Data processing system for equipment intelligent operation and maintenance fault monitoring

Abstract

The invention discloses a data processing system for equipment intelligent operation and maintenance fault monitoring, which comprises an edge co-processing module, a predictive maintenance decision module and an industrial augmented reality AR verification module, and is characterized in that the edge co-processing module transmits a vibration harmonic analysis result to the predictive maintenance decision module in real time through a 5G URLLC network; a maintenance instruction generated by the predictive maintenance decision module is issued to the anti-explosion AR terminal through an MQTT protocol, and operation verification data of the anti-explosion AR terminal is fed back to the edge co-processing module through the MES system. According to the method, three core problems of poor dynamic adaptability, difficulty in multi-fault coupling detection and uncontrollable manual operation in industrial equipment maintenance are solved through a technical chain of edge cooperative computing-knowledge graph decision-AR verification closed loop.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial Internet of Things and predictive maintenance, and particularly to a data processing system for equipment intelligent operation and maintenance fault monitoring. Background Art

[0002] With the rapid development of Industry 4.0, the requirements for the real-time performance and accuracy of equipment fault prediction and maintenance in the field of intelligent manufacturing are increasing day by day. Traditional industrial operation and maintenance systems mostly adopt regular maintenance based on fixed thresholds or single-dimensional fault detection methods, which are difficult to cope with the dynamic coupling problems of multi-source heterogeneous data in complex industrial scenarios. Especially in high-load and high-noise environments such as automobile manufacturing and intelligent warehousing, equipment failures are often caused by the interaction of multiple factors such as mechanical vibration, electrical harmonics, and thermal stress. The traditional solutions have the following technical bottlenecks: 1. Limitations in data processing Existing systems usually rely on a single sensor (such as vibration or current) for fault detection, lacking the ability to analyze multi-source signals synergistically, and it is difficult to capture mechanical-electrical coupling faults (such as the positioning drift of a robotic arm caused by harmonic resonance).

[0003] 2. Static nature of maintenance decisions Traditional resource allocation strategies use fixed threshold partitioning (such as dividing low / medium / high loads according to load rates of 50% and 80%), which cannot adapt to production fluctuations (such as the load peak-valley ratio in a welding shop reaching 2.8:1), resulting in resource waste or response delays.

[0004] 3. Lack of verification closed-loop Most maintenance systems only generate operation instructions and lack real-time verification of manual operations, which are prone to secondary faults caused by operation errors (such as equipment resonance caused by deviation in bolt tightening torque). Summary of the Invention

[0005] The purpose of the present invention is to provide a data processing system for equipment intelligent operation and maintenance fault monitoring to solve the problems raised in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solution: A data processing system for equipment intelligent operation and maintenance fault monitoring, including: (1) Edge collaborative processing module, including: Industrial-grade edge computing nodes, using Xilinx Zynq-7030 type FPGA chips and configured with a dynamic resource allocation unit, and the dynamic resource allocation unit performs the following operations according to the load fluctuations of workshop equipment:

[0007] Wherein, is the peak load, is the average load applicable condition: 0.5 ≤ ≤ 2.8. The formula coefficients are obtained through regression analysis of historical load data, where a logarithmic function is used to smooth sudden changes in resource allocation. The coefficient 0.68 represents the baseline resource ratio, and 0.12 is the fluctuation correction factor. The applicability has been verified by Monte Carlo simulation (error ≤ 5%); A multi-source signal processing unit integrating the following functions: Vibration signal analysis: Perform order tracking on the vibration signals of the robotic arm joints with a resolution of 0.2 Hz, covering the rotational speed frequencies from 1 to 3 times; Current harmonic detection: Real-time monitor the 5th / 7th / 11th harmonic distortion rates of the AGV drive motor, with alarm thresholds of 3%, 2.5%, and 2% respectively; (2) A predictive maintenance decision-making module, including: A dynamic knowledge graph engine that constructs a fault model containing the following specific nodes: A guideway flatness drift node that is dynamically updated based on laser interferometer data with an accuracy of 0.5 μm / m; A harmonic resonance coupling fault node that marks the interaction between grid harmonics and mechanical vibrations; The relationships between the nodes are defined as follows: The guideway flatness drift causes the positioning error of the robotic arm, and the robotic arm positioning error triggers overcurrent in the servo motor; A maintenance strategy optimization unit that uses an improved TD3 algorithm to generate decision instructions, and its reward function is extended to:

