Mine equipment state monitoring and maintenance method based on Internet of Things

Abstract

The invention relates to the technical field of mine equipment monitoring and maintenance, and discloses a mine equipment state monitoring and maintenance method based on the Internet of Things, which comprises the following steps: S1, multi-source data acquisition: synchronously acquiring equipment operation parameters and environmental parameters through vibration, temperature and pressure sensors and environmental sensors deployed on mine equipment; s2, edge calculation processing: performing filtering and anomaly detection on the sensor data at a mine edge gateway to generate a compressed data packet; s3, online learning and model dynamic updating; s4, new equipment is accessed; s5, maintenance decision and resource scheduling; and S6, performing closed-loop feedback optimization. High-error historical sub-trees are dynamically eliminated, new sub-trees are generated, model parameters are adjusted in combination with maintenance feedback data, on the basis that historical effective knowledge is reserved, equipment dynamic features are extracted, accurate matching of prediction results and maintenance strategies is ensured, and the self-learning ability and maintenance decision reliability of the model are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of mine equipment monitoring and maintenance, and in particular to a mine equipment status monitoring and maintenance method based on the Internet of Things. Background Art

[0002] Mine equipment status monitoring and maintenance are the core links of mine safety production. Its goal is to perceive the equipment operating status in real time, warn of potential failures and plan efficient maintenance strategies. The underground environment has complex conditions such as high dust, high humidity, gas accumulation and strong electromagnetic interference, which place strict requirements on sensor data acquisition accuracy, anti-interference ability and real-time analysis. The Internet of Things technology supports multi-dimensional collection of equipment operating parameters and synchronous monitoring of environmental parameters through sensor networks, edge computing and cloud collaborative architecture, providing a data foundation for equipment health management. Traditional equipment monitoring and maintenance mostly rely on static models and fixed threshold alarm mechanisms.

[0003] However, in current technologies, traditional technologies suffer from problems such as accumulated prediction errors and delayed systematic maintenance due to delayed static model updates, lack of maintenance feedback, and reliance on manual configuration for cold starts of new equipment. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides a mine equipment status monitoring and maintenance method based on the Internet of Things, which solves the problems of delayed static model updates, lack of maintenance feedback, and reliance on manual configuration for cold starts of new equipment, which lead to accumulated prediction errors and delayed systematic maintenance.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a mine equipment status monitoring and maintenance method based on the Internet of Things, comprising:

[0006] S1. Multi-source data acquisition: Vibration, temperature, pressure and environmental sensors deployed on mine equipment are used to synchronously collect equipment operating parameters and environmental parameters;

[0007] S2, edge computing processing: Filtering and anomaly detection of sensor data at the mine edge gateway to generate compressed data packets;

[0008] S3. Online learning and dynamic model updating: This model inputs collected multi-source data and real-time updated maintenance effect data into the incremental random forest model, outputs predictions of failure type, probability, and remaining life, and matches them with a pre-set failure mode library.

[0009] S4, New device access: When a new device is connected to the system, its device type and sensor configuration are automatically extracted, and the monitoring model is dynamically expanded based on the prediction results of S3 to generate initial fault prediction rules for the new device;

[0010] S5. Maintenance decision-making and resource scheduling: Generate maintenance work orders based on model prediction results and equipment priority, and mobilize equipment maintenance parts inventory;

[0011] S6. Closed-loop feedback optimization: Feedback maintenance effect data to the online learning model to iteratively optimize prediction accuracy and maintenance strategies.

[0012] Preferably, the environmental sensors in S1 include non-dispersive infrared gas sensors, laser scattering dust sensors and digital temperature and humidity sensors;

[0013] The acquisition frequency of the equipment operating parameters is dynamically adjusted according to the equipment type, including a mining machine vibration signal sampling rate ≥10kHz, a conveyor motor current signal sampling rate ≥2kHz, and all sensor data are synchronized at the microsecond level through the IEEE 1588PTP protocol.

[0014] Preferably, the step S2 specifically includes the following steps:

[0015] S201, performing frequency band separation and noise suppression on the vibration signal by using a variational mode decomposition and Kalman filter fusion algorithm;

[0016] S202, performing anomaly detection on the temperature data using a lightweight neural network model to identify temperature values ​​that exceed a normal fluctuation range;

[0017] S203: Compress the data packet using a lossless compression algorithm, so that the amount of compressed data does not exceed 40% of the original data.

