Mining electromechanical equipment fault self-diagnosis and early warning system based on intelligent sensing

Abstract

The invention discloses a mining electromechanical equipment fault self-diagnosis and early warning system based on intelligent sensing, and the system comprises a sensor array which comprises a vibration sensor, a temperature sensor and a current transformer, and is used for collecting the vibration acceleration, temperature and three-phase current effective value of mining electromechanical equipment; the edge computing node is in communication connection with the sensor array and is used for preprocessing the data acquired in real time and performing fault diagnosis and early warning based on the preprocessed data; the equipment topology modeling module is used for constructing a linkage topology network of the mining electromechanical equipment, marking the upstream and downstream relationship of each piece of equipment and distributing a unique equipment ID (Identity); and the fault feature database is used for storing normal operation state feature data of the equipment and feature evolution modes of typical faults. Therefore, the fault diagnosis accuracy in a high-dust or vibration scene can be remarkably improved, a primary fault source in a multi-equipment linkage scene can be quickly positioned, and early warning of progressive faults such as bearing wear and insulation aging can be realized.

Description

Technical Field

[0001] The present invention relates to the field of electromechanical engineering, and in particular to a fault self-diagnosis and early warning system for mining electromechanical equipment based on intelligent sensing. Background Art

[0002] Intelligent sensing-based fault diagnosis and early warning technologies have been widely used in the operation and maintenance of mining electromechanical equipment. Existing systems deploy sensors to collect equipment operating data and combine them with data analysis to monitor and warn of equipment abnormalities. However, in actual applications, existing systems have found that in complex underground environments such as high dust and strong vibration, sensors are susceptible to interference from environmental noise, resulting in distortion of collected vibration, current, and other data, which in turn affects the accuracy of fault diagnosis. Furthermore, when equipment in mining production lines operates in tandem, a single device failure may be caused by anomalies in upstream and downstream equipment. Existing systems often struggle to quickly distinguish primary from secondary faults, leading to delayed fault location. Furthermore, for progressive faults such as bearing wear and insulation aging, traditional early warning methods based on fixed thresholds struggle to capture slow parameter drift, resulting in insufficient lead time for early warnings and a lack of sufficient time for preventive maintenance. Therefore, improving the reliability of sensor data in complex environments, rapidly locating fault sources in multi-device scenarios, and enhancing early prediction capabilities for progressive faults have become pressing technical challenges in the intelligent operation and maintenance of mining electromechanical equipment. Summary of the Invention

[0003] The present invention aims to solve the technical problems in the above-mentioned technologies at least to some extent.

[0004] To this end, the present invention discloses a fault self-diagnosis and early warning system for mining electromechanical equipment based on intelligent sensing, comprising:

[0005] A sensor array, comprising at least a vibration sensor, a temperature sensor and a current transformer, for synchronously collecting vibration acceleration, temperature and three-phase current effective value of mining electromechanical equipment;

[0006] an edge computing node, communicatively connected to the sensor array, for performing preprocessing on the real-time collected data and performing fault diagnosis and early warning based on the preprocessed data;

[0007] The equipment topology modeling module is used to build a linkage topology network for mining electromechanical equipment, mark the upstream and downstream relationships of each device, and assign unique device IDs;

[0008] Fault feature database, which stores the feature data of normal operation status of storage devices and the feature evolution patterns of typical faults.

[0009] The intelligent sensing-based self-diagnosis and early warning system for mining electromechanical equipment disclosed in the present invention can:

[0010] Improve the reliability of sensor data in complex environments through multi-source data noise reduction preprocessing, and significantly improve the accuracy of fault diagnosis in high dust or vibration scenarios;

[0011] Relying on the correlation analysis of device topology network and linkage data, the primary source of faults in multi-device linkage scenarios can be quickly located, shortening troubleshooting time;

[0012] With the help of trend prediction models and fault pattern matching, slow parameter drift can be identified in advance, and early warning of progressive faults such as bearing wear and insulation aging can be achieved, providing sufficient response time for preventive maintenance.