[0008] where, is the actual output, is the theoretical maximum output of the equipment, is the single maintenance cost, is the preset maximum allowable cost, is the downtime caused by the current maintenance, is the historical average downtime; (3) An industrial augmented reality (AR) verification module, including: An explosion-proof AR terminal that integrates the following functions: Dynamic fastening strategy guidance: When the detected resonance frequency deviation of the equipment is ≥ 2%, regenerate the bolt fastening sequence and provide visual guidance through a finite element stress nephogram with a grid accuracy ≤ 3 mm; An acoustic torque verification unit, whose feature extraction method includes: Collect the signals in the 2 kHz - 8 kHz frequency band of the wrench knocking sound; Calculate the MFCC coefficients and match the template library through the dynamic time warping (DTW) algorithm; If the distance ≤ 0.35, it is determined that the torque meets the standard; Maintenance record synchronization interface, which uploads AR operation logs to the workshop MES system in real time; The dynamic resource allocation strategy is applicable to equipment clusters with a load fluctuation range of 0.5 ≤ ≤ 2.8 in industrial scenarios, including but not limited to stamping, welding, and assembly workshops: Load fluctuation range in the stamping workshop: 0.7 ≤ ≤ 2.1; Load fluctuation range in the welding workshop: 1.2 ≤ ≤ 2.8; Load fluctuation range in the assembly workshop: 0.5 ≤ ≤ 1.6.

[0009] The detection method for the harmonic resonance coupling fault node is as follows: Synchronously collect vibration acceleration signals (sampling rate ≥ 10 kHz) and grid current signals (accuracy class 0.5); Mark as a harmonic resonance risk when the following conditions are met: | |≤ 0.5 Hz where, is the vibration main frequency, is the grid fundamental frequency.

[0010] The acoustic torque verification function is enabled when the background noise ≤ 78 dB(A), and is only applicable to torque wrenches with a range of 5 - 60 Nm.

[0011] The detection data source for the guide rail flatness drift node is a laser interferometer, model: Renishaw XL - 80, detection period: perform a full guide rail scan every 8 hours, alarm threshold: flatness deviation ≥ 0.1 mm / m.

[0012] The implementation effects in the welding workshop include: the mechanical arm failure downtime is reduced from 12.3 hours / month to 1.5 hours / month; the maintenance misdetection rate is reduced from 8.7% to 0.8%, test conditions: noise ≤ 75 dB, sensor calibration period ≤ 7 days.

[0013] The synchronous alignment method for the harmonic resonance detection is PTP protocol (IEEE 1588) clock synchronization.

[0014] The target policy smoothing noise truncation threshold of the TD3 algorithm is 0.5.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. Improvement in dynamic resource allocation efficiency: Based on the dynamic resource allocation formula of the logarithmic function, when the load fluctuation range is [specific range], the resource utilization rate is increased from 72% to 89%; 2. Breakthrough in the detection accuracy of multi-source coupled faults: Through the synchronous analysis of vibration and current and the dynamic association of the deformation of the guide rail (accuracy 0.5μm / m) and the overcurrent of the servo motor by the knowledge graph, the false detection rate is reduced from 8.7% to 0.8%, and the fault location time is shortened by 82%; 3. Augmented reality closed-loop verification system: Through acoustic torque verification and finite element stress nephogram guidance (mesh accuracy ≤ 3mm), overstress damage is avoided, the manual operation error rate is reduced by 90%, and the MTBF (Mean Time Between Failures) of the equipment after maintenance is increased from 1200 hours to 3500 hours. Brief Description of the Drawings