[0018] Preferably, the step S3 specifically includes the following steps:

[0019] S301, aligning the equipment operation data with the instantaneous values ​​of the environmental parameters in the time domain to generate a multi-dimensional joint feature vector, and injecting the effect data after maintenance and repair;

[0020] S302: Eliminate historical subtrees with an error rate exceeding 25%, generate new subtrees 1.5 times the number of eliminated subtrees, and adjust subtree weights based on the fault recurrence interval. For every 20% reduction in the recurrence interval, the corresponding subtree weight decreases by 10%. Finally, update the operating parameters based on the performance recovery degree. For every 10% increase in the performance recovery degree, the corresponding feature weight increases by 5%.

[0021] S303. The probability of the fault type is calculated by weighting the subtree voting results and the confidence of the preset fault mode library. The remaining life prediction value is generated by quantile aggregation of the regression results of each subtree. The time window is output and matched with the fault mode library to generate component-level positioning results. The maintenance strategy library is called to generate spare parts call instructions and priority sorting.

[0022] Preferably, the S4 specifically includes the following steps:

[0023] S401, quickly matching device type through unique identification code MAC;

[0024] S402, automatically matching the sampling rate according to the device power;

[0025] S403. Migrate high-confidence fault rules from similar devices, and automatically remove weight restrictions based on the prediction error rate to perform progressive model fusion.

[0026] Preferably, the maintenance decision in S5 includes:

[0027] Repair direction decision: If a single, clear fault is predicted, a repair work order is directly triggered and linked to the spare parts inventory;

[0028] Composite maintenance: If the prediction is that multiple faults are superimposed, a combined maintenance plan is generated;

[0029] Preventive maintenance: If an early potential failure is predicted, a manual inspection work order for the equipment potential failure is triggered.

[0030] Preferably, the S6 specifically includes the following steps:

[0031] S601. Collect equipment operation data after maintenance, synchronize environmental parameters and maintenance records;

[0032] S602. Calculate the fault type prediction accuracy and the remaining life prediction error rate, and count the number of false alarms or missed alarms caused by environmental interference;

[0033] S603: Eliminate subtrees with high error rates through incremental random forest, add subtrees based on maintenance data, and assign initial weights.

[0034] Preferably, in said S4, when a new device is connected, the baseline monitoring strategy is automatically loaded, and initial model parameters are generated based on historical similar device data;

[0035] If multiple related devices experience abnormalities at the same time, a global maintenance sequence is generated by analyzing the fault propagation path, and key nodes that affect system safety are prioritized.

[0036] A mine equipment status monitoring and maintenance system based on the Internet of Things, comprising:

[0037] Multi-source data acquisition module, used to synchronously collect equipment operating parameters and environmental parameters through vibration, temperature, pressure and environmental sensors, and dynamically configure sampling rate and time synchronization according to equipment type;

[0038] Edge computing processing module, used to filter, extract features and detect anomalies of sensor data and generate compressed data packets;

[0039] Intelligent analysis and decision-making module, which is used to dynamically update fault prediction rules through an incremental random forest model and output fault type, probability and remaining life prediction results;

[0040] Dynamic expansion access module, used to automatically match the device type library when new devices are added, migrate fault rules for similar devices, and generate an initial monitoring model;

[0041] Maintenance execution scheduling module, which generates maintenance work orders based on equipment priority, optimizes repair paths, and schedules spare parts inventory;

[0042] The closed-loop feedback optimization module is used to collect maintenance effect data and feed it back to the model to iteratively optimize prediction accuracy and maintenance strategies.

[0043] Preferably, the system further comprises:

[0044] The visualization module is used to monitor the equipment's operating status, maintenance work order progress, and fault prediction results in real time through display devices, and provides the following interactive functions:

[0045] 3D digital twin view: Dynamically renders the equipment's operating status based on real-time equipment data and simulation results;

[0046] Prediction results are displayed in layers: equipment anomalies are marked by failure probability and remaining life;

[0047] Maintenance work order tracking: Displays the location of maintenance personnel, spare parts transportation route and estimated completion time.

[0048] The present invention provides a mine equipment status monitoring and maintenance method based on the Internet of Things.

[0049] It has the following beneficial effects:

[0050] 1. The present invention dynamically eliminates high-error historical subtrees and generates new subtrees, adjusts model parameters based on maintenance feedback data, and extracts dynamic characteristics of equipment while retaining historical effective knowledge, ensuring accurate matching of prediction results with maintenance strategies, significantly improving the model's self-learning ability and the reliability of maintenance decisions.

