Fault diagnosis method for mine unmanned vehicle

Abstract

The invention discloses a well mining unmanned vehicle fault diagnosis method, and relates to the field of unmanned driving, and the method comprises the steps: collecting the multi-sensor data of a well mining unmanned vehicle, and the multi-sensor data comprise vehicle hardware state data, system operation state data and environment perception data; preprocessing the multi-sensor data to obtain a multi-source data set; according to the multi-source data set, the vehicle state is analyzed through a fault diagnosis algorithm, and potential faults are obtained; grading the potential faults to obtain fault grades; and executing a corresponding fault processing strategy according to the fault level. In order to solve the problem of low vehicle fault diagnosis precision caused by the fact that sensor data is easily interfered by environmental factors in a well mining environment in the prior art, the fault source accurate positioning capability and diagnosis accuracy are improved by constructing a multi-level rule base and a fault causal relationship network.

Description

Technical Field

[0001] This invention relates to the field of unmanned driving technology, and specifically to a fault diagnosis method for unmanned vehicles in underground mining operations. Background Technology

[0002] In unmanned vehicle systems in underground mines, the accuracy of fault diagnosis directly affects production safety and operational efficiency. However, sensor data in underground mining environments are easily affected by environmental factors such as dust, humidity, and temperature, posing a significant challenge to fault diagnosis. Traditional fault diagnosis methods struggle to achieve accurate diagnosis in such environments, necessitating the development of fault diagnosis technologies adapted to the unique conditions of underground mines.

[0003] Existing fault diagnosis technologies are mainly divided into two categories: traditional methods and intelligent methods. Traditional methods rely primarily on physical models, statistical analysis, and expert experience, typically identifying faults by analyzing the equipment's operating status based on its physical model or historical data. Intelligent methods, on the other hand, utilize artificial intelligence technologies such as machine learning and deep learning for fault diagnosis. With the development of artificial intelligence technology, intelligent fault diagnosis methods are gradually becoming mainstream. These methods can automatically learn fault characteristics from large amounts of data and achieve high-precision fault diagnosis.

[0004] However, in the unique environment of underground mines, existing fault diagnosis methods often struggle to cope with complex operating conditions and ever-changing fault scenarios. Existing fault diagnosis solutions suffer from the following main drawbacks: data from a single sensor cannot comprehensively reflect the operating status of equipment, especially in complex environments like underground mines where sensor data is easily affected by environmental factors such as dust, humidity, and temperature, leading to inaccurate diagnostic results; manual diagnosis is inefficient and relies heavily on expert experience, making it difficult to adapt to the complex and ever-changing fault scenarios in underground mines; and the lack of effective fusion and analysis of multi-source data prevents real-time, accurate fault prediction and diagnosis.

[0005] Chinese invention patent publication number CN118746348A discloses a load measurement system and method for a hydraulically articulated unmanned mining truck in underground mines. The system includes a data acquisition module and an automatic driving controller that communicates with it. The data acquisition module includes a pressure sensor and a tire pressure and temperature sensor. The automatic driving controller includes a communication module, a load calculation module, a behavior planning module, and a fault diagnosis module. The fault diagnosis module is used to determine whether the mining truck and the automatic driving controller are faulty based on the data transmitted from each module. However, this definition of fault refers only to whether the pressure sensor and tire pressure and temperature sensor are functioning normally, which is clearly incomplete. Summary of the Invention

[0006] The technical problem this invention aims to solve is that sensor data in underground mining environments is easily affected by environmental factors, leading to low accuracy in vehicle fault diagnosis. To address this, an unmanned vehicle fault diagnosis method for underground mining is provided. By constructing a multi-level rule base and a fault causal relationship network, the method improves the ability to accurately locate fault sources and the accuracy of diagnosis.

[0007] The technical solution of the present invention is: a fault diagnosis method for unmanned vehicles in underground mines, comprising the following steps: (1) collecting multi-sensor data of unmanned vehicles in underground mines, wherein the multi-sensor data includes vehicle hardware status data, system operating status data and environmental perception data; (2) preprocessing the multi-sensor data to obtain a multi-source dataset; (3) analyzing the vehicle status through a fault diagnosis algorithm based on the multi-source dataset to obtain potential faults; (4) classifying the potential faults to obtain fault levels; (5) executing corresponding fault handling strategies according to the fault levels.

[0008] The acquisition of potential faults in step (3) of the above scheme includes: (3.1) dividing the multi-source dataset into a first dataset and a second dataset according to the data acquisition time; (3.2) the first dataset is the power-on stage data and the second dataset is the autonomous driving stage data; (3.3) detecting vehicle system parameters and generating self-test results based on the first dataset; (3.4) detecting vehicle operating status parameters and communication status parameters through a fault monitoring algorithm based on the second dataset and generating autonomous driving monitoring results; (3.5) identifying potential faults based on the self-test results and autonomous driving monitoring results. (3.6) The vehicle system parameters include host system memory, vehicle hardware status, sensor status, vehicle positioning, and fuel and air pressure; (3.7) The potential faults include recoverable faults and unrecoverable faults; (3.8) The autonomous vehicle system includes multiple functional nodes, including path planning nodes, motion planning nodes, control nodes, MCU nodes, and VCU nodes, and each node is connected through a communication link.

[0009] The self-test results of step (3.3) in the above scheme include: (3.3.1) setting the longest detection time. (3.3.2) Parallel detection of host system memory, vehicle hardware status, sensor status, vehicle positioning, and fuel and air pressure; (3.3.3) Based on preset thresholds for host system memory, vehicle hardware status, sensor status, vehicle positioning, and fuel and air pressure, determine whether each detection result is normal; (3.3.4) If within the preset maximum detection time... If all test results are normal within the specified time, a self-test pass result is generated, and the testing process ends; (3.3.5) If within the longest testing time... At the end, if any test result is abnormal, a self-test failure result is generated; (3.3.6) Determine the system status based on the self-test result. If the self-test passes, the vehicle enters the driverless state. If the self-test fails, the vehicle remains powered on; (3.3.7) Add the self-test result to the fault list and publish it at a preset frequency through the fault diagnosis result topic.