[0016] Figure 1 It is a schematic diagram of the system framework of the present invention; Figure 2 It is a schematic diagram of the edge data processing flow of the present invention; Figure 3 It is a schematic diagram of the predictive maintenance decision-making process of the present invention; Figure 4 It is a schematic diagram of the AR-assisted operation process of the present invention. Detailed Embodiment

[0017] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0018] Please refer to Figures 1-4 , the present invention provides a technical solution: A data processing system for intelligent operation and maintenance fault monitoring of equipment, including: Edge collaborative processing module: The edge computing node uses Xilinx Zynq-7030 FPGA (125k logic units, power consumption ≤ 8W), equipped with an IP67 protection shell; Vibration sensor: PCB Piezotronics 356A01 (frequency response range 0.5Hz - 10kHz, sensitivity 100mV / g); Current acquisition module: HIOKI PW3390 (bandwidth DC - 5MHz, accuracy ±0.2%).

[0019] Dynamic resource allocation: Load data source: 12-month operation logs of 56 robotic arms in a certain automotive welding workshop (sampling interval 10 seconds); Resource allocation formula:

[0020] The unit is %CPU, and the priority settings are as follows: Vibration order analysis occupies 60% of the FPGA resources, and current harmonic detection occupies 25%.

[0021] Predictive maintenance decision-making module: Knowledge graph construction: Node type: Guide rail flatness drift: The data source for guide rail flatness detection is a laser interferometer, which is compatible with devices of the same accuracy (≤0.5μm / m) such as Renishaw XL-80; Harmonic resonance coupling: Vibration signals and grid current signals are synchronously collected (PTP protocol, time alignment error ≤1ms).

[0022] Node relationship logic: Guide rail flatness drift leads to robotic arm positioning errors, and robotic arm positioning errors trigger overcurrent in the servo motor; TD3 algorithm optimization: Reward function:

[0023] Training data: 10,000 groups of historical fault cases (covering scenarios such as transmission wear and electrical aging).

[0024] Industrial augmented reality AR verification module: Explosion-proof AR terminal: Model: RealWear HMT-1Z1 (ATEX Zone 2 certified, brightness 1000 nit); Dynamic fastening strategy: Generate bolt sequences based on the finite element analysis results of ANSYS Mechanical (mesh size ≤3mm); Acoustic torque verification: Collect wrench tapping sound signals (frequency band 2kHz - 8kHz), calculate MFCC coefficients (20 Mel filters), set the DTW distance threshold to 0.35, and when the background noise > 78dB(A), enable a band-pass filter (2kHz~8kHz) and an adaptive noise cancellation (ANC) algorithm to ensure a signal-to-noise ratio ≥15dB.

[0025] Implementation steps: Edge data processing flow: Step 101: Install vibration sensors at the robotic arm joints and connect current probes to the power supply line of the AGV drive motor; Step 102: The FPGA node calculates the vibration order spectrum (resolution 0.2Hz) and harmonic distortion rate (5 / 7 / 11th harmonics) in real time; Step 103: When the detected harmonic distortion rate exceeds the threshold (3% / 2.5% / 2%), trigger an update of the cloud federated learning model; Federated learning threshold update process: Data collection: Extract the following data from the workshop MES system every month: Average monthly actual monitored harmonic distortion rate; Recurrence rate of equipment failures after maintenance; Number of false alarm events in the current month; Threshold update rule: The harmonic distortion rate threshold is updated using weighted moving average:

[0026] Where: is the updated harmonic distortion rate threshold (unit: %), is the original threshold (unit: %), is the average monthly actual monitored harmonic distortion rate (unit: %), is the number of false alarm events in the current month, is the total number of detections in the current month. Coefficients 0.9, 0.1, 0.2 are determined by historical data regression ( = 0.85), and the weight coefficients are optimized by the gradient descent method with the goal of minimizing the weighted sum of the false alarm rate and the missed detection rate ( = 0.85).

[0027] Verification and deployment: The updated threshold needs to be verified online for 72 hours (false alarm rate ≤ 1.5%); It is batch issued to the edge computing nodes through the 5G URLLC network.