[0051] 2. This invention integrates vibration, temperature, pressure, and environmental sensors, dynamically configures sampling rates, and implements microsecond-level time synchronization protocols to achieve comprehensive acquisition of both equipment operating and environmental parameters. The synchronous integration of environmental and equipment status data enhances signal feature extraction capabilities under complex operating conditions, providing a highly consistent, low-noise, multi-source input foundation for fault prediction.

[0052] 3. The present invention solves the model cold start problem when new equipment is connected through the migration of the equipment's unique identification code and historical rules, realizes the rapid monitoring and integration of new equipment and the coordinated handling of multi-equipment fault chains, ensures the scalability of the system while blocking the spread of risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flow chart of a method for monitoring and maintaining mine equipment status based on the Internet of Things of the present invention;

[0054] Figure 2 This is an architectural diagram of a mine equipment status monitoring and maintenance system based on the Internet of Things of the present invention. DETAILED DESCRIPTION

[0055] The following will clearly and completely describe the technical solution of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0056] Please see the attached Figure 1 The embodiment of the present invention provides a method for monitoring and maintaining the status of mine equipment based on the Internet of Things, comprising:

[0057] S1. Multi-source data acquisition: Vibration, temperature, pressure and environmental sensors deployed on mine equipment are used to synchronously collect equipment operating parameters and environmental parameters;

[0058] S2, edge computing processing: Filtering and anomaly detection of sensor data at the mine edge gateway to generate compressed data packets;

[0059] S3. Online learning and dynamic model updating: This model inputs collected multi-source data and real-time updated maintenance effect data into the incremental random forest model, outputs predictions of failure type, probability, and remaining life, and matches them with a pre-set failure mode library.

[0060] S4, New device access: When a new device is connected to the system, its device type and sensor configuration are automatically extracted, and the monitoring model is dynamically expanded based on the prediction results of S3 to generate initial fault prediction rules for the new device;

[0061] S5. Maintenance decision-making and resource scheduling: Generate maintenance work orders based on model prediction results and equipment priority, and mobilize equipment maintenance parts inventory;

[0062] S6. Closed-loop feedback optimization: Feedback maintenance effect data to the online learning model to iteratively optimize prediction accuracy and maintenance strategies.

[0063] Specifically, S1 deploys vibration, temperature, pressure sensors and environmental sensors (including gas, dust, temperature and humidity sensors) to collect full-dimensional data on the mechanical status and operating environment of mine equipment, and realizes the synchronous collection and time alignment of equipment operating parameters (such as vibration spectrum, motor current harmonics, hydraulic oil pressure fluctuations) and environmental parameters (gas concentration, dust concentration, temperature and humidity) and monitoring, providing a high-precision, multi-dimensional data foundation for subsequent fault prediction and maintenance decisions.

[0064] S2 filters and detects anomalies in sensor data at the mine edge gateway and generates compressed data packets. It uses variational mode decomposition (VMD) and Kalman filtering to suppress noise and enhance signals on the raw data, extracting effective frequency band features. It then performs real-time anomaly detection on the processed data based on the isolation forest algorithm, marking abnormal data segments. Finally, it compresses the filtered and anomaly-marked data using a lossless compression algorithm, retaining key statistical features (such as mean and variance) and anomaly tag information. This achieves denoising of the raw data, real-time anomaly identification, and effective compression of the transmitted data volume, providing high signal-to-noise ratio, lightweight structured data for subsequent analysis.

[0065] S3 dynamically updates model parameters to adapt to changes in equipment operating conditions by inputting multi-source data and real-time maintenance effect data into an incremental random forest model. It calculates the weighted error rate of historical subtrees based on the new input data, eliminates subtrees with an error rate exceeding 25%, and generates new subtrees. It also adjusts feature weights based on maintenance effect data (such as the steady-state operating time of equipment after maintenance and the interval between fault recurrences). It ultimately outputs the fault type, its probability, and the remaining life prediction results, and quickly locates the diagnostic basis by matching the pre-set fault pattern library.

