-electrical overload and their probabilities of occurrence, as well as the specific component relationships involved, such as hydraulic pipelines, tunneling arms, and drive motors.

[0138] Next, multi-level warning thresholds are set based on the severity and probability of the fault. Typically, three levels are established based on probability: Level 1 warning (probability above 70%, requiring immediate action); Level 2 warning (50%-70%, requiring close monitoring); and Level 3 warning (30%-50%, requiring regular checks). Warnings are not triggered if the probability is below 30%. Simultaneously, considering the impact of the fault on equipment operation, thresholds are appropriately lowered for faults that may lead to downtime or safety accidents (such as motor overload causing short circuits) to ensure early warning.

[0139] The probability of coupled faults in the roadheader's detection map is compared with the early warning threshold to match the corresponding early warning level. If the probability of a coupled fault is 75% and it involves a critical power system, a Level 1 early warning is triggered; a mechanical structure-related fault with a probability of 60% triggers a Level 2 early warning. For faults that trigger early warnings, their core characteristics need to be extracted, including the systems involved, key components, and possible development trends, as the core content of the early warning signal.

[0140] The generated warning signals need to be transmitted through multiple channels, including audible and visual alarms on the equipment control panel (red for Level 1, yellow for Level 2, and blue for Level 3), real-time pop-ups in the central control system, and push notifications to administrators' mobile terminals. The signals clearly indicate the warning level, fault type, associated components, and preliminary handling suggestions. For example, a Level 1 warning signal might display: "Emergency: Hydraulic leakage may cause motor overload; it is recommended to immediately stop the machine and check the cylinder seals and drive motor."

[0141] After an early warning signal is generated, the operating logs of the associated equipment are synchronized, automatically recording the warning time, fault information, and handling status, providing a basis for subsequent tracing and analysis. Simultaneously, the system automatically adjusts equipment operating parameters according to the warning level; for example, during a level two warning, the tunneling speed is appropriately reduced to decrease the risk of the fault worsening.

[0142] By analyzing the detection spectrum of the tunneling and anchoring machine, setting thresholds for comparison, and generating multi-level and multi-channel early warning signals, the machine was able to promptly identify and provide early warnings of potential coupling faults in the integrated tunneling and anchoring machine. This gave the staff more time to handle the situation and effectively reduced downtime losses and safety risks caused by the escalation of the fault.

[0143] In summary, the multi-sensor fusion-based method for detecting the operating status of an integrated tunneling and anchoring machine provided in this application has the following technical effects:

[0144] This application uses multiple sensor terminals to monitor the roadheader-anchor machine in real time, obtaining hydraulic, mechanical, and electrical monitoring sequences with work node identifiers. A reliable state grid space is constructed, and anomaly detection is performed under the guidance of work node identifiers to obtain ternary anomaly detection results. Combined with historical fault event sets, multi-dimensional fault trend extrapolation is performed, and a coupled fault prediction model is introduced to predict and obtain the roadheader-anchor machine detection map. This allows for accurate detection of the roadheader-anchor machine's operating status, making the equipment operating status detection results in the mining environment more comprehensive, accurate, and reliable. It achieves the technical effect of multi-source data fusion analysis of the roadheader-anchor machine, improving the accuracy of anomaly detection, and realizing accurate fault prediction for the roadheader-anchor machine.

[0145] Example 2, as Figure 2 As shown, based on the same inventive concept as in Embodiment 1 above, this application provides a multi-sensor fusion-based operating status detection system for an integrated tunneling and anchoring machine, the system comprising:

[0146] The roadheader monitoring sequence acquisition module 1 is used to monitor the roadheader in real time through multiple sensor terminals and obtain the roadheader monitoring sequence with operation node identifiers. The roadheader monitoring sequence includes hydraulic monitoring sequence, mechanical monitoring sequence and electrical monitoring sequence.

[0147] The trusted state grid space construction module 2 is used to perform trusted state feature fitting and analysis on the tunneling and anchoring integrated machine according to the tunneling and anchoring operation control decision, and construct a trusted state grid space.

[0148] The ternary anomaly detection result acquisition module 3 is used to guide the trusted state grid space to perform anomaly detection on the tunneling and anchoring machine monitoring sequence according to the operation node identifier, and obtain the ternary anomaly detection result.

[0149] The fault prediction result acquisition module 4 is used to perform multi-dimensional fault trend prediction on the three-dimensional anomaly detection result based on the historical fault event set of the tunneling and anchoring machine, and obtain the first fault prediction result, the second fault prediction result and the third fault prediction result.

[0150] The roadheader detection map acquisition module 5 is used to introduce the roadheader coupled fault prediction model to perform coupled fault prediction on the first fault inference result, the second fault inference result and the third fault inference result to obtain the roadheader detection map.

