System and method for comprehensive monitoring and predictive maintenance of power equipment based on digital twinning

Abstract

The invention provides a digital twinning-based power equipment comprehensive monitoring and predictive maintenance system and method. The system comprises a physical layer, a data acquisition and transmission layer, a digital twinning model layer, a data analysis and intelligent early warning layer and a decision support and visualization layer. The method comprises the steps of data acquisition, model driving, simulation, calculation evaluation, early warning and feedback optimization. According to the method, equipment faults are predicted in advance by means of a digital twinning technology and a machine learning algorithm. The system can accurately analyze the health state of the equipment and predict the remaining service life of key parts by continuously monitoring equipment operation parameters such as vibration, temperature and pressure and combining historical data and an equipment operation mechanism. Once the health index of the equipment is lower than the preset threshold value or the remaining service life is close to the safety margin, the system immediately and automatically gives an early warning, and maintenance personnel can arrange maintenance work in advance, so that unplanned shutdown caused by sudden failure of the equipment is avoided, and the production continuity is ensured.

Description

Technical Field

[0001] This invention belongs to the field of industrial automation and equipment monitoring technology, specifically relating to a comprehensive monitoring and predictive maintenance system and method for power equipment based on digital twins. Background Technology

[0002] Traditional power monitoring systems (such as those monitoring air compressors, water pumps, fans, and central air conditioning systems) primarily rely on SCADA (Supervisory Control and Data Acquisition) systems for real-time data acquisition and alarms. While these systems can display the current operating status of the equipment, they suffer from three main drawbacks: passive response, information silos, and reliance on experience.

[0003] Passive response systems can only generate alarms after a fault occurs or parameters exceed limits, failing to anticipate faults and leading to unplanned downtime and production disruptions. Information silos, where monitoring data, maintenance records, and performance data are independent and lack effective correlation and analysis, make it difficult to extract value from historical data. For example, equipment temperature monitoring data is stored in a SCADA system, while maintenance records are manually entered into Excel spreadsheets, making it impossible to correlate historical temperature trends with bearing wear and missing opportunities for fault prediction. Reliance on experience means maintenance decisions heavily depend on the personal experience of maintenance personnel, lacking scientific data support and easily leading to over-maintenance or under-maintenance. Some companies, to avoid faults, mandate monthly bearing replacements for air compressors, even though some bearings could still operate normally for 3-6 months, resulting in wasted spare parts costs. Summary of the Invention

[0004] The purpose of this invention is to provide a digital twin-based integrated monitoring and predictive maintenance system and method for power equipment that solves the problems of passive response, information silos, and reliance on experience in traditional systems.

[0005] A digital twin-based integrated monitoring and predictive maintenance system for power equipment includes a physical layer, a data acquisition and transmission layer, a digital twin model layer, a data analysis and intelligent early warning layer, and a decision support and visualization layer. The physical layer includes various power equipment and sensors installed on the equipment, used to collect real-time operating data of the equipment; The data acquisition and transmission layer collects sensor data through IoT gateways, performs protocol conversion and edge computing preprocessing, and then uploads the data to the cloud platform or central server. The digital twin model layer constructs a high-fidelity digital twin model in virtual space that corresponds one-to-one with the physical device. The model integrates the device's geometric model, physical characteristics, operating mechanism, and historical data. The data analysis and intelligent early warning layer includes a health status assessment module, a fault prediction module, and an adaptive alarm module. The decision support and visualization layer provides a visual monitoring interface that displays the device's 3D model, real-time data, health status, and prediction results, and automatically generates maintenance work orders and maintenance suggestions.

[0006] Furthermore, the health status assessment module scores the device's health status based on real-time and historical data, using machine learning algorithms combined with preset formulas.

[0007] Furthermore, the preset formula used by the health status assessment module is: in, For parameter weights, ; These are real-time parameter values; , These are the upper and lower limits of the normal range for the parameters; The degradation coefficient is fitted using historical failure data; The equipment's operating time is measured in months. ∈[0,1].