[0066] S4 automatically extracts the unique identification code (MAC / IMEI) and sensor configuration information of newly added devices, matches them to a pre-set device type library to obtain standard monitoring parameters, and dynamically generates initial fault prediction rules based on historical data from similar devices. It calls S3's incremental random forest model, migrates high-confidence fault rules (confidence ≥ 85%) from similar devices, assigns initial model weights based on the new device's sensor type (such as vibration, temperature) and installation location (mining face, transportation line), and simultaneously loads the baseline monitoring strategy to enable instant expansion of the monitoring model after the new device is connected, avoiding cold start issues.

[0067] S5 automatically generates maintenance work orders and associates them with maintenance resources based on the model's output of failure type, probability, and remaining life predictions. This enables efficient dispatch of maintenance tasks and precise resource scheduling, shortening equipment maintenance response time.

[0068] S6 drives the continuous optimization of models and strategies by feeding maintenance effect data into the online learning model in real time. After cleaning, the maintenance data is input into the incremental random forest model, which analyzes the source of prediction deviation based on the confusion matrix and dynamically adjusts the subtree weights and feature importance distribution. At the same time, the maintenance strategy rule base is iteratively updated according to the maintenance results to achieve adaptive collaborative optimization of the prediction model and maintenance strategy.

[0069] Environmental sensors in S1 include non-dispersive infrared gas sensor, laser scattering dust sensor and digital temperature and humidity sensor;

[0070] The collection frequency of equipment operating parameters is dynamically adjusted according to the equipment type, including a mining machine vibration signal sampling rate ≥10kHz and a conveyor motor current signal sampling rate ≥2kHz. All sensor data is synchronized at the microsecond level through the IEEE 1588PTP protocol.

[0071] Specifically, the non-dispersive infrared gas sensor detects methane concentration based on the absorption characteristics of the infrared spectrum, the laser scattering dust sensor analyzes the concentration of suspended particulate matter in the air through the principle of light scattering, and the digital temperature and humidity sensor uses a capacitive sensing unit to synchronously measure ambient temperature and humidity. Through the collaborative work of multiple types of environmental sensors, the gas, dust, temperature and humidity parameters in the mine are collected in real time, constructing a multi-dimensional feature input of the equipment operating environment. The synchronous collection and fusion of data from environmental sensors and equipment status sensors (vibration, temperature, pressure) provides an environmental correlation analysis basis for fault prediction models, effectively enhancing the robustness and prediction accuracy of equipment status monitoring under complex working conditions.

[0072] By dynamically adjusting sensor sampling frequencies based on equipment type, for example, mining machine vibration signals use a high sampling rate of ≥10kHz to capture high-frequency impact characteristics, while conveyor motor current signals use a sampling rate of ≥2kHz to analyze harmonic distortion and load fluctuations. This ensures complete acquisition of the operating characteristics of different equipment. Furthermore, all sensor data is synchronized at the microsecond level (deviation ≤1μs) using the IEEE 1588PTP protocol, eliminating time misalignment issues with multi-source data.

[0073] S2 specifically includes the following steps:

[0074] S201, performing frequency band separation and noise suppression on the vibration signal by using a variational mode decomposition and Kalman filter fusion algorithm;

[0075] S202, performing anomaly detection on the temperature data using a lightweight neural network model to identify temperature values ​​that exceed a normal fluctuation range;

[0076] S203: Compress the data packet using a lossless compression algorithm, so that the amount of compressed data does not exceed 40% of the original data.

[0077] Specifically, S201 uses a fusion algorithm of variational mode decomposition (VMD, with 8 decomposition layers) and Kalman filtering to perform frequency band separation and noise suppression on vibration signals. VMD decomposes the vibration signal into preset frequency bands, separating the characteristic frequency bands of equipment operation. Kalman filtering dynamically adjusts the process noise covariance matrix to suppress high-frequency noise, thereby effectively extracting key features from the vibration signal (such as the bearing spalling impact waveform) and eliminating noise interference from the complex underground environment, providing a high signal-to-noise ratio signal input for subsequent anomaly detection.

[0078] S202 uses a lightweight neural network model to perform real-time anomaly detection on temperature data. The model input is temperature time series data, and the output is an anomaly probability value. By setting a dynamic threshold to identify temperature anomalies that exceed the normal fluctuation range (such as motor overheating and cooling failure), real-time detection of temperature anomalies is achieved.