[0010] The step (3.4) in the above scheme to generate unmanned driving monitoring results includes: (3.4.1) setting the fault monitoring frequency during the unmanned driving phase. (3.4.2) Parallel monitoring of vehicle operating status parameters, including path planning status, motion planning status, control status, MCU status and VCU status; (3.4.3) Parallel monitoring of vehicle communication status parameters, including communication status between nodes, communication status between nodes and controllers, and communication status of the vehicle's internal CAN bus; (3.4.4) Determining whether each monitoring result is normal based on preset operating parameter thresholds and communication parameter thresholds; (3.4.5) Adding the detected abnormal statuses to the local fault list and publishing them at a preset frequency through the fault diagnosis result topic.

[0011] The step (3.5) in the above scheme, based on the self-test results and the autonomous driving monitoring results, identifies potential faults, including: (3.5.1) at a preset frequency. It receives fault information from self-test results and autonomous driving monitoring results, and converts the fault information into a data structure including fault ID, node ID, timestamp, parameter value, and threshold to form an initial fault dataset. (3.5.2) Construct rules including the hardware layer. Communication layer rules Software layer rules and environmental perception layer rules Fault diagnosis rule base (3.5.3) Based on the initial fault dataset Using the fault diagnosis rule base Perform rule matching to generate a fault feature analysis result set containing fault source s, type t, severity g, and impact range i. (3.5.4) Calculate the conditional probability between faults based on the fault feature analysis result set F. ,when At that time, mark the fault pair A causal relationship exists, generating a set of fault causal relationships. , For the preset threshold; (3.5.5), based on the fault feature analysis result set F and the fault causal relationship set C, the decision tree model D is applied to divide the fault into a set of recoverable faults. and unrecoverable fault sets Generate a fault classification result set .

[0012] The hardware layer rules in step (3.5.2) of the above scheme. Including: (3.5.2.1), signal-to-noise ratio Monitoring rules, when Less than the threshold The time was determined to be dust interference; the communication layer rules This includes: (3.5.2.2), data packet round-trip time. Monitoring rules, when Greater than The time is determined to be an abnormal communication delay; (3.5.2.3), heartbeat packet loss count. Monitoring rules, when Greater than The time is determined to be a communication interruption; the software layer rules This includes: (3.5.2.4), algorithm execution time. Monitoring rules, when Greater than The time is determined to be an algorithm performance anomaly; the environment perception layer rules This includes: (3.5.2.5), sensor data cross-validation bias. Calculation rules, when Greater than The system was determined to have inconsistent sensor data.

[0013] The above scheme describes step (3.5.3) for generating a set of fault feature analysis results. This includes: (3.5.3.1) applying the sliding time window algorithm to the initial fault dataset D0. Perform time segmentation, where t is the current time point and Δt is the window width, to generate a time series data segment set. (3.5.3.2) Based on the time series data segment set The following feature vector is calculated: (3.5.3.2.1) Fault duration. ,in, This is the fault end time. The fault start time; (3.5.3.2.2), fault fluctuation frequency. ,in, Number of state changes; (3.5.3.2.3), Fault repetition rate ,in, This represents the number of times the fault occurred. Total observation time; (3.5.3.2.4), fault parameter deviation. ,in, These are actual parameter values. The threshold is used; (3.5.3.2.5) the extracted feature vector is... As a rule base The input is used to perform a rule matching process: (3.5.3.3) Apply hardware layer rules to each feature vector. Generate hardware fault feature set (3.5.3.4) Apply communication layer rules to each feature vector. Generate a communication fault feature set (3.5.3.5) Apply software layer rules to each feature vector. Generate software fault feature set (3.5.3.6) Apply the environment-aware layer rule to each feature vector. Generate environmental fault feature set Merge fault feature sets from each layer This forms a complete set of fault characteristic analysis results. Where: s is the source of the fault; t is the type of fault; g is the severity of the fault; and i is the scope of the fault's impact.

[0014] The step (3.5.4) in the above scheme generates the fault causal relationship set. This includes: (3.5.4.1) Constructing the system functional node dependency graph. Where V is the set of system functional nodes and E is the set of dependencies between nodes; (3.5.4.2) Based on the timestamp information in the fault feature analysis result set F, for each pair of faults Calculate the chronological order and generate a time series correlation matrix. ,in, Indicates a fault Fault occurred Before, Indicates a fault Fault occurred Then or simultaneously; (3.5.4.3) Perform path analysis on the system functional node dependency graph G to identify the nodes To the node All possible propagation paths ,in, Fault The node in question Fault The node in question; (3.5.4.4), the temporal correlation matrix Using the propagation path set P as input, the Bayesian network algorithm is applied to calculate the fault conditional probability: (3.5.4.4.1) For each pair of faults... Calculate in fault Fault under the condition of occurrence Conditional probability of occurrence (3.5.4.4.2) Regarding conditional probability Application path weight adjustment ,in, Set the weight factor for path p; (3.5.4.4.3) Set the conditional probability threshold. ,when At that time, determine the fault. With fault A causal relationship exists; (3.5.4.4.4) Construct a fault causal relationship diagram based on the calculation results. ,in, For the set of faulty nodes, The set of causal relationships between faults is as follows: (3.5.4.4.4.1) Fault nodes without incoming edges are labeled as the primary fault set O; (3.5.4.4.4.2) Fault nodes with incoming edges are labeled as the secondary fault set D; (3.5.4.4.5) The fault causal relationship graph is then... Transform into a set of fault causal relationships ,in, This is the primary fault. This is a secondary fault.

[0015] The above scheme describes step (3.5.5) for generating the fault classification result set. This includes: (3.5.5.1) constructing the decision tree model D; (3.5.5.2) analyzing the feature vectors in the fault feature analysis result set F. For each fault By traversing the decision tree model D, a preliminary fault judgment result set is obtained. (3.5.5.3) According to the fault causal relationship set C: (3.5.5.4) For the primary fault The preliminary fault diagnosis results As a result of the recoverability assessment; (3.5.5.5) for secondary faults To obtain the primary fault associated with it. The result of the recoverability assessment ,when When determined to be irreversible, force to The judgment result is updated to irreversible; (3.5.5.6) Generate the judgment result set after association processing. (3.5.5.7) Based on the final judgment result set All faults are divided into two categories: (3.5.5.7.1) Faults that are determined to be recoverable are classified into the recoverable fault set. (3.5.5.7.2) Faults deemed unrecoverable shall be classified into the unrecoverable fault set. .