[0013] In addition, the intelligent sensing-based self-diagnosis and early warning system for mining electromechanical equipment disclosed in the present invention may also have the following additional technical features:

[0014] In one embodiment of the present invention, the preprocessing of vibration data by the edge computing node includes:

[0015] Perform sliding average filtering on the real-time vibration data stream, create a sliding window of preset length, cyclically calculate the mean of the data in the window, and eliminate abnormal pulses that exceed the mean ± preset multiple standard deviation.

[0016] In one embodiment of the present invention, the preprocessing of temperature data by the edge computing node includes:

[0017] An exponential smoothing algorithm is used to iteratively update the temperature data. The current temperature value is calculated by combining the measured value and the previous smoothed value according to a preset weight coefficient to suppress measurement drift caused by dust adhesion.

[0018] In one embodiment of the present invention, the edge computing node is further configured to:

[0019] If the proportion of high-frequency components in the preprocessed vibration spectrum exceeds a first threshold, wavelet packet decomposition is performed on the vibration signal, and after decomposition to a preset number of layers, the fault characteristic frequency band signal is reconstructed.

[0020] In one embodiment of the present invention, the edge computing node performs fault diagnosis based on the preprocessed data, including:

[0021] The Euclidean distance between the current data and the normal state feature data of the equipment in the fault feature database is calculated. If the Euclidean distance exceeds a second threshold, an abnormality warning is generated and the deviated parameter item is marked.

[0022] In one embodiment of the present invention, a dual-axis acceleration sensor is deployed on key equipment in the sensor array to synchronously collect X-axis and Y-axis vibration data; the edge computing node extracts the instantaneous amplitude envelope of the dual-axis vibration data through Hilbert transform and calculates the vibration impact value.

[0023] In one embodiment of the present invention, when a device triggers an alarm, the edge computing node traverses upstream and downstream devices in the order of the topological network constructed by the device topology modeling module:

[0024] If the current of the upstream device continuously exceeds the first proportion of the rated current within the preset time before the alarm, and the time lag between the vibration shock of the current device and the current fluctuation of the upstream device is less than the third threshold, then it is determined that the source of the fault is the upstream device;

[0025] If the speed of the downstream device drops suddenly to below the second proportion of the rated speed, and occurs simultaneously with the current device torque suddenly increasing to above the third proportion of the rated torque, it is determined that the fault is caused by an abnormality of the downstream device.

[0026] In one embodiment of the present invention, the edge computing node's early warning of a progressive failure of a device includes:

[0027] A sliding window of preset duration is established for the vibration kurtosis value, current total harmonic distortion rate, and temperature rise rate, and the exponentially weighted moving average trend slope of each parameter is calculated;

[0028] If the absolute value of the trend slope for a consecutive preset number of days exceeds a fourth threshold, an early warning is triggered.

[0029] In one embodiment of the present invention, the edge computing node is further configured to:

[0030] The parameter trend curve extracted in real time is dynamically time-warped matched with the typical failure modes in the fault feature database. If the similarity exceeds the fifth threshold and the cumulative operating time of the equipment exceeds the preset time threshold, the remaining service life is predicted based on historical failure data statistics.

[0031] In one embodiment of the present invention, the edge computing node is an industrial controller that supports floating-point operations.

[0032] Additional contents and advantages of the present invention will be given in the following description or can be understood through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The technical solutions and beneficial effects of the present invention will become apparent and easily understood from the following contents in conjunction with the accompanying drawings, in which:

[0034] Figure 1This is a workflow diagram of the intelligent sensing-based mining electromechanical equipment fault self-diagnosis and early warning system of the present invention;

[0035] Figure 2 Another work flow chart of the intelligent sensing-based mining electromechanical equipment fault self-diagnosis and early warning system of the present invention;

[0036] Figure 3 Another work flow chart of the intelligent sensing-based mining electromechanical equipment fault self-diagnosis and early warning system of the present invention;

[0037] Figure 4 This is a system block diagram of the intelligent sensing-based mining electromechanical equipment fault self-diagnosis and early warning system of the present invention. DETAILED DESCRIPTION

[0038] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0039] The following will describe the mining electromechanical equipment fault self-diagnosis and early warning system based on intelligent sensing disclosed by the present invention with reference to the accompanying drawings.