[0079] S203 compresses the filtered and anomaly-marked data through the LZW lossless compression algorithm, retaining statistical features (mean, variance, peak-to-peak value) and anomaly marking information, and the piecewise linear fitting algorithm prioritizes retaining the original sampling rate of the abnormal data segment. On the premise of ensuring data integrity, the compressed data supports direct cloud-based parsing and feature extraction without the need for decompression preprocessing.

[0080] S3 specifically includes the following steps:

[0081] S301, aligning the equipment operation data with the instantaneous values ​​of the environmental parameters in the time domain to generate a multi-dimensional joint feature vector, and injecting the effect data after maintenance and repair;

[0082] S302: Eliminate historical subtrees with an error rate exceeding 25%, generate new subtrees 1.5 times the number of eliminated subtrees, and adjust subtree weights based on the fault recurrence interval. For every 20% reduction in the recurrence interval, the corresponding subtree weight decreases by 10%. Finally, update the operating parameters based on the performance recovery degree. For every 10% increase in the performance recovery degree, the corresponding feature weight increases by 5%.

[0083] S303. The probability of the fault type is calculated by weighting the subtree voting results and the confidence of the preset fault mode library. The remaining life prediction value is generated by quantile aggregation of the regression results of each subtree. The time window is output and matched with the fault mode library to generate component-level positioning results. The maintenance strategy library is called to generate spare parts call instructions and priority sorting.

[0084] Specifically, S301 aligns equipment operation data with environmental parameters in the time domain and integrates maintenance feedback data to construct a multi-dimensional joint feature vector. This solves the problem of feature fragmentation caused by differences in sampling frequency in multi-source data. The model's input feature representation capability is significantly enhanced, capturing the dynamic coupling relationship between equipment status and environmental changes. This improves the robustness of fault feature extraction under complex working conditions, thereby reducing the misjudgment rate caused by asynchronous or isolated data analysis.

[0085] S302 achieves simultaneous optimization of the model structure and equipment status evolution by eliminating historical subtrees with high errors, dynamically generating new subtrees, and adjusting subtree weights. By reducing the weight of corresponding subtrees based on the shortening ratio of the fault recurrence interval, it suppresses the interference of inefficient rules on the prediction results. At the same time, it increases the weight of key features based on the improvement in performance recovery, strengthening the model's ability to learn effective maintenance strategies.

[0086] S303 calculates the failure probability by fusing the subtree voting results with the confidence of the preset fault mode library, and generates a remaining life prediction window based on the quantile aggregation of the subtree regression results. It quantifies the uncertainty of the prediction results, achieves component-level positioning by matching the fault mode library, and calls the maintenance strategy library to generate spare parts call instructions and priority sorting. The prediction results are directly mapped into executable maintenance decisions, realizing closed-loop support from fault prediction to maintenance actions, significantly shortening the time spent on fault diagnosis and maintenance response, and improving operation and maintenance efficiency.

[0087] S4 specifically includes the following steps:

[0088] S401, quickly matching device type through unique identification code MAC;

[0089] S402, automatically matching the sampling rate according to the device power;

[0090] S403. Migrate high-confidence fault rules from similar devices, and automatically remove weight restrictions based on the prediction error rate to perform progressive model fusion.

[0091] Specifically, S401 uses the unique device identification code (such as MAC address) to quickly match the pre-set device type library and automatically identify the type of new equipment (such as mining machine, conveyor, ventilator). Based on the mapping relationship between the identification code and the device type library, the corresponding standard sensor configuration mode is called to achieve zero manual intervention in type identification when new equipment is connected, avoiding monitoring deviations caused by manual configuration errors.

[0092] S402 automatically matches the sampling rate rules based on the device power to ensure the complete acquisition of high-frequency impact signals (such as the transient waveform of a bearing fault) while avoiding oversampling of low-power devices (reducing the edge computing load).

[0093] S403 migrates fault rules with a confidence level ≥ 85% from historical data of similar devices, generates an initial fault prediction rule base for new devices, and assigns initial weights to independent subtrees. When the operating data of the new device accumulates to 24 hours and the prediction error rate is less than 5%, the weight limit is automatically lifted and gradually increased to the global model weight, solving the problem of lack of cold start data for new devices. The model fusion process also smoothly transitions, avoiding the risk of misjudgment caused by sudden changes in prediction results.

[0094] Maintenance decisions in S5 include:

[0095] Repair direction decision: If a single, clear fault is predicted, a repair work order is directly triggered and linked to the spare parts inventory;

[0096] Composite maintenance: If the prediction is that multiple faults are superimposed, a combined maintenance plan is generated;

[0097] Preventive maintenance: If an early potential failure is predicted, a manual inspection work order for the equipment potential failure is triggered.