[0016] The step (4) in the above scheme to obtain the fault level includes: (4.1) constructing the fault level evaluation matrix M; and based on each fault element in the fault feature analysis result set F. Query the fault level assessment matrix M to obtain the basic fault level score. (4.2) Scoring of all potential faults The scores are sorted, and the highest score is selected as the base score for the current fault level of the system. (4.3) Apply the fault level mapping rule to the basic score Convert to fault level H: (4.3.1), when When, level H is mapped to level 0 fault; (4.3.2), when When, level H is mapped to level 1 fault; (4.3.3), when When, level H is mapped to level 2 fault; (4.3.4), when At that time, level H is mapped to level 3 fault.

[0017] The beneficial effects of this invention are: (1) This application divides the diagnostic process into two stages: self-testing during startup and monitoring during unmanned driving. Differentiated detection strategies are designed for different operating stages, solving the technical deficiency of traditional methods that cannot adapt to different vehicle operating states. During startup, a maximum detection time is set. Parallel detection is performed to improve system timeliness; during the autonomous driving phase, detection is conducted at a fixed frequency. Continuous monitoring ensures timely fault detection, fundamentally improving the accuracy and coverage of diagnosis.

[0018] (2) In the underground mining environment, sensor data is easily affected by environmental factors such as dust, humidity and temperature. Traditional fault diagnosis technology uses single sensor data analysis and simple threshold judgment, which has the defects of high false alarm rate and inability to identify the original fault. This application realizes accurate identification and location of faults in complex environments through multi-source data fusion, staged diagnosis strategy and four-layer rule base architecture. Attached Figure Description

[0019] Figure 1 This is an exemplary flowchart of a fault diagnosis method for unmanned vehicles in underground mines, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the power-on self-test phase according to an embodiment of this application; Figure 3 This is a schematic diagram of the operational phase detection according to some embodiments of this application; Figure 4 This is a schematic diagram illustrating the extraction of fault features according to some embodiments of this application; Figure 5 This is a schematic diagram of a fault scoring process according to some embodiments of this application. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments and accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments implemented by those skilled in the art without creative effort are within the protection scope of the present invention.

[0021] like Figure 1 As shown, 1. Startup and Initialization: Fault monitoring node startup: The system begins operation.

[0022] Self-node initialization: The fault monitoring node first completes its own initialization.

[0023] Subscription Information: Nodes begin subscribing to two types of key information: one is the "self-inspection results" from other monitored nodes, and the other is the "diagnostic results" or instructions from the upper-level platform.

[0024] 2. Main Loop and Routine Monitoring: After initialization, the system enters a high-frequency 10Hz main loop (i.e., runs 10 times per second) for continuous monitoring.

[0025] In each loop, the system first performs a self-check on its own nodes.

[0026] Scenario 1: Self-test successful and stable operation If the self-test is successful and the cumulative running time of the nodes has exceeded 2 minutes (indicating that the system has entered a stable state), a "reset" operation will be performed: the fault counters of all nodes will be incremented by 1. This continuously incrementing counter is key to determining whether a node is "continuously unresponsive".

[0027] Scenario 2: Self-test failed or in the initial startup phase If the self-test fails, or the running time is less than 2 minutes (still in the startup phase), the fault monitoring node will only publish the self-test result once (e.g., the power-on self-test result) and will not perform counter operations.

[0028] 3. Information processing and fault diagnosis: In the main loop, the system will process the subscribed information synchronously: Update self-check / diagnostic results: Update local data based on the latest status of other subscribed nodes.

[0029] Reset Counter: Once a normal self-test result or diagnostic information is received from a node, the fault counter corresponding to that node is immediately reset to zero. This indicates that the node has just "reported that everything is fine" and is active.

[0030] Fault diagnosis: If a node's fault counter continues to increase, it means that the node has not reported its status for an extended period of time. When a node's counter reaches 25 (corresponding to a 10Hz frequency in the main loop, i.e., no response for 2.5 seconds), it is determined to be a "node shutdown fault" (a persistent fault).

[0031] If the counter value is small, it may be considered a "transient fault" (a temporary fault).

[0032] 4. Fault Statistics and Recovery: Fault statistics: The system will summarize the fault status of all nodes (including shutdown faults and transient faults).

[0033] Execute recovery logic: Based on the statistical results, the system will initiate the corresponding "node failure recovery logic", which may include attempting to restart the node or taking degraded operation strategies.

[0034] 5. Results Release: Fault classification: Based on the severity and scope of the fault, the overall vehicle condition is classified into fault levels (e.g., normal, minor fault, serious fault, etc.).

[0035] Release of overall real-time diagnostic results: Finally, the fault monitoring node will release the aggregated, graded overall real-time diagnostic results for use by other vehicle systems (such as decision and control modules) or ground monitoring platforms to take appropriate measures.

[0036] Multi-sensor data from unmanned vehicles in underground mines is collected. The multi-sensor data includes vehicle hardware status data, system operating status data, and environmental perception data. The multi-sensor data is preprocessed to obtain a multi-source dataset. Based on the multi-source dataset, the vehicle status is analyzed using a fault diagnosis algorithm to identify potential faults. The potential faults are classified to obtain fault levels. Based on the fault levels, corresponding fault handling strategies are executed.

[0037] In the unique environment of underground mines, unmanned vehicles need to collect multi-dimensional sensor data to ensure reliable system operation. This includes vehicle hardware status data such as: power system parameters (engine or motor temperature, speed, output power, and torque) to adapt to the high-load, frequent start-stop conditions of underground mines; chassis system parameters (suspension system pressure, braking system status, and steering system angle) to adapt to the uneven road surfaces of underground mines; electrical system parameters (battery charge / fuel level, main controller temperature, and power supply voltage fluctuations) to address the susceptibility of electrical systems to the humid environment of underground mines; onboard computing unit status (CPU utilization, memory usage, and storage space availability) to monitor computing resources; and communication module status (signal strength, data transmission rate, and communication stability indicators) to address the limited communication conditions underground.

[0038] System operation status data includes: positioning system status: GPS signal quality (wellhead area), inertial navigation drift rate, and laser SLAM positioning reliability; autonomous driving algorithm operation status: path planning update frequency, motion planning time overhead, and control execution cycle; functional node status: operation status of path planning nodes, motion planning nodes, control nodes, MCU nodes, and VCU nodes; communication status: round-trip time (RTT) between nodes, heartbeat packet loss count, and CAN bus load rate; and vehicle motion status: speed, acceleration, attitude angle, and wheel slippage rate, adapting to slippery road surfaces in underground mines.