[0008] Furthermore, the fault prediction module analyzes the degradation trend of equipment operating parameters, combines mechanistic models and data-driven models, and uses preset formulas to predict the remaining service life of key components.

[0009] Furthermore, the preset formula used by the fault prediction module is: in, This represents the maximum permissible wear of the component. The current wear amount is inferred from vibration or temperature data. The wear rate per unit time predicted by the LSTM model is in mm / day; the RUL result is rounded to the nearest integer and is in days.

[0010] Furthermore, the adaptive alarm module dynamically adjusts the alarm threshold based on the health score and fault prediction results.

[0011] Furthermore, the power equipment includes at least one of an air compressor, a water pump, a fan, and a central air conditioner; the sensor includes at least one of a vibration sensor, a temperature sensor, a pressure sensor, and a current sensor.

[0012] A method for integrated monitoring and predictive maintenance of power equipment based on digital twins, based on the integrated monitoring and predictive maintenance system for power equipment based on digital twins as described in any one of claims 1-7, includes the following steps: S1: Real-time acquisition of diverse operational data from power equipment; S2: Inject data into the corresponding digital twin model to drive the model to run synchronously with the physical device; S3: Use digital twin models for simulation to simulate the operating status of equipment under different working conditions; S4: Based on simulation results and real-time data, the health index of the equipment and the remaining service life of key components are calculated by combining the health index calculation formula and the remaining service life prediction formula with the machine learning model. S5: When the health index is lower than the preset threshold or the remaining service life is lower than the safety margin, the system automatically generates early warning information and maintenance suggestions; S6: After performing maintenance, maintenance personnel will feed the maintenance records back to the system for optimizing and calibrating parameters in the digital twin model and prediction algorithm.

[0013] Furthermore, the preset threshold in S5 is 0.6; the safety margin is 7 days.

[0014] Furthermore, the parameters to be optimized and calibrated in S6 include those in the preset formula used by the health status assessment module. , k, and the remaining lifetime formula .

[0015] The beneficial effects of this invention are as follows: By leveraging digital twin technology and machine learning algorithms, this invention can predict equipment failures in advance. Through continuous monitoring of equipment operating parameters, such as vibration, temperature, and pressure, combined with historical data and equipment operating mechanisms, the system can accurately analyze the health status of the equipment and predict the remaining service life of key components. Once the equipment health index is found to be below a preset threshold or its remaining service life is nearing its safety margin, the system immediately and automatically issues an early warning. Maintenance personnel can then schedule maintenance work in advance, avoiding unplanned downtime caused by sudden equipment failures and ensuring production continuity.

[0016] This invention breaks away from the traditional system's separation of monitoring data, maintenance records, and performance data, achieving deep data fusion and correlation analysis. The digital twin model integrates the equipment's geometric model, physical characteristics, operating mechanisms, and historical data, providing a comprehensive and accurate data foundation for data analysis. Based on this data, through quantitative formulas and data-driven models, it provides a scientific basis for maintenance decisions, moving away from the previous reliance on individual maintenance personnel's experience. This effectively avoids over-maintenance or under-maintenance, achieving precise maintenance and reducing maintenance costs.

[0017] This invention's system possesses self-learning and self-evolution capabilities, thanks to its unique closed-loop optimization mechanism. After performing maintenance work, maintenance personnel promptly feed back maintenance records to the system. The system uses this feedback data to optimize and calibrate the parameters and prediction algorithms of the digital twin model. As maintenance data accumulates, the model and algorithm can continuously learn and adapt to changes in equipment operation, gradually improving prediction accuracy and continuously optimizing system performance to better meet the needs of power equipment monitoring and maintenance. Attached Figure Description

[0018] Figure 1 This is a diagram of the overall system architecture of the present invention; Figure 2 This is a flowchart of the predictive maintenance process of the present invention. Detailed Implementation

[0019] The present invention will now be further described with reference to the accompanying drawings.