[0098] Specifically, through a hierarchical decision-making mechanism, maintenance strategies are automatically matched based on fault prediction results: for a single clear fault (such as bearing damage), a maintenance work order is directly generated and linked to the spare parts inventory to achieve rapid response and accurate resource invocation; for multiple fault superposition scenarios (such as the coexistence of seal leakage and motor overheating), a combined maintenance plan is generated to systematically eliminate hidden dangers; for early potential faults (such as oil contamination warnings), manual inspection work orders are triggered for early intervention, realizing intelligent adaptation of maintenance strategies and efficient linkage of resources, ensuring immediate handling of high-priority faults, collaborative repair of complex faults, and active prevention and control of potential risks, comprehensively optimizing maintenance process efficiency and equipment reliability.

[0099] S6 specifically includes the following steps:

[0100] S601. Collect equipment operation data after maintenance, synchronize environmental parameters and maintenance records;

[0101] S602. Calculate the fault type prediction accuracy and the remaining life prediction error rate, and count the number of false alarms or missed alarms caused by environmental interference;

[0102] S603: Eliminate subtrees with high error rates through incremental random forest, add subtrees based on maintenance data, and assign initial weights.

[0103] Specifically, S601 collects the operating data of the equipment after maintenance, synchronizes environmental parameters and maintenance records, and builds a complete maintenance effect evaluation data set. This enables full-dimensional data correlation between equipment status, environmental changes, and maintenance operations. This provides a traceable, multi-factor coupled analysis foundation for subsequent model optimization, ensuring the timeliness and integrity of feedback data.

[0104] S602 calculates the fault type prediction accuracy, the remaining life prediction error rate, and counts the number of false positives or missed negatives caused by environmental interference to quantify the model's prediction performance. This identifies the model's prediction blind spots for specific fault types (such as seal leakage) or complex environmental scenarios, locates the root cause of the error, and provides a quantitative basis for targeted optimization.

[0105] S603 dynamically eliminates subtrees with an error rate exceeding 25% through an incremental random forest model, and generates new subtrees based on maintenance data (the number is 1.5 times the number of eliminated subtrees). It assigns initial weights (0.2-0.5) to the new subtrees to balance the contribution of new and old knowledge, enabling the model to quickly adapt to maintenance feedback, strengthen the learning weights of verified fault rules, and suppress the interference of inefficient subtrees on prediction results.

[0106] In S4, when a new device is connected, the baseline monitoring strategy is automatically loaded and the initial model parameters are generated based on historical data of similar devices;

[0107] If multiple related devices experience abnormalities at the same time, a global maintenance sequence is generated by analyzing the fault propagation path, and key nodes that affect system safety are prioritized.

[0108] Specifically, by automatically loading the baseline monitoring strategy (such as the default sampling rate and alarm threshold) when a new device is connected and generating initial model parameters based on the high-confidence fault rules migrated from historical similar device data, the cold start problem of new device monitoring is solved, ensuring that it is immediately included in the system monitoring scope. At the same time, when multiple related devices (such as motors and reducers in the conveyor drive chain) have abnormalities at the same time, the key nodes (such as failure of the main motor will cause the entire line to shut down) are identified through fault propagation path analysis (such as cause-and-effect diagrams and fault trees), a global maintenance sequence is generated, and high-impact nodes are prioritized, achieving zero-delay access to new equipment monitoring and coordinated maintenance of complex fault chains, avoiding the spread of systemic risks.

[0109] Please see the attached Figure 2 , a mine equipment status monitoring and maintenance system based on the Internet of Things, including:

[0110] Multi-source data acquisition module, used to synchronously collect equipment operating parameters and environmental parameters through vibration, temperature, pressure and environmental sensors, and dynamically configure sampling rate and time synchronization according to equipment type;

[0111] Edge computing processing module, used to filter, extract features and detect anomalies of sensor data and generate compressed data packets;

[0112] Intelligent analysis and decision-making module, which is used to dynamically update fault prediction rules through an incremental random forest model and output fault type, probability and remaining life prediction results;

[0113] Dynamic expansion access module, used to automatically match the device type library when new devices are added, migrate fault rules for similar devices, and generate an initial monitoring model;

[0114] Maintenance execution scheduling module, which generates maintenance work orders based on equipment priority, optimizes repair paths, and schedules spare parts inventory;

[0115] The closed-loop feedback optimization module is used to collect maintenance effect data and feed it back to the model to iteratively optimize prediction accuracy and maintenance strategies.