[0039] Environmental perception data includes: LiDAR data (point cloud data density, effective detection range, and laser attenuation characteristics in dusty environments of underground mines); visual sensor data (image clarity, contrast, and brightness, adapting to low-light environments in underground mines); gas sensor data (methane concentration and carbon monoxide concentration, monitoring potential hazardous gases in underground mines); temperature and humidity sensor data (trends in ambient temperature and humidity); dust concentration sensor data (concentration of suspended particulate matter in the air); and obstacle sensor data (ultrasonic and millimeter-wave radar detection results, improving navigation safety in confined spaces in underground mines).

[0040] The data acquisition frequency is set with different sampling rates according to the importance and rate of change of the parameters. Key safety parameters are sampled at a high frequency of 100Hz or higher, general state parameters are sampled at a medium frequency of 10Hz, and environmental parameters are sampled at a low frequency of 1Hz, so as to ensure the real-time performance of the data while optimizing the system resource usage.

[0041] Multi-sensor data is preprocessed to obtain a multi-source dataset. Specifically, the 3σ principle or box plot method can be used to identify sensor anomalies caused by dust and vibration in the underground mining environment. An adaptive median filter is applied to lidar data to reduce dust interference. Kalman filtering is applied to inertial navigation data to reduce vibration noise. Bilateral filtering is applied to visual data to preserve edge features. A temperature and humidity compensation algorithm is applied to address sensor drift caused by the high temperature and humidity environment in underground mining. Attenuation compensation is performed on laser ranging data based on dust concentration. The friction coefficient estimate is corrected based on humidity.

[0042] This solution establishes a multi-stage fault diagnosis system for unmanned vehicles in underground mining environments, including a complete process such as phased dataset partitioning, power-on self-test, operation monitoring, rule base matching, fault feature analysis, causal relationship identification, and fault classification. This system is specifically optimized for the unique environments of underground mines, such as high dust levels, high humidity, limited communication, and confined spaces.

[0043] Based on the data collection time, the multi-source dataset was divided into data from the power-on phase (dataset 1) and data from the autonomous driving phase (dataset 2). The dividing point was set at the time when the system completed its self-check and entered autonomous driving mode.

[0044] like Figure 2 As shown, the data during the power-on phase mainly includes static detection data during system initialization, such as hardware self-test, sensor calibration, and positioning system initialization; while the data during the autonomous driving phase includes dynamic monitoring data during actual vehicle operation, such as planning execution status, control command response, and communication quality.

[0045] During the startup phase, set the maximum detection time. (Typical value is 180 seconds). Exceeding this time will forcibly terminate the self-test and result in a failure. The self-test process employs a parallel detection mechanism to simultaneously detect five key parameters: Host System Memory Check: Verify that available memory meets the minimum requirements (usually 30% of total memory), check for abnormal memory allocation, and ensure sufficient computing resources to handle sudden data processing demands in underground mining environments. Vehicle Hardware Status Check: Powertrain Self-Check: Check motor / engine temperature (<85℃) and controller response time (<50ms); Steering System Self-Check: Perform standard steering angle test (±30°) and verify actual response deviation (<2°); Braking System Self-Check: Test braking pressure (>80% of rated value) and response delay (<200ms); Battery System Self-Check: Verify voltage stability (fluctuation <5%) and sufficient capacity (>80%). Sensor Status Check: LiDAR Self-Check: Signal strength verification, point cloud density detection (>85% of nominal value), blind spot scanning; Camera Self-Check: Image clarity test, dynamic range verification, automatic exposure adjustment capability; Inertial Navigation Unit Self-Check: Zero drift test, standard attitude test. Vehicle positioning self-check: Fusion positioning accuracy assessment (wellhead area <20cm, downhole area <50cm); initial position confirmation (map matching >90%); positioning system stability test (static drift <10cm / min). Fuel and air pressure detection: Fuel / electricity detection (>120% of planned task requirements); air pressure system detection (brake air pressure >0.8MPa, auxiliary system air pressure >0.6MPa); Based on preset thresholds, the results of each test are judged. If all tests pass, a self-check pass result is generated, and the vehicle enters autonomous driving mode; if any test fails, a self-check failure result is generated, the vehicle remains powered on, and detailed fault information is published at a frequency of 10Hz through the fault diagnosis result topic. The parallel detection mechanism significantly shortens self-check time and improves operational efficiency.

[0046] like Figure 3 As shown, the fault monitoring frequency is set during the autonomous driving phase. (Typical value is 20Hz), a parallel monitoring mechanism is used to monitor two major categories of parameters simultaneously: vehicle operation status parameter monitoring: path planning status monitoring: detection planning cycle (<200ms), path smoothness (curvature <0.3 / m), path feasibility (minimum distance to obstacles >0.5m); motion planning status monitoring: speed planning rationality (acceleration <0.5m / s²), trajectory tracking error (<10cm); control status monitoring: control command execution delay (<50ms), control deviation (lateral <5cm, longitudinal <10cm); MCU status monitoring: processor load (<80%), execution cycle stability (jitter <10%); VCU status monitoring: drive motor response time (<100ms), steering execution accuracy (<1°).

[0047] Vehicle communication status parameter monitoring: Inter-node communication status: Round-trip time (RTT) (<50ms), packet loss rate (<1%); Node-controller communication status: Command response time (<100ms), control command completion rate (>99%); CAN bus communication status: Bus load rate (<70%), error frame rate (<0.1%), message loss rate (<0.5%); For underground mining environments, the threshold of the lidar in high dust areas will be dynamically adjusted (e.g., the point cloud density threshold is reduced to 70%), and the communication parameter thresholds in weak signal areas will also be relaxed accordingly (e.g., the RTT threshold is extended to 100ms).

[0048] When the monitoring results are determined to be abnormal, the system adds the fault information (including fault node, fault type, parameter value, threshold, and timestamp) to the local fault list and publishes it through the fault diagnosis result topic at a frequency of 10Hz, ensuring that fault information can be cached locally even in areas with poor communication.

[0049] like Figure 4 As shown, a four-layer fault diagnosis rule base is constructed for the special environment of underground mines. Hardware layer rules Signal-to-noise ratio (SNR) monitoring rules: When It was determined to be dust interference ( Adjustments are made dynamically based on dust concentration (typical value is 15dB); temperature monitoring rules: when the temperature... or rate of temperature change The time is judged as an abnormal heat dissipation; voltage fluctuation monitoring rules: when the voltage fluctuation rate The system was initially determined to be experiencing power instability.