[0040] like Figure 1 、 Figure 2 、 Figure 3 and Figure 4 As shown:

[0041] A fault self-diagnosis and early warning system for mining electromechanical equipment based on intelligent sensing, comprising:

[0042] A sensor array, comprising at least a vibration sensor, a temperature sensor and a current transformer, for synchronously collecting vibration acceleration, temperature and three-phase current effective value of mining electromechanical equipment;

[0043] Edge computing nodes, which are connected to the sensor array, are used to pre-process the real-time collected data and perform fault diagnosis and early warning based on the pre-processed data;

[0044] The equipment topology modeling module is used to build a linkage topology network for mining electromechanical equipment, mark the upstream and downstream relationships of each device, and assign unique device IDs;

[0045] Fault feature database, which stores the feature data of normal operation status of storage devices and the feature evolution patterns of typical faults.

[0046] The preprocessing of vibration data by edge computing nodes includes:

[0047] Perform sliding average filtering on the real-time vibration data stream, create a sliding window of preset length, cyclically calculate the mean of the data in the window, and eliminate abnormal pulses that exceed the mean ± preset multiple standard deviation.

[0048] The edge computing node preprocesses temperature data including:

[0049] An exponential smoothing algorithm is used to iteratively update the temperature data. The current temperature value is calculated by combining the measured value and the previous smoothed value according to a preset weight coefficient to suppress measurement drift caused by dust adhesion.

[0050] Edge computing nodes are also used to:

[0051] If the proportion of high-frequency components in the preprocessed vibration spectrum exceeds a first threshold, wavelet packet decomposition is performed on the vibration signal, and after decomposition to a preset number of layers, the fault characteristic frequency band signal is reconstructed.

[0052] The edge computing node performs fault diagnosis based on preprocessed data, including:

[0053] The Euclidean distance between the current data and the normal state feature data of the equipment in the fault feature database is calculated. If the Euclidean distance exceeds a second threshold, an abnormality warning is generated and the deviated parameter item is marked.

[0054] Key devices in the sensor array deploy dual-axis acceleration sensors to synchronously collect X-axis and Y-axis vibration data; the edge computing node extracts the instantaneous amplitude envelope of the dual-axis vibration data through Hilbert transform and calculates the vibration impact value.

[0055] When an alarm is triggered by a device, the edge computing node traverses the upstream and downstream devices in the order of the topology network constructed by the device topology modeling module:

[0056] If the current of the upstream device continuously exceeds the first proportion of the rated current within the preset time before the alarm, and the time lag between the vibration shock of the current device and the current fluctuation of the upstream device is less than the third threshold, then it is determined that the source of the fault is the upstream device;

[0057] If the speed of the downstream device drops suddenly to below the second proportion of the rated speed, and occurs simultaneously with the current device torque suddenly increasing to above the third proportion of the rated torque, it is determined that the fault is caused by an abnormality of the downstream device.

[0058] Edge computing nodes provide early warnings for progressive device failures, including:

[0059] A sliding window of preset duration is established for the vibration kurtosis value, current total harmonic distortion rate, and temperature rise rate, and the exponentially weighted moving average trend slope of each parameter is calculated;

[0060] If the absolute value of the trend slope for a consecutive preset number of days exceeds a fourth threshold, an early warning is triggered.

[0061] Edge computing nodes are also used to:

[0062] The parameter trend curve extracted in real time is dynamically time-warped matched with the typical failure modes in the fault feature database. If the similarity exceeds the fifth threshold and the cumulative operating time of the equipment exceeds the preset time threshold, the remaining service life is predicted based on historical failure data statistics.