[0151] Furthermore, the trusted state grid space construction module 2 is used to perform the following steps:

[0152] A 3D model of the tunneling and anchoring machine is obtained by performing a 3D modeling process. Multiple simulations are then performed on the tunneling and anchoring machine model based on the tunneling and anchoring operation control decisions to obtain multiple fitted state datasets. Hydraulic feature reliability analysis is performed on the multiple fitted state datasets to construct a hydraulic reliability mesh space. Mechanical feature reliability analysis is performed on the multiple fitted state datasets to construct a mechanical reliability mesh space. Electrical feature reliability analysis is performed on the multiple fitted state datasets to construct an electrical reliability mesh space. Finally, the reliability state mesh space is generated based on the hydraulic reliability mesh space, the mechanical reliability mesh space, and the electrical reliability mesh space.

[0153] Furthermore, the trusted state grid space construction module 2 is used to perform the following steps:

[0154] Hydraulic features are extracted from the multiple fitted state datasets to obtain multiple hydraulic response fitting sets; temporal synchronization correlation is performed on the multiple hydraulic response fitting sets to obtain each time-series hydraulic response set; multidimensional central tendency analysis is performed on each time-series hydraulic response set to construct a reliable sequence of hydraulic features for each time-series; and gridding is performed on each reliable sequence of hydraulic features to generate the hydraulic reliable grid space.

[0155] Furthermore, the ternary anomaly detection result acquisition module 3 is used to perform the following steps:

[0156] The trusted state grid space is mapped according to the operation node identifier to obtain a matching hydraulic grid space, a matching mechanical grid space, and a matching electrical grid space. Anomaly detection is performed on the hydraulic monitoring sequence according to the matching hydraulic grid space to obtain hydraulic anomaly detection results. Anomaly detection is performed on the mechanical monitoring sequence according to the matching mechanical grid space to obtain mechanical anomaly detection results. Anomaly detection is performed on the electrical monitoring sequence according to the matching electrical grid space to obtain electrical anomaly detection results. The three-dimensional anomaly detection results are generated by combining the hydraulic anomaly detection results and the mechanical anomaly detection results.

[0157] Furthermore, the fault deduction result acquisition module 4 is used to perform the following steps:

[0158] The historical fault event set is classified to obtain a hydraulic fault event set, a mechanical fault event set, and an electrical fault event set. Based on a fault triggering sensitivity threshold, hydraulic fault trend inference is performed on the hydraulic anomaly detection results according to the hydraulic fault event set to generate a first fault inference result. Based on the fault triggering sensitivity threshold, mechanical fault trend inference is performed on the mechanical anomaly detection results according to the mechanical fault event set to generate a second fault inference result. Based on the fault triggering sensitivity threshold, electrical fault trend inference is performed on the electrical anomaly detection results according to the electrical fault event set to generate a third fault inference result.

[0159] Furthermore, the fault deduction result acquisition module 4 is used to perform the following steps:

[0160] Fault type identification is performed based on the hydraulic fault event set to obtain top-level fault factors; direct factor tracing is performed on the top-level fault factors based on the hydraulic fault event set to obtain direct fault factors; bottom-level factor tracing is performed on the direct fault factors based on the hydraulic fault event set to obtain bottom-level fault factors; Boolean logic association is performed on the top-level fault factors, the direct fault factors, and the bottom-level fault factors to obtain a hydraulic fault deduction channel; the hydraulic anomaly detection result is input into the hydraulic fault deduction channel to obtain multiple hydraulic fault deduction paths; trigger sensitivity optimization is performed on the multiple hydraulic fault deduction paths based on the fault trigger sensitivity threshold to obtain the first fault deduction result.

[0161] Furthermore, the fault deduction result acquisition module 4 is used to perform the following steps:

[0162] Triggering factors are identified based on the multiple hydraulic fault deduction paths to obtain multiple hydraulic fault factors; based on the hydraulic anomaly detection results, the multiple hydraulic fault factors are mapped and compared to obtain multiple fault triggering sensitivity coefficients; based on the multiple fault triggering sensitivity coefficients, the multiple hydraulic fault deduction paths are optimized and filtered according to the fault triggering sensitivity threshold to generate the first fault deduction result.

[0163] Furthermore, the anchor-breaking machine monitoring sequence acquisition module 1 is used to perform the following steps:

[0164] Based on the multi-sensor terminal, a roadheader monitoring dataset is obtained; the roadheader monitoring dataset is cleaned and classified to generate the roadheader monitoring sequence.

[0165] Furthermore, the tunneling and anchoring machine detection map acquisition module 5 is used to perform the following steps:

[0166] Based on the roadheader detection map, a roadheader early warning signal is generated.

[0167] The multi-sensor fusion-based tunneling and anchoring machine operation status detection system provided in this embodiment of the invention can execute the multi-sensor fusion-based tunneling and anchoring machine operation status detection method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0168] Although this application makes various references to certain modules in the system according to the embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy distinction between each other and are not used to limit the scope of protection of this invention.