[0050] Communication layer rules : Data packet round-trip time (RTT) monitoring rules: when The time was determined to be an abnormal communication delay. (50ms in areas with good communication, 100ms in areas with weak communication); heartbeat packet loss count. Monitoring rules: When The time is determined to be a communication interruption ( (3 for critical nodes, 5 for non-critical nodes); Bandwidth utilization monitoring rule: When bandwidth utilization is greater than... The time was determined to be communication congestion.

[0051] Software layer rules Algorithm execution time Monitoring rules: When The algorithm is judged to be abnormal at this time. (150% of the nominal execution time); Memory growth rate monitoring rules: when the memory growth rate is greater than... This is immediately identified as a memory leak; thread blocking monitoring rules: when the blocking time of a critical thread exceeds [a certain threshold], it is considered a memory leak. The thread is identified as abnormal.

[0052] Environmental perception layer rules Sensor data cross-validation bias Calculation rule: When The time was determined to be due to inconsistent sensor data. (10% in normal environments, 20% in high-dust environments); Environmental change rate monitoring rules: when the environmental parameter change rate is greater than... The time frame is determined to be a sudden change in the environment; the stability rule for the perception result is: when the change in the perception result in n consecutive frames is greater than... It was determined to be perceptually unstable at that time.

[0053] Each rule has multiple threshold levels, automatically selecting the appropriate threshold based on different areas of the underground mine (mine entrance, transport roadway, working face) to achieve environmental adaptability diagnosis. The multi-level rule base structure is designed directly for the characteristics of the underground mine environment, improving rule matching accuracy. The dynamic threshold mechanism adapts to the environmental differences in different areas of the underground mine, balancing diagnostic sensitivity and accuracy.

[0054] The system receives the initial fault dataset. Then, the following feature extraction and analysis steps are performed: applying the sliding time window algorithm. Segment the fault data by time, and set the window width. Different time settings are applied to different areas of the underground mine: 30 seconds for the shaft entrance area, 60 seconds for the roadway area, and 90 seconds for the working face area, to adapt to the environmental complexity of different areas.

[0055] Calculate the feature vector: For the fault data within each time window, calculate the four-dimensional feature vector, including the fault duration. Total duration of the fault; Fault fluctuation frequency The number of times the fault state changes per unit time reflects the fault stability; fault repetition rate. The frequency of repeated fault occurrences reflects the periodicity of the fault; the deviation of fault parameters. The degree of deviation between the actual parameter value and the threshold reflects the severity of the fault; in high dust environments, the system adds smoothing processing to the calculation of the fluctuation frequency Ff to reduce environmental noise interference; in the working area with large vibration, a vibration compensation coefficient is introduced into the calculation of the parameter deviation Pd.

[0056] eigenvectors The rules are input into four layers of rule bases for matching: hardware layer rule matching, mainly targeting sensor hardware anomalies, power supply anomalies, and actuator anomalies; communication layer rule matching, mainly targeting communication delays, packet loss, and insufficient bandwidth; software layer rule matching, mainly targeting algorithm anomalies, resource consumption anomalies, and execution logic anomalies; and environmental perception layer rule matching, mainly targeting environmental interference and inconsistent perception results. The rule matching adopts a fuzzy inference mechanism to quantify and score the matching degree between feature vectors and rule conditions, thereby improving matching accuracy.

[0057] By merging the four-level matching results, a complete set of fault feature analysis results is formed. Fault source s: determined based on the rule layer with the highest matching degree, such as hardware source, communication source, software source, or environment source; Fault type t: determined based on specific rules, such as parameter abnormality, functional abnormality, communication abnormality, or performance degradation; Fault severity g: mainly based on parameter deviation. Mapping, while considering duration The map is divided into four levels: minor, moderate, severe, or fatal; the scope of the fault impact i: the scope of impact is determined based on the analysis of associated nodes, and is divided into single node, local system, or global system.

[0058] Based on the fault feature analysis result set F, the system performs fault causal relationship analysis to identify primary and secondary faults and constructs a system functional node dependency graph. The node set V contains functional nodes such as path planning, motion planning, control, MCU, and VCU, while the edge set E represents the dependencies between nodes. For underground mining environments, the modeling of dependencies between sensors and control nodes has been particularly strengthened, such as key dependencies like lidar and obstacle perception, and IMU and positioning systems.

[0059] Based on the fault timestamps, the temporal sequence relationship between faults is calculated, and a time-series correlation matrix is ​​generated. To address the communication delay issue in underground mines, a time compensation mechanism is introduced, taking into account the time synchronization errors of different nodes. Path analysis is performed on the node dependency graph G to identify possible fault propagation paths. To address the unique environment of underground mines, an "environmental factor propagation weight" was designed, such as increasing the weight of dust concentration on the propagation path from lidar to sensing to planning.

[0060] Calculating fault condition probabilities using Bayesian network algorithms: Basic condition probability calculation: Path weight adjustment: Among them, weighting factors Considering path length, node criticality, and environmental factors; conditional probability threshold determination: setting a threshold. (This can be dynamically adjusted according to the mining area), when Determine the causal relationship between faults in time.

[0061] Constructing a fault cause-effect graph Nodes without incoming edges are marked as primary fault set O (the root cause of the fault); nodes with incoming edges are marked as secondary fault set D (caused by other faults). In high-dust areas of underground mines, the system automatically reduces the conditional probability threshold of environmental factor-related faults (e.g., to 0.55) to improve the environmental interference identification rate. The fault causal relationship graph is converted into a fault causal relationship set. Clearly identify which are primary faults (that need to be addressed first) and which are secondary faults (that will automatically disappear after the primary fault is resolved).

[0062] Based on the fault feature analysis result set F and the fault causal relationship set C, the system applies a decision tree model for fault classification: a decision tree model D for optimization of the underground mining environment is constructed, and the main decision nodes include: fault source judgment (hardware / communication / software / environment); fault duration analysis (temporary / continuous); fault fluctuation characteristic assessment (stable / fluctuating); fault severity judgment (minor / moderate / serious / fatal); and environmental impact factor evaluation (high / medium / low).