[0063] Edge computing nodes are industrial controllers that support floating-point operations.

[0064] In an embodiment of the present invention, for sensor data noise reduction and fault diagnosis in high dust or vibration environments:

[0065] The sensor array synchronously collects equipment vibration, temperature, and current data (sampling frequency 10kHz) and transmits it to the controller via the industrial bus;

[0066] Perform sliding average filtering on the vibration data, create a 50ms sliding window, cyclically calculate the mean of the data in the window, and eliminate outliers exceeding ±3σ;

[0067] Exponential smoothing is applied to the temperature data: current value = α × measured value + (1-α) × previous smoothing value, α = 0.3, to suppress drift caused by dust adhesion;

[0068] If the high-frequency component (>500Hz) of the vibration spectrum after preprocessing accounts for more than 40%, the wavelet packet decomposition is triggered and decomposed into 3 layers to reconstruct the fault characteristic frequency band signal;

[0069] Calculate the Euclidean distance between the current data and the normal state feature library of the equipment, traverse the 10-dimensional features in the feature library, such as the vibration fundamental frequency and current harmonic coefficient, and accumulate the square root of the difference;

[0070] If the distance is greater than the normal mean + 2 times the standard deviation, an "abnormal warning" is output.

[0071] In an embodiment of the present invention, for locating the source of a fault in a multi-device linkage scenario:

[0072] Build a device topology network (e.g., motor-reducer-conveyor belt), assign an ID to each device, and mark the upstream and downstream relationships;

[0073] The main motor is equipped with a dual-axis acceleration sensor, which uses Hilbert transform to extract the instantaneous amplitude envelope of the vibration signal and calculate the impact value (vibration peak value - mean value).

[0074] Real-time monitoring of all device currents (ΔI>15% of rated current) and vibration shock values ​​(>10g). When device A triggers an alarm, the system loops through its upstream device B and downstream device C according to the topology network.

[0075] If the current of device B is continuously overloaded (>120% of the rated current) within 5 seconds before the alarm of device A, and the time lag between the vibration shock of device A and the current fluctuation of device B is less than 200ms, the fault source is determined to be device B;

[0076] If a sudden drop in speed (<80% of rated speed) of device C and a sudden increase in torque (>130% of rated torque) of device A occur simultaneously, the fault is determined to be a jam of device C causing an overload of device A.

[0077] Prioritize early warning of primary fault equipment and generate fault chain reports (such as "motor overload-reducer wear").

[0078] In an embodiment of the present invention, for early warning of progressive failure and remaining life prediction:

[0079] A 7-day sliding window was established to cyclically calculate the EWMA trend slope of vibration kurtosis, THD, and temperature rise rate (the weight decays exponentially over time, and the weight of recent data is greater than that of distant data);

[0080] If the absolute value of the slope exceeds the normal fluctuation range (e.g., kurtosis slope > 0.1 / day) for three consecutive days, an early warning is triggered;

[0081] Extract real-time trend curves (such as the 7-day trend of kurtosis value) and perform DTW matching with typical patterns in the fault feature database (such as the slowly rising curve of bearing wear before the sudden change of kurtosis);

[0082] If the similarity is greater than 85% and the cumulative operating time of the equipment exceeds the threshold (such as 5000 hours for bearings), the remaining life (for example, "remaining life: 200±50 hours") is predicted by combining the Weibull distribution model (simplified logic: based on historical failure data statistics).

[0083] Specifically, in complex underground environments such as those with high dust or strong vibration, vibration sensors are preferentially installed in locations prone to mechanical failure, such as motor bearing seats and reducer input and output shafts. Temperature sensors are embedded in winding ends or gearbox oil reservoirs, and current transformers are connected in series to the motor's three-phase power supply circuit. These three sensors achieve microsecond-level data synchronization through a synchronous clock module. Taking a belt conveyor system as an example, the sensor array can synchronously collect the drive motor's X-axis and Y-axis vibration acceleration (with a sampling frequency set to 10kHz), stator winding temperature, and three-phase current RMS values. The raw data is transmitted in real time via industrial Ethernet to the high-speed cache of the edge computing node.