[0116] Specifically, the multi-source data acquisition module enables simultaneous, high-precision collection of equipment operating and environmental parameters. Combined with the real-time filtering and anomaly detection of the edge computing processing module, this provides lightweight, high-signal-to-noise ratio data input for the cloud-based intelligent analysis and decision-making module. The dynamic expansion access module supports rapid access to new equipment and initial model construction, eliminating cold start issues. The maintenance execution scheduling module generates optimal maintenance plans based on model predictions, accurately scheduling resources. The closed-loop feedback optimization module continuously iterates models and strategies based on maintenance effect data. Through multi-module collaboration, full-lifecycle monitoring of equipment status, real-time fault prediction, and dynamic optimization of maintenance strategies are achieved.

[0117] The system also includes:

[0118] The visualization module is used to monitor the equipment's operating status, maintenance work order progress, and fault prediction results in real time through display devices, and provides the following interactive functions:

[0119] 3D digital twin view: Dynamically renders the equipment's operating status based on real-time equipment data and simulation results;

[0120] Prediction results are displayed in layers: equipment anomalies are marked by failure probability and remaining life;

[0121] Maintenance work order tracking: Displays the location of maintenance personnel, spare parts transportation route and estimated completion time.

[0122] Specifically, the three-dimensional digital twin view is used to render the equipment operating status in real time, and the fault prediction results and the remaining life progress bar are superimposed to intuitively present the equipment health status. The layered display function of the prediction results marks abnormal equipment according to the failure probability and remaining life. The maintenance work order tracking function dynamically displays the location of the maintenance personnel (UWB positioning), the spare parts transportation route and the estimated completion time, and superimposes the equipment disassembly instructions through the AR interface to achieve multi-dimensional visual interaction of equipment status, maintenance process and prediction results, significantly improving the maintenance decision-making efficiency in complex mine environments, reducing the intensity of manual inspections, and supporting collaborative operations of multiple terminals (AR glasses, mobile terminals, PC billboards).

[0123] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A mine equipment status monitoring and maintenance method based on the Internet of Things, characterized in that: include: S1. Multi-source data acquisition: Vibration, temperature, pressure and environmental sensors deployed on mine equipment are used to synchronously collect equipment operating parameters and environmental parameters; S2, edge computing processing: Filtering and anomaly detection of sensor data at the mine edge gateway to generate compressed data packets; S3. Online learning and dynamic model updating: This model inputs collected multi-source data and real-time updated maintenance effect data into the incremental random forest model, outputs predictions of failure type, probability, and remaining life, and matches them with a pre-set failure mode library. S4, New device access: When a new device is connected to the system, its device type and sensor configuration are automatically extracted, and the monitoring model is dynamically expanded based on the prediction results of S3 to generate initial fault prediction rules for the new device; S5. Maintenance decision-making and resource scheduling: Generate maintenance work orders based on model prediction results and equipment priority, and mobilize equipment maintenance parts inventory; S6. Closed-loop feedback optimization: Feedback maintenance effect data to the online learning model to iteratively optimize prediction accuracy and maintenance strategies.

2. The method for monitoring and maintaining mine equipment status based on the Internet of Things according to claim 1, characterized in that: The environmental sensors in S1 include a non-dispersive infrared gas sensor, a laser scattering dust sensor, and a digital temperature and humidity sensor; The acquisition frequency of the equipment operating parameters is dynamically adjusted according to the equipment type, including a mining machine vibration signal sampling rate ≥10kHz, a conveyor motor current signal sampling rate ≥2kHz, and all sensor data are synchronized at the microsecond level through the IEEE 1588PTP protocol.

3. The method for monitoring and maintaining mine equipment status based on the Internet of Things according to claim 1, characterized in that: The S2 specifically includes the following steps: S201, performing frequency band separation and noise suppression on the vibration signal by using a variational mode decomposition and Kalman filter fusion algorithm; S202, performing anomaly detection on the temperature data using a lightweight neural network model to identify temperature values ​​that exceed a normal fluctuation range; S203: Compress the data packet using a lossless compression algorithm, so that the amount of compressed data does not exceed 40% of the original data.