[0063] Using feature vectors Traverse the decision tree to obtain the preliminary fault judgment result set. To address the unique characteristics of underground mining, the decision tree employs a special processing path for environment-related faults (such as lidar anomalies caused by dust), enhancing environmental adaptability. It performs association processing based on the fault causal relationship set C: for primary faults, the decision tree's judgment result is maintained; for secondary faults, the recoverability judgment of the associated primary fault is checked. If the primary fault is unrecoverable, the secondary fault is also marked as unrecoverable. Specifically, when multiple primary faults are associated with the same secondary fault, the "strictest principle" is applied: if any primary fault is unrecoverable, the secondary fault is marked as unrecoverable.

[0064] All faults are divided into two categories: recoverable fault sets. Faults that can be automatically recovered through software restart, parameter adjustment, or environmental improvement; unrecoverable fault sets. Faults that require manual intervention, hardware repair, or system reset to resolve; For underground mining environments, the system is specially designed with "environmental adaptability judgment rules". For example, in high dust areas, sensor malfunctions may be judged as recoverable (waiting for dust to decrease), while in normal areas, they may be judged as unrecoverable (sensor malfunction).

[0065] like Figure 5As shown in Figure S4, potential faults are classified to obtain fault levels. A two-dimensional fault level evaluation matrix M is constructed, which is specifically designed according to the characteristics of unmanned vehicle faults in the underground mining environment: Row dimension (16 rows): The combination of fault source s and fault type t, including: hardware source - parameter anomaly, hardware source - functional anomaly, hardware source - communication anomaly, hardware source - performance degradation; communication source - parameter anomaly, communication source - functional anomaly, communication source - communication anomaly, communication source - performance degradation; software source - parameter anomaly, software source - functional anomaly, software source - communication anomaly, software source - performance degradation; environmental source - parameter anomaly, environmental source - functional anomaly, environmental source - communication anomaly, environmental source - performance degradation.

[0066] Column dimension (12 columns): The combination of fault severity g and impact range i, including: minor-single node, minor-local system, minor-global system; normal-single node, normal-local system, normal-global system; severe-single node, severe-local system, severe-global system; fatal-single node, fatal-local system, fatal-global system.

[0067] The matrix element values ​​are preset fault level scores (0-100 points), set according to the following principles: the larger the impact range, the higher the score; the higher the severity of the fault, the higher the score; among fault types, functional abnormalities have a higher weight than parameter abnormalities; among fault sources, hardware and software control faults have a higher weight than environmental and communication faults.

[0068] Special considerations for the underground mining environment: In high dust areas, the scoring of environmental source-related faults has an attenuation coefficient (×0.8); in areas with weak signals, the scoring of communication anomaly faults has an attenuation coefficient (×0.7); and the scoring of safety-critical function (such as braking system) related faults has an amplification coefficient (×1.5).

[0069] Examples of key elements in the matrix: Hardware source - functional abnormality - fatal - global system: 100 points (highest level); Environment source - parameter abnormality - minor - single node: 5 points (lowest non-zero level); Software source - functional abnormality - severe - local system: 75 points; Communication source - communication abnormality - general - global system: 60 points.

[0070] For each fault element in the fault feature analysis result set F The system executes the following scoring process: based on the fault element The value is retrieved from the fault level assessment matrix M, corresponding to the score value of the coordinate. For example, for a hardware source-functional anomaly-critical-local system fault, querying M[hardware source-functional anomaly][critical-local system] yields a base score of 85. The environmental parameters of the vehicle's current location (dust concentration, humidity, temperature, communication quality) are read; the environmental adaptability adjustment formula is applied: Dust adjustment: when the dust concentration is greater than a threshold... At (8 mg / m³), the basic environmental-related fault score × , Humidity adjustment: When the humidity is greater than the threshold When (85%), the basic score for electrical faults × , Communication quality adjustment: When the signal strength is less than the threshold At that time, the basic score for communication-related faults is × , .

[0071] Short-term fault decay: fault duration less than (10 seconds) and the first occurrence, score ×0.8; Long-term fault weighting: fault duration > Tl (5 minutes), score × (1 + 0.1 × min (5, T / Tl)); Repeated fault weighting: fault recurrence count n > 3 within 24 hours, score × (1 + 0.05 × min (10, n)). High-speed driving mode (>3m / s): safety-critical fault score ×1.2; Special operating condition mode (e.g., uphill / downhill >10°): stability-related fault score ×1.3; Full load mode: power system fault score ×1.2, braking system fault score ×1.3. After the above adjustments, the final score for each fault is obtained. .

[0072] Analyze the scores of all potential faults to determine the current fault level of the system: score all faults. Sort in descending order; filter for related faults: if a causal relationship exists. ,and If it is the primary fault, it will be given priority in the scoring comparison. Remove duplicate impacts: For multiple faults originating from the same source and originating from the same node, only the highest score is used. The highest score after sorting is selected as the base score for the current fault level of the system. For combinations of primary and secondary faults, a weighted formula is used: When multiple high-scoring faults exist simultaneously (the first three faults all score > 60 points), the combined effect formula is applied: In key areas of underground mines (such as turns and intersections), the weight of the combined effect is increased by 50%. A fault memory decay mechanism is implemented: if a fault occurs in the past... If a high level is reached within a certain time (usually 10 minutes), when it occurs again, even if the score is lower, it will still maintain a score no lower than 70% of the previous score; this mechanism is particularly suitable for handling intermittent faults in underground mining environments.

[0073] Apply fault level mapping rules to base scores Convert to fault level H: Fault level classification standard: Level 0 fault (normal state): Level 1 fault (fault of concern): , Level 2 fault (general fault): , Level 3 fault (critical fault): , The maximum is 100. Differential mapping for underground mining areas: Different threshold adjustment coefficients are applied to different underground mining areas: Wellhead area: Standard threshold ( , ); Transport roadway: Threshold fine-tuning ( , ), improve environmental interference tolerance; working face area: threshold increased ( , It adapts to dust and humidity interference.

[0074] Dust concentration dynamic threshold: An adaptive dust concentration sensing mechanism is designed. When a high dust concentration environment is detected: Slight dust (<10 mg / m³): Standard threshold; Moderate dust (10 to 20 mg / m³): Add 5 points. Add 10 points; High concentration of dust (>20mg / m³): Add 10 points. Add 15 points.