[0020] like Figures 1-2 As shown, a digital twin-based integrated monitoring and predictive maintenance system for power equipment includes a physical layer, a data acquisition and transmission layer, a digital twin model layer, a data analysis and intelligent early warning layer, and a decision support and visualization layer. The physical layer consists of various power equipment (such as air compressors and water pumps) and sensors installed on the equipment (such as vibration sensors, temperature sensors, pressure sensors, and current sensors) to collect real-time operating data of the equipment. Taking an air compressor as an example, a vibration sensor is installed at the bearing housing to collect vibration acceleration data (sampling frequency 1000Hz); a temperature sensor is embedded in the motor winding to collect winding temperature; and a pressure sensor is installed at the air outlet to collect exhaust pressure.

[0021] The data acquisition and transmission layer collects sensor data through IoT gateways, performs protocol conversion (such as converting the sensor's RS485 protocol to MQTT protocol) and edge computing preprocessing (such as removing outliers and downsampling data), and then uploads the data to the cloud platform or central server. Edge computing can complete preliminary data processing locally on the gateway, reducing the amount of data uploaded and lowering network bandwidth usage.

[0022] The digital twin model layer is a high-fidelity digital twin model that corresponds one-to-one with the physical equipment in virtual space. This model not only includes the geometric model of the equipment (such as the three-dimensional structure of the cylinder, bearings, impeller and other components of an air compressor, with an accuracy of 0.1mm), but also integrates its physical properties (such as the elastic modulus and thermal conductivity of materials), operating mechanisms (such as the thermodynamic process of compressed air), and historical data (operating parameters, fault records and maintenance records for the past 3 years).

[0023] The data analysis and intelligent early warning layer includes a health status assessment module, a fault prediction module, and an adaptive alarm module. The health status assessment module uses machine learning algorithms (such as support vector machines and deep learning) to score the health status of devices based on real-time and historical data. The score is calculated using the following formula: in, For parameter weights (e.g., vibration acceleration weight 0.3, temperature weight 0.2), use these weights. ; These are real-time parameter values; , These are the upper and lower limits of the normal range for parameters (e.g., the normal range for air compressor bearing temperature is 30-80℃). The degradation coefficient is fitted using historical fault data (e.g., k=0.05 for bearing wear). For equipment operating time (monthly); ∈[0, 1], the closer to 1, the better the health status. When HI is 0.9, the equipment is in an excellent health status; when HI is 0.6, the equipment is in a critical health status.

[0024] The fault prediction module analyzes the degradation trend of equipment operating parameters and combines mechanistic models (such as bearing wear mechanism models) and data-driven models (such as LSTM models) to predict the remaining useful life (RUL) of key components (such as bearings and impellers). The prediction calculation uses the following formula: in, This represents the maximum permissible wear of the component (e.g., the bearing wear limit of 0.1 mm). The current wear amount is inferred from vibration or temperature data (inferred from the peak factor of vibration data; for example, if the peak factor is greater than 5, the wear amount is inferred to be 0.05 mm). The wear rate per unit time predicted by the LSTM model (e.g., 0.002 mm / day); if the RUL result is an integer (days) and falls below the safety margin (e.g., 7 days), an early warning is triggered.

[0025] Adaptive alarm module: Dynamically adjusts alarm thresholds based on health scores and fault prediction results. For example, when the device's HI is 0.9, the temperature alarm threshold is set to 80℃; when the HI drops to 0.7, the temperature alarm threshold is lowered to 70℃, realizing the transformation from "fixed threshold alarm" to "state-based intelligent early warning".

[0026] The decision support and visualization layer provides users with a visual monitoring interface, displaying a 3D model of the equipment (which can be rotated 360° for viewing) and real-time data (presented in dashboard format, such as vibration acceleration of 2.5 mm / s²).2 The system automatically generates maintenance work orders (including maintenance parts, maintenance time, and required spare parts) and maintenance suggestions (such as "It is recommended to replace the bearing in 15 days, and the machine needs to be shut down for 2 hours before replacement"). The system also displays the temperature (55℃), health status (marked by color, green for excellent, yellow for critical, and red for poor), prediction results (e.g., "Bearing life remaining 12 days").