[0084] The preprocessing process of vibration data by the edge computing node includes a multi-stage filtering mechanism. First, a sliding average filter is implemented, creating a sliding window with a preset duration of 50ms. This window duration is optimized based on the fundamental frequency vibration period of most mining electromechanical equipment (usually in the range of 10-100ms). As the window slides point by point along the time axis, the mean and standard deviation of the 1000 data points in the window are calculated in real time. Abnormal pulses exceeding the mean ±3 times the standard deviation are identified as environmental noise (such as interference signals generated by dust particles hitting the sensor) and are eliminated. After this processing, the signal-to-noise ratio of the vibration signal can be improved by 15-20dB, effectively suppressing the impact of random noise in high-dust environments on subsequent analysis.

[0085] To address the measurement drift caused by dust accumulation in temperature data, the edge computing node uses an exponential smoothing algorithm for iterative correction. In specific implementation, the preset weight coefficient α is set to 0.3. This coefficient, verified by long-term underground measured data, strikes a balance between response speed and smoothing effect. The current temperature value is calculated from the measured value and the previous smoothed value using the formula "current value = 0.3 × measured value + 0.7 × previous smoothed value." This effectively filters out the slow drift caused by dust accumulation forming an insulating layer on the temperature sensor surface, keeping the temperature measurement error within ±1.5°C.

[0086] When the preprocessed vibration spectrum analysis shows that the high-frequency component accounts for more than 40% (the first threshold), it indicates that the equipment may have early wear or impact failure. At this time, the wavelet packet decomposition algorithm is triggered. The decomposition level is preset to 3, which subdivides the vibration signal frequency band into 8 sub-bands (0-125Hz, 125-250Hz, ..., 875-1000Hz). By reconstructing the target frequency band signal containing the characteristic frequency of the bearing outer race fault (such as 6.28× the rotational speed), it highlights the weak fault characteristics and provides a clearer feature vector for subsequent fault type identification.

[0087] The fault diagnosis process is based on the Euclidean distance method of multi-parameter fusion. The edge computing node first retrieves the normal state characteristic data of the current device from the fault feature database. This includes 10-dimensional characteristic parameters such as the vibration fundamental frequency amplitude, the proportion of the first three harmonic components, the current total harmonic distortion (THD), and the bearing temperature gradient. By calculating the Euclidean distance between real-time data and normal state data in 10-dimensional space, when the distance value exceeds the "normal state mean + 2 times the standard deviation" (the second threshold), the device is judged to have entered an abnormal state, and an early warning message is automatically generated and the specific deviated parameter items are marked, such as "the vibration fundamental frequency amplitude exceeds the normal range by 25%" and "the current THD increases to 8%."

[0088] For critical equipment (such as the main drainage pump motor), a sensor array deploys dual-axis accelerometers to simultaneously collect vibration data along the mutually perpendicular X and Y axes. Edge computing nodes perform envelope demodulation on the dual-axis signals using a Hilbert transform, extracting the instantaneous amplitude envelope and calculating the vibration impact value (defined as the difference between the peak and mean envelope values). This impact value is sensitive to transient shocks caused by collisions between bearing rollers and raceways. When the impact value exceeds 10g, the system automatically triggers the fault tracing process.

[0089] In multi-device linkage scenarios, the tree-like topology network constructed by the device topology modeling module (e.g., "power transformer-high-voltage switchgear-motor-reducer-load device") serves as the core basis for fault source location. When a device triggers an alarm, the edge computing node prioritizes traversing upstream power supply devices according to the topological hierarchy, then checks downstream load devices. For example, in the case of a belt conveyor motor overload alarm, if the current of the upstream high-voltage switchgear continuously exceeds 120% of the rated current (the first threshold) within 5 seconds before the alarm, and the time lag between the increase in motor vibration shock value and the current fluctuation is less than 200ms (the third threshold), the fault is determined to be caused by an abnormality in the power supply system. If the speed of the downstream belt pulley suddenly drops to 80% of the rated speed (the second threshold) while the motor torque suddenly increases to 130% of the rated torque (the third threshold), the motor overload is determined to be caused by a belt jam. The system automatically locates the fault to the downstream pulley and generates a diagnostic report containing the fault chain timeline.