[0075] Task criticality considerations: Adjust thresholds based on the criticality of the current task: Standard transportation tasks: standard thresholds; Special material transportation (e.g., explosive materials): Reduce by 5 points. Deduct 10 points; Emergency tasks (such as emergency rescue): Add 5 points. Add 10 points to increase the task completion priority.

[0076] Fault level labeling and recording: Each fault is labeled with a final level H (level 0-3); the history of fault level changes is recorded, including timestamps, fault descriptions, original scores, adjustment coefficients, and final levels; a fault level change trend analysis mechanism is established to identify fault deterioration or improvement trends.

[0077] Based on the fault level, execute the corresponding fault handling strategy. When the system does not detect a fault or all fault scores are 0, execute the standard operating strategy: maintain standard control parameters, including speed, acceleration, and steering angle limits; execute the predetermined task plan, and operate normally according to the schedule and path plan.

[0078] When fault level H is mapped to level 1 (fault of concern), the system only reports the fault and does not execute a response: Detailed fault information is recorded in the vehicle fault log, including fault ID, timestamp, parameter values, score, and level; the operator is prompted to pay attention to the fault in yellow text on the monitoring interface; and the fault diagnosis results are discussed at a standard frequency (…). It publishes fault information to the upstream system.

[0079] When a Level 2 fault is detected during the system initialization phase (power-on self-test): immediately terminate the self-test process and generate a self-test failure result; lock the vehicle control system and prohibit entry into autonomous driving mode; send detailed fault information and location data to the control center; trigger an audible and visual alarm to alert on-site personnel to the fault.

[0080] When a Level 2 fault is detected during the autonomous driving phase: Enter the "Safely Complete Current Task" mode: reduce the vehicle speed to a safe speed (usually 60% of the normal speed); increase the safe distance from other vehicles and obstacles (increase by 30%); avoid new task scheduling and reject new route planning requests; plan the nearest safe parking point, and drive to the safe parking point after completing the current road segment task; after arriving at the safe parking point, automatically switch to manual mode and wait for manual handling.

[0081] When fault level H is mapped to level 3 (serious fault), the system executes an emergency response: immediately activates the emergency stop procedure, calculates the optimal deceleration based on the current vehicle speed and road conditions; stops the vehicle at the current position as soon as possible while ensuring safety; special handling for narrow underground tunnels: avoid stopping at bends or intersections as much as possible, and move at low speed to the nearest safe area if necessary.

[0082] The invention and its embodiments have been described above illustratively. This description is not restrictive, and the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. The accompanying drawings are only one embodiment of the invention, and the actual structure is not limited thereto. No reference numerals in the claims should limit the scope of the claims. Therefore, if a person skilled in the art is inspired by this description and designs a similar structure and embodiment without departing from the spirit of the invention, such design should fall within the scope of protection of this patent. Furthermore, the word "comprising" does not exclude other elements or steps, and the word "a" preceding an element does not exclude the inclusion of "a plurality" of that element. Multiple elements stated in the product claims may also be implemented by a single element through software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any specific order.

Claims

1. A fault diagnosis method for unmanned vehicles in underground mines, characterized by: Includes the following steps: (1) Collect multi-sensor data of unmanned vehicles in underground mines, including vehicle hardware status data, system operation status data and environmental perception data; (2) Preprocess the multi-sensor data to obtain a multi-source dataset; (3) Analyze the vehicle status using a fault diagnosis algorithm based on the multi-source dataset to obtain potential faults; (4) Classify the potential faults to obtain fault levels; (5) Execute corresponding fault handling strategies based on the fault levels.

2. The fault diagnosis method for unmanned vehicles in underground mines as described in claim 1, characterized in that: The step (3) of acquiring potential faults includes: (3.1) dividing the multi-source dataset into a first dataset and a second dataset according to the data acquisition time; (3.2) the first dataset is the data during the power-on phase and the second dataset is the data during the autonomous driving phase; (3.3) detecting vehicle system parameters and generating self-test results based on the first dataset; (3.4) detecting vehicle operating status parameters and communication status parameters through a fault monitoring algorithm based on the second dataset and generating autonomous driving monitoring results; (3.5) identifying potential faults based on the self-test results and the autonomous driving monitoring results.

3. The fault diagnosis method for unmanned vehicles in underground mines as described in claim 2, characterized in that: The self-test results of step (3.3) include: (3.3.1) setting the maximum test time. (3.3.2) Parallel detection of host system memory, vehicle hardware status, sensor status, vehicle positioning, and fuel and air pressure; (3.3.3) Based on preset thresholds for host system memory, vehicle hardware status, sensor status, vehicle positioning, and fuel and air pressure, determine whether each detection result is normal; (3.3.4) If within the preset maximum detection time... If all test results are normal within the specified time, a self-test pass result is generated, and the testing process ends; (3.3.5) If within the longest testing time... At the end, if any test result is abnormal, a self-test failure result is generated; (3.3.6) Determine the system status based on the self-test result. If the self-test passes, the vehicle enters the driverless state. If the self-test fails, the vehicle remains powered on; (3.3.7) Add the self-test result to the fault list and publish it at a preset frequency through the fault diagnosis result topic.

4. The fault diagnosis method for unmanned vehicles in underground mines as described in claim 3, characterized in that: The step (3.4) of generating unmanned driving monitoring results includes: (3.4.1) setting the fault monitoring frequency for the unmanned driving phase. (3.4.2) Parallel monitoring of vehicle operating status parameters, including path planning status, motion planning status, control status, MCU status and VCU status; (3.4.3) Parallel monitoring of vehicle communication status parameters, including communication status between nodes, communication status between nodes and controllers, and communication status of the vehicle's internal CAN bus; (3.4.4) Determining whether each monitoring result is normal based on preset operating parameter thresholds and communication parameter thresholds; (3.4.5) Adding the detected abnormal statuses to the local fault list and publishing them at a preset frequency through the fault diagnosis result topic.