[0027] A method for integrated monitoring and predictive maintenance of power equipment based on digital twins includes the following steps: S1: Real-time acquisition of multi-dimensional operating data of power equipment; the acquisition frequency is adjusted according to the parameter type, vibration data is acquired once every 10 seconds, and temperature and pressure data are acquired once every 1 minute; S2: Inject data into the corresponding digital twin model to drive the model to run synchronously with the physical device; transmit data to the virtual model in real time through a data interface (such as an API interface) to drive the model to run synchronously with the physical device, with the synchronization delay controlled within 1 second; S3: Use digital twin models for simulation to simulate the operating status of equipment under different working conditions; for example, simulate the energy consumption and temperature change trend of the equipment when the intake pressure is 0.8MPa and the ambient temperature is 35℃. S4: Based on simulation results and real-time data, the health index of the equipment and the remaining service life of key components are calculated by combining the health index calculation formula and the remaining service life prediction formula with the machine learning model. Based on simulation results and real-time data, the health index and remaining lifespan of key components of the equipment are calculated every 5 minutes using machine learning models (such as support vector machines and LSTM) combined with the health index calculation formula and the remaining lifespan prediction formula. S5: When the health index is lower than the preset threshold or the remaining service life is lower than the safety margin, the system automatically generates early warning information and maintenance suggestions; When the health index is lower than the preset threshold (e.g., 0.6) or the remaining service life is lower than the safety margin (e.g., 7 days), the system will automatically generate warning information (e.g., "Health index of air compressor bearing #1 is 0.58, remaining service life is 6 days, maintenance needs to be arranged as soon as possible") and maintenance suggestions via SMS and platform pop-up window; S6: After performing maintenance, maintenance personnel will feed the maintenance records back to the system for optimizing and calibrating parameters in the digital twin model and prediction algorithm.

[0028] After maintenance personnel perform maintenance, they will send maintenance records (such as maintenance time, replacement part model, and post-maintenance equipment parameters) back to the system via a mobile app. The system will then use this data to optimize the physical parameters of the digital twin model (such as adjusting the bearing friction coefficient) and calibrate the parameters in the prediction algorithm to improve the accuracy of subsequent predictions.

[0029] Taking predictive maintenance of air compressors as an example, the present invention will be described in detail. Vibration sensors (to collect bearing vibration data), temperature sensors (to collect motor temperature), pressure sensors (to collect exhaust pressure), and current sensors (to collect operating current) are installed on the air compressor to form the physical layer. An IoT gateway is configured as the data acquisition and transmission layer. A digital twin model of the air compressor is built (integrating the air compressor's geometry, compressed air generation mechanism, and operating data from the past 3 years). A data analysis and intelligent early warning layer is deployed (with vibration acceleration weight set to 0.3, temperature weight to 0.2, pressure weight to 0.3, and current weight to 0.2, and a degradation coefficient k=0.02 is fitted using historical fault data) and a decision support and visualization layer are also deployed.

[0030] Sensors collect real-time data on the air compressor's vibration, temperature, pressure, and current. This data is processed and injected into a digital twin model, driving the model to operate synchronously with the air compressor. The digital twin model simulates the air compressor's operation under load rates of 80%, 90%, and 100%, obtaining parameter variation trends under different conditions. Combining real-time data and simulation results, a health index HI≈0.55 is calculated using a support vector machine model. An LSTM model is used to predict the bearing's wear rate per unit time. =0.002mm / day, given the maximum allowable wear of the bearing. =0.1mm, current wear amount =0.08mm, calculate the remaining service life RUL=10 days; the health index 0.55 is lower than the preset threshold of 0.6, the system automatically generates a warning message, prompting "the air compressor is in poor health condition, it is recommended to pay close attention to the bearing condition"; the maintenance personnel inspect the air compressor bearing, find slight wear and replace it, and feed the maintenance record back to the system, the system calibrates the deterioration coefficient k=0.018, and optimizes the accuracy of subsequent health index calculations.