[0090] Early warning of progressive failures relies on dynamic monitoring of parameter trends. Edge computing nodes establish a 7-day sliding window for vibration kurtosis, current total harmonic distortion (THD), and temperature rise rate, and use an exponentially weighted moving average (EWMA) algorithm to calculate the trend slope. This algorithm assigns a 60% weight to the data from the last three days, 30% to the middle two days, and 10% to the two most recent days, highlighting recent data trends. When the absolute value of the slope for three consecutive days exceeds the normal fluctuation range (e.g., kurtosis slope > 0.1°C / day, temperature rise rate > 0.5°C / day), an early warning is triggered, alerting operations and maintenance personnel to potential equipment degradation trends.

[0091] The remaining life prediction process combines dynamic time warping (DTW) matching with reliability statistical models. Edge computing nodes perform DTW matching on parameter trend curves extracted in real time (such as a seven-day series of changes in vibration kurtosis values) with typical failure patterns in the fault signature database (such as the exponential growth trend of kurtosis values ​​in the initial stages of bearing wear). The dynamic time warping distance is calculated and converted into a similarity percentage. When the similarity exceeds 85% (the fifth threshold) and the cumulative operating time of the equipment exceeds 5,000 hours (a preset operating time threshold corresponding to the bearing design life), a simplified Weibull distribution model constructed based on historical failure data is used to predict the remaining useful life. For example, if historical data shows that the average remaining useful life of similar bearings with a similar kurtosis growth trend is 200 hours, the output prediction result is "Remaining useful life: 200 ± 50 hours," providing a quantitative basis for preventive replacement.

[0092] At the hardware implementation level, the edge computing node uses an industrial controller that supports floating-point operations, equipped with a dual-core processor, 8GB of memory, and a 128GB solid-state drive, ensuring that the entire process of data preprocessing, feature extraction, and fault diagnosis is completed within 10ms. The controller communicates with the sensor array and the upper-level monitoring system through industrial bus protocols such as ModbusTCP and CANopen. It has multiple interfaces such as RS485 and Ethernet to meet the access requirements of equipment from different manufacturers underground. Its built-in real-time operating system (RTOS) supports multi-task scheduling, ensuring that tasks such as vibration signal processing, temperature data smoothing, and topology network traversal are executed in order according to priority, and the overall system response delay is controlled within 50ms.

[0093] In summary, the intelligent sensing-based self-diagnosis and early warning system for mining electromechanical equipment disclosed in the present invention can improve the reliability of sensor data in complex environments through multi-source data noise reduction preprocessing, and significantly improve the accuracy of fault diagnosis in high dust or vibration scenarios; relying on the equipment topology network and linkage data correlation analysis, it can quickly locate the source of primary faults in multi-device linkage scenarios and shorten the troubleshooting time; with the help of trend prediction models and fault pattern matching, it can identify slow parameter drift in advance, realize early warning of progressive faults such as bearing wear and insulation aging, and provide sufficient response time for preventive maintenance.

[0094] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A mine electromechanical equipment fault self-diagnosis and early warning system based on intelligent sensing, characterized in that: include: A sensor array, comprising at least a vibration sensor, a temperature sensor and a current transformer, for synchronously collecting vibration acceleration, temperature and three-phase current effective value of mining electromechanical equipment; An edge computing node, in communication with the sensor array, configured to preprocess the real-time collected data and perform fault diagnosis and early warning based on the preprocessed data; The equipment topology modeling module is used to build a linkage topology network for mining electromechanical equipment, mark the upstream and downstream relationships of each device, and assign unique device IDs; Fault feature database, which stores the feature data of normal operation status of storage devices and the feature evolution patterns of typical faults.