[0169] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application. In some cases, the actions or steps described in this application can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Claims

1. A method for detecting the running state of a combined excavator-anchor machine based on multi-sensor fusion, characterized in that, The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises:

2. The multi-sensor fusion-based running state detection method for a combined anchor-excavation machine according to claim 1, characterized in that, The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: According to the each time sequence hydraulic characteristic credible sequence, grid processing is performed to generate the hydraulic credible grid space. 3.The method of claim 1, wherein, According to the historical fault event set of the combined tunneling and anchoring machine, multi-dimensional fault trend deduction is performed on the ternary anomaly detection result, including: The historical fault event set is classified to obtain a hydraulic fault event set, a mechanical fault event set, and an electrical fault event set; Based on a fault trigger sensitive threshold, hydraulic fault trend deduction is performed on a hydraulic anomaly detection result according to the hydraulic fault event set to generate the first fault deduction result; Based on the fault trigger sensitive threshold, mechanical fault trend deduction is performed on a mechanical anomaly detection result according to the mechanical fault event set to generate the second fault deduction result; Based on the fault trigger sensitive threshold, electrical fault trend deduction is performed on an electrical anomaly detection result according to the electrical fault event set to generate the third fault deduction result.

4. The method according to claim 3, wherein, Based on a fault trigger sensitive threshold, hydraulic fault trend deduction is performed on a hydraulic anomaly detection result according to the hydraulic fault event set to generate the first fault deduction result, including: Fault type identification is performed according to the hydraulic fault event set to obtain each fault top-level factor; Direct factor tracing is performed on the each fault top-level factor according to the hydraulic fault event set to obtain each fault direct factor; Bottom-level factor tracing is performed on the each fault direct factor according to the hydraulic fault event set to obtain each fault bottom-level factor; Boolean logic association is performed according to the each fault top-level factor, the each fault direct factor, and the each fault bottom-level factor to obtain a hydraulic fault deduction channel; The hydraulic anomaly detection result is input into the hydraulic fault deduction channel to obtain a plurality of hydraulic fault deduction paths; Trigger sensitive optimization is performed on the plurality of hydraulic fault deduction paths according to the fault trigger sensitive threshold to obtain the first fault deduction result.

5. The method according to claim 4, wherein the method is characterized by, Trigger sensitive optimization is performed on the plurality of hydraulic fault deduction paths according to the fault trigger sensitive threshold to obtain the first fault deduction result, including: Trigger factor identification is performed according to the plurality of hydraulic fault deduction paths to obtain a plurality of hydraulic fault factors; According to the hydraulic anomaly detection result, the plurality of hydraulic fault factors are respectively mapped and compared to obtain a plurality of fault trigger sensitive coefficients; Based on the plurality of fault trigger sensitive coefficients, the plurality of hydraulic fault deduction paths are optimized and screened according to the fault trigger sensitive threshold to generate the first fault deduction result.

6. The multi-sensor fusion-based running state detection method for a combined anchor-excavation machine according to claim 1, characterized in that, Real-time monitoring of the combined tunneling and anchoring machine is performed through a multi-sensing terminal, including: According to the multi-sensing terminal, a tunneling and anchoring machine monitoring data set is obtained; The tunneling and anchoring machine monitoring data set is cleaned and classified to generate the tunneling and anchoring machine monitoring sequence.

7. The multi-sensor fusion-based running state detection method for a combined anchor-excavation machine according to claim 1, characterized in that, According to the tunneling and anchoring machine detection map, a tunneling and anchoring machine early warning signal is generated.

8. A running state detection system for a combined excavator-anchor machine based on multi-sensor fusion, characterized in that, A system for implementing the combined tunneling and anchoring machine operation state detection method based on multi-sensing fusion according to any one of claims 1-7, the system comprising: The anchor machine monitoring sequence acquisition module is configured to monitor the anchor-digging integrated machine in real time through the multi-sensing terminal, to obtain an anchor machine monitoring sequence with a work node identifier, and to include a hydraulic monitoring sequence, a mechanical monitoring sequence, and an electrical monitoring sequence in the anchor machine monitoring sequence; The trusted state grid space construction module is configured to fit and analyze the trusted state characteristics of the anchor-digging integrated machine according to the anchor-digging work control decision, and to construct a trusted state grid space; The ternary anomaly detection result acquisition module is configured to guide the trusted state grid space to perform anomaly detection on the anchor machine monitoring sequence according to the work node identifier, and to obtain a ternary anomaly detection result; The fault deduction result acquisition module is configured to perform multi-dimensional fault trend deduction on the ternary anomaly detection result according to a historical fault event set of the anchor-digging integrated machine, and to obtain a first fault deduction result, a second fault deduction result, and a third fault deduction result; The anchor machine detection atlas acquisition module is configured to introduce an anchor machine coupling fault prediction model to perform coupling fault prediction on the first fault deduction result, the second fault deduction result, and the third fault deduction result, and to obtain an anchor machine detection atlas.

Images