Claims

1. A comprehensive monitoring and predictive maintenance system for power equipment based on digital twins, characterized in that, It includes the physical layer, data acquisition and transmission layer, digital twin model layer, data analysis and intelligent early warning layer, and decision support and visualization layer; The physical layer includes various power equipment and sensors installed on the equipment, used to collect real-time operating data of the equipment; The data acquisition and transmission layer collects sensor data through IoT gateways, performs protocol conversion and edge computing preprocessing, and then uploads the data to the cloud platform or central server. The digital twin model layer constructs a high-fidelity digital twin model in virtual space that corresponds one-to-one with the physical device. The model integrates the device's geometric model, physical characteristics, operating mechanism, and historical data. The data analysis and intelligent early warning layer includes a health status assessment module, a fault prediction module, and an adaptive alarm module. The decision support and visualization layer provides a visual monitoring interface that displays the device's 3D model, real-time data, health status, and prediction results, and automatically generates maintenance work orders and maintenance suggestions.

2. The integrated monitoring and predictive maintenance system for power equipment based on digital twins according to claim 1, characterized in that, The health status assessment module scores the equipment's health status based on real-time and historical data, using machine learning algorithms combined with preset formulas.

3. The integrated monitoring and predictive maintenance system for power equipment based on digital twins according to claim 2, characterized in that, The preset formula used by the health status assessment module is: in, For parameter weights, ; These are real-time parameter values; , These are the upper and lower limits of the normal range for the parameters; The degradation coefficient is fitted using historical failure data; The equipment's operating time is measured in months. ∈[0,1].

4. The integrated monitoring and predictive maintenance system for power equipment based on digital twins according to claim 1, characterized in that, The fault prediction module analyzes the degradation trend of equipment operating parameters, combines mechanistic models and data-driven models, and uses preset formulas to predict the remaining service life of key components.

5. The integrated monitoring and predictive maintenance system for power equipment based on digital twins according to claim 4, characterized in that, The preset formula used by the fault prediction module is: in, This represents the maximum permissible wear of the component. The current wear amount is inferred from vibration or temperature data. The wear rate per unit time predicted by the LSTM model is in mm / day; the RUL result is rounded to the nearest integer and is in days.

6. The integrated monitoring and predictive maintenance system for power equipment based on digital twins according to claim 1, characterized in that, The adaptive alarm module dynamically adjusts the alarm threshold based on the health score and fault prediction results.

7. The integrated monitoring and predictive maintenance system for power equipment based on digital twins according to claim 1, characterized in that, The power equipment includes at least one of an air compressor, a water pump, a fan, and a central air conditioner; the sensor includes at least one of a vibration sensor, a temperature sensor, a pressure sensor, and a current sensor.

8. A method for integrated monitoring and predictive maintenance of power equipment based on digital twins, based on the integrated monitoring and predictive maintenance system for power equipment based on digital twins as described in any one of claims 1-7, characterized in that, Includes the following steps: S1: Real-time acquisition of diverse operational data from power equipment; S2: Inject data into the corresponding digital twin model to drive the model to run synchronously with the physical device; S3: Use digital twin models for simulation to simulate the operating status of equipment under different working conditions; S4: Based on simulation results and real-time data, the health index of the equipment and the remaining service life of key components are calculated by combining the health index calculation formula and the remaining service life prediction formula with the machine learning model. S5: When the health index is lower than the preset threshold or the remaining service life is lower than the safety margin, the system automatically generates early warning information and maintenance suggestions; S6: After performing maintenance, maintenance personnel will feed the maintenance records back to the system for optimizing and calibrating parameters in the digital twin model and prediction algorithm.

9. A method for integrated monitoring and predictive maintenance of power equipment based on digital twins according to claim 8, characterized in that, The preset threshold in S5 is 0.6; the safety margin is 7 days.

10. A method for integrated monitoring and predictive maintenance of power equipment based on digital twins according to claim 8, characterized in that, The parameters to be optimized and calibrated in S6 include those in the preset formula used by the health status assessment module. , k, and the remaining lifetime formula .

Images