4. The method for monitoring and maintaining mine equipment status based on the Internet of Things according to claim 1, characterized in that: The S3 specifically includes the following steps: S301, aligning the equipment operation data with the instantaneous values ​​of the environmental parameters in the time domain to generate a multi-dimensional joint feature vector, and injecting the effect data after maintenance and repair; S302: Eliminate historical subtrees with an error rate exceeding 25%, generate new subtrees 1.5 times the number of eliminated subtrees, and adjust subtree weights based on the fault recurrence interval. For every 20% reduction in the recurrence interval, the corresponding subtree weight decreases by 10%. Finally, update the operating parameters based on the performance recovery degree. For every 10% increase in the performance recovery degree, the corresponding performance feature weight increases by 5%. S303. The probability of the fault type is calculated by weighting the subtree voting results and the confidence of the preset fault mode library. The remaining life prediction value is generated by quantile aggregation of the regression results of each subtree. The time window is output and matched with the fault mode library to generate component-level positioning results. The maintenance strategy library is called to generate spare parts call instructions and priority sorting.

5. The method for monitoring and maintaining mine equipment status based on the Internet of Things according to claim 1, characterized in that: The S4 specifically includes the following steps: S401, quickly matching device type through unique identification code MAC; S402, automatically matching the sampling rate according to the device power; S403. Migrate high-confidence fault rules from similar devices, and automatically remove weight restrictions based on the prediction error rate to perform progressive model fusion.

6. The method for monitoring and maintaining mine equipment status based on the Internet of Things according to claim 1, characterized in that: The maintenance decision in S5 includes: Repair direction decision: If a single, clear fault is predicted, a repair work order is directly triggered and linked to the spare parts inventory; Composite maintenance: If the prediction is that multiple faults are superimposed, a combined maintenance plan is generated; Preventive maintenance: If an early potential failure is predicted, a manual inspection work order for the equipment potential failure is triggered.

7. The method for monitoring and maintaining mine equipment status based on the Internet of Things according to claim 1, characterized in that: The S6 specifically includes the following steps: S601. Collect equipment operation data after maintenance, synchronize environmental parameters and maintenance records; S602. Calculate the fault type prediction accuracy and the remaining life prediction error rate, and count the number of false alarms or missed alarms caused by environmental interference; S603: Eliminate subtrees with high error rates through incremental random forest, add subtrees based on maintenance data, and assign initial weights.

8. The method for monitoring and maintaining mine equipment status based on the Internet of Things according to claim 1, characterized in that: In S4, when a new device is added, the baseline monitoring strategy is automatically loaded and initial model parameters are generated based on historical similar device data; If multiple related devices experience abnormalities at the same time, a global maintenance sequence is generated by analyzing the fault propagation path, and key nodes that affect system safety are prioritized.

9. A mine equipment status monitoring and maintenance system based on the Internet of Things, characterized in that: A method for monitoring and maintaining the condition of mine equipment based on the Internet of Things as described in any one of claims 1 to 8, comprising: Multi-source data acquisition module, used to synchronously collect equipment operating parameters and environmental parameters through vibration, temperature, pressure and environmental sensors, and dynamically configure sampling rate and time synchronization according to equipment type; Edge computing processing module, used to filter, extract features and detect anomalies of sensor data and generate compressed data packets; Intelligent analysis and decision-making module, which is used to dynamically update fault prediction rules through an incremental random forest model and output fault type, probability and remaining life prediction results; Dynamic expansion access module, used to automatically match the device type library when new devices are added, migrate fault rules for similar devices, and generate an initial monitoring model; Maintenance execution scheduling module, which generates maintenance work orders based on equipment priority, optimizes repair paths, and schedules spare parts inventory; The closed-loop feedback optimization module is used to collect maintenance effect data and feed it back to the model to iteratively optimize prediction accuracy and maintenance strategies.

10. The mine equipment status monitoring and maintenance system based on the Internet of Things according to claim 9, characterized in that: The system further comprises: The visualization module is used to monitor the equipment's operating status, maintenance work order progress, and fault prediction results in real time through display devices, and provides the following interactive functions: 3D digital twin view: Dynamically renders the equipment's operating status based on real-time equipment data and simulation results; Prediction results are displayed in layers: equipment anomalies are marked by failure probability and remaining life; Maintenance work order tracking: Displays the location of maintenance personnel, spare parts transportation route and estimated completion time.

Images