5. A fault diagnosis method for unmanned vehicles in underground mines as described in any one of claims 2-4, characterized in that: Step (3.5) identifies potential faults based on the self-test results and the autonomous driving monitoring results, including: (3.5.1) at a preset frequency It receives fault information from self-test results and autonomous driving monitoring results, and converts the fault information into a data structure including fault ID, node ID, timestamp, parameter value, and threshold to form an initial fault dataset. (3.5.2) Construct rules including the hardware layer. Communication layer rules Software layer rules and environmental perception layer rules Fault diagnosis rule base (3.5.3) Based on the initial fault dataset Using the fault diagnosis rule base Perform rule matching to generate a fault feature analysis result set containing fault source s, type t, severity g, and impact range i. (3.5.4) Calculate the conditional probability between faults based on the fault feature analysis result set F. ,when At that time, mark the fault pair A causal relationship exists, generating a set of fault causal relationships. , For the preset threshold; (3.5.5), based on the fault feature analysis result set F and the fault causal relationship set C, the decision tree model D is applied to divide the fault into a set of recoverable faults. and unrecoverable fault sets Generate a fault classification result set .

6. A fault diagnosis method for unmanned vehicles in underground mines as described in any one of claims 5, characterized in that: The hardware layer rules in step (3.5.2) Including: (3.5.2.1), signal-to-noise ratio Monitoring rules, when Less than the threshold The time was determined to be dust interference; the communication layer rules This includes: (3.5.2.2), data packet round-trip time. Monitoring rules, when Greater than The time is determined to be an abnormal communication delay; (3.5.2.3), heartbeat packet loss count. Monitoring rules, when Greater than The time is determined to be a communication interruption; the software layer rules This includes: (3.5.2.4), algorithm execution time. Monitoring rules, when Greater than The time is determined to be an algorithm performance anomaly; the environment perception layer rules This includes: (3.5.2.5), sensor data cross-validation bias. Calculation rules, when Greater than The system was determined to have inconsistent sensor data.

7. A fault diagnosis method for unmanned vehicles in underground mines as described in any one of claims 5, characterized in that: The step (3.5.3) generates a set of fault feature analysis results. include: (3.5.3.1) Apply the sliding time window algorithm to the initial fault dataset D0. Perform time segmentation, where t is the current time point and Δt is the window width, to generate a time series data segment set. (3.5.3.2) Based on the time series data segment set The following feature vector is calculated: (3.5.3.2.1) Fault duration. ,in, This is the fault end time. The fault start time; (3.5.3.2.2), fault fluctuation frequency. ,in, Number of state changes; (3.5.3.2.3), Fault repetition rate ,in, This represents the number of times the fault occurred. Total observation time; (3.5.3.2.4), fault parameter deviation. ,in, These are actual parameter values. The threshold is used; (3.5.3.2.5) the extracted feature vector is... As a rule base The input is used to perform a rule matching process: (3.5.3.3) Apply hardware layer rules to each feature vector. Generate hardware fault feature set (3.5.3.4) Apply communication layer rules to each feature vector. Generate a communication fault feature set (3.5.3.5) Apply software layer rules to each feature vector. Generate software fault feature set (3.5.3.6) Apply the environment-aware layer rule to each feature vector. Generate environmental fault feature set Merge fault feature sets from each layer This forms a complete set of fault characteristic analysis results. Where: s is the source of the fault; t is the type of fault; g is the severity of the fault; and i is the scope of the fault's impact.

8. A fault diagnosis method for unmanned vehicles in underground mines as described in claim 5, characterized in that: The step (3.5.4) generates the fault causal relationship set. This includes: (3.5.4.1) Constructing the system functional node dependency graph. Where V is the set of system functional nodes and E is the set of dependencies between nodes; (3.5.4.2) Based on the timestamp information in the fault feature analysis result set F, for each pair of faults Calculate the chronological order and generate a time series correlation matrix. ,in, Indicates a fault Fault occurred Before, Indicates a fault Fault occurred Then or simultaneously; (3.5.4.3) Perform path analysis on the system functional node dependency graph G to identify the nodes To the node All possible propagation paths ,in, Fault The node in question Fault The node in question; (3.5.4.4), the temporal correlation matrix Using the propagation path set P as input, the Bayesian network algorithm is applied to calculate the fault conditional probability: (3.5.4.4.1) For each pair of faults... Calculate in fault Fault under the condition of occurrence Conditional probability of occurrence (3.5.4.4.2) Regarding conditional probability Application path weight adjustment ,in, Set the weight factor for path p; (3.5.4.4.3) Set the conditional probability threshold. ,when At that time, determine the fault. With fault A causal relationship exists; (3.5.4.4.4) Construct a fault causal relationship diagram based on the calculation results. ,in, For the set of faulty nodes, The set of causal relationships between faults is as follows: (3.5.4.4.4.1) Fault nodes without incoming edges are labeled as the primary fault set O; (3.5.4.4.4.2) Fault nodes with incoming edges are labeled as the secondary fault set D; (3.5.4.4.5) The fault causal relationship graph is then... Transform into a set of fault causal relationships ,in, This is the primary fault. This is a secondary fault.

9. A fault diagnosis method for unmanned vehicles in underground mines as described in claim 8, characterized in that: The step (3.5.5) generates the fault classification result set. This includes: (3.5.5.1) constructing the decision tree model D; (3.5.5.2) analyzing the feature vectors in the fault feature analysis result set F. For each fault By traversing the decision tree model D, a preliminary fault judgment result set is obtained. (3.5.5.3) According to the fault causal relationship set C: (3.5.5.4) For the primary fault The preliminary fault diagnosis results As a result of the recoverability assessment; (3.5.5.5) for secondary faults To obtain the primary fault associated with it. The result of the recoverability assessment ,when When determined to be irreversible, force to The judgment result is updated to irreversible; (3.5.5.6) Generate the judgment result set after association processing. (3.5.5.7) Based on the final judgment result set All faults are divided into two categories: (3.5.5.7.1) Faults that are determined to be recoverable are classified into the recoverable fault set. (3.5.5.7.2) Faults deemed unrecoverable shall be classified into the unrecoverable fault set. .

10. A fault diagnosis method for unmanned vehicles in underground mines as described in claim 9, characterized in that: The step (4) of obtaining the fault level includes: (4.1) constructing the fault level evaluation matrix M; and based on each fault element in the fault feature analysis result set F. Query the fault level assessment matrix M to obtain the basic fault level score. (4.2) Scoring of all potential faults The scores are sorted, and the highest score is selected as the base score for the current fault level of the system. (4.3) Apply the fault level mapping rule to the basic score Convert to fault level H: (4.3.1), when When, level H is mapped to level 0 fault; (4.3.2), when When, level H is mapped to level 1 fault; (4.3.3), when When, level H is mapped to level 2 fault; (4.3.4), when At that time, level H is mapped to level 3 fault.

Images