2. The intelligent sensing-based self-diagnosis and early warning system for mining electromechanical equipment according to claim 1, characterized in that: The edge computing node preprocesses the vibration data including: Perform sliding average filtering on the real-time vibration data stream, create a sliding window of preset length, cyclically calculate the mean of the data in the window, and eliminate abnormal pulses that exceed the mean ± preset multiple standard deviation.

3. The intelligent sensing-based self-diagnosis and early warning system for mining electromechanical equipment according to claim 2, characterized in that: The edge computing node preprocesses the temperature data including: An exponential smoothing algorithm is used to iteratively update the temperature data. The current temperature value is calculated by combining the measured value and the previous smoothed value according to a preset weight coefficient to suppress measurement drift caused by dust adhesion.

4. The intelligent sensing-based self-diagnosis and early warning system for mining electromechanical equipment faults according to claim 2 or 3, characterized in that: The edge computing node is also used for: If the proportion of high-frequency components in the preprocessed vibration spectrum exceeds a first threshold, wavelet packet decomposition is performed on the vibration signal, and after decomposition to a preset number of layers, the fault characteristic frequency band signal is reconstructed.

5. The intelligent sensing-based self-diagnosis and early warning system for mining electromechanical equipment faults according to claim 1, characterized in that: The edge computing node performs fault diagnosis based on the preprocessed data, including: The Euclidean distance between the current data and the normal state feature data of the equipment in the fault feature database is calculated. If the Euclidean distance exceeds a second threshold, an abnormality warning is generated and the deviated parameter item is marked.

6. The intelligent sensing-based self-diagnosis and early warning system for mining electromechanical equipment faults according to claim 1, characterized in that: The key equipment in the sensor array deploys a dual-axis acceleration sensor for synchronously collecting X-axis and Y-axis vibration data; the edge computing node extracts the instantaneous amplitude envelope of the dual-axis vibration data through Hilbert transform and calculates the vibration impact value.

7. The intelligent sensing-based self-diagnosis and early warning system for mining electromechanical equipment according to claim 6, characterized in that: When a device triggers an alarm, the edge computing node traverses upstream and downstream devices in the order of the topology network constructed by the device topology modeling module: If the current of the upstream device continuously exceeds the first proportion of the rated current within the preset time before the alarm, and the time lag between the vibration shock of the current device and the current fluctuation of the upstream device is less than the third threshold, then it is determined that the source of the fault is the upstream device; If the speed of the downstream device drops suddenly to below the second proportion of the rated speed, and occurs simultaneously with the current device torque suddenly increasing to above the third proportion of the rated torque, it is determined that the fault is caused by an abnormality of the downstream device.

8. The intelligent sensing-based self-diagnosis and early warning system for mining electromechanical equipment faults according to claim 1, characterized in that: The edge computing node's early warning of progressive failures of the device includes: A sliding window of preset duration is established for the vibration kurtosis value, current total harmonic distortion rate, and temperature rise rate, and the exponentially weighted moving average trend slope of each parameter is calculated; If the absolute value of the trend slope for a consecutive preset number of days exceeds a fourth threshold, an early warning is triggered.

9. The intelligent sensing-based self-diagnosis and early warning system for mining electromechanical equipment according to claim 8, characterized in that: The edge computing node is also used for: The parameter trend curve extracted in real time is dynamically time-warped matched with the typical failure modes in the fault feature database. If the similarity exceeds the fifth threshold and the cumulative operating time of the equipment exceeds the preset time threshold, the remaining service life is predicted based on historical failure data statistics.

10. The intelligent sensing-based self-diagnosis and early warning system for mining electromechanical equipment faults according to claim 1, characterized in that: The edge computing node is an industrial controller that supports floating-point operations.

Images