[0028] Predictive maintenance decision-making process: Step 201: The laser interferometer scans the flatness of the guide rail every 8 hours, and the data is uploaded to the knowledge graph engine; Step 202: If the detected deviation of the guide rail flatness ≥ 0.1 mm / m, the engine generates a maintenance work order and pushes it to the AR terminal; Step 203: The TD3 algorithm generates a maintenance strategy based on the real-time device status (load rate, MTBF), and issues an instruction through the 5G URLLC network (end-to-end delay ≤ 80 ms).

[0029] AR-assisted operation process: Step 301: The technician wears the AR terminal to scan the equipment QR code and retrieves the 3D disassembly and assembly guide; Step 302: Fill in the grease according to the AR scale prompt (error ±5 mL), and perform acoustic verification by knocking the wrench after the operation; Step 303: The acoustic signal is matched with the template library through the DTW algorithm. If the distance ≤ 0.35, it is determined to be qualified, and the data is synchronized to the MES system.

[0030]

[0031] Test item 1: The 2 kHz - 8 kHz frequency band effectively avoids common workshop noises (<1 kHz), and the MFCC + DTW algorithm has strong robustness; Test item 3: The logarithmic distribution formula is adapted to high-fluctuation scenarios, covering 95% of the load peak events; Test item 4: The accuracy reaches the nominal value of the laser interferometer (±0.5μm / m), and the threshold setting complies with the ASME B5.54 guideway maintenance standard; Test item 5: Dual verification of visual guidance + acoustic feedback significantly reduces human errors; Test item 6: The real-time scheduling algorithm combines the device health status (such as battery capacity > 30%) to ensure conflict prediction.

[0032] Example 1: Automobile welding workshop Test conditions: Equipment: 56 welding robotic arms (load fluctuation range 1.2 ≤ ≤ 2.8); Noise environment: ≤ 75dB(A), sensor calibration period 7 days.

[0033] Results:

[0034] Example 2: Intelligent warehousing scenario Test conditions: Number of AGVs: 50 (peak flow 50 units / hour, 5G URLLC network); Navigation method: SLAM laser navigation (positioning accuracy ±10mm).

[0035] Results:

[0036] The present invention solves the three core problems in industrial equipment maintenance, namely poor dynamic adaptability, difficult multi-fault coupling detection, and uncontrollable manual operation, through the technical chain of edge collaborative computing - knowledge graph decision-making - AR verification closed-loop.

[0037] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or device.

Claims

1. A data processing system for equipment intelligent operation and maintenance fault monitoring, characterized in that: include: (1) Edge collaborative processing module, including: The industrial edge computing node is configured with a dynamic resource allocation unit, which performs the following operations according to the load fluctuation of the workshop equipment: in, is the peak load, For average load, the applicable conditions are: 0.5 ≤ ≤ 2.8; Multi-source signal processing unit, integrating the following functions: Vibration signal analysis: order tracking of the vibration signal of the robot arm joint, with a resolution of 0.2Hz, covering 1~3 times the rotation speed frequency; Current harmonic detection: real-time monitoring of the 5th / 7th / 11th harmonic distortion rate of the AGV drive motor, with alarm thresholds of 3%, 2.5%, and 2% respectively; (2) Predictive maintenance decision module, including: A dynamic knowledge graph engine that builds a fault model containing the following unique nodes: Guideway flatness drift node, based on laser interferometer data: accuracy 0.5μm / m dynamic update; Harmonic resonance coupled fault nodes mark the interaction between grid harmonics and mechanical vibrations; The relationship between the nodes is defined as: The guide rail flatness drift causes the robot arm positioning error, and the robot arm positioning error triggers the servo motor overcurrent; The maintenance strategy optimization unit uses the improved TD3 algorithm to generate decision instructions, and its reward function is expanded to: in, is the actual output, is the theoretical maximum output of the equipment, is the single maintenance cost, To preset the maximum allowable cost, The downtime caused by the maintenance. is the historical average downtime duration; (3) Industrial augmented reality (AR) verification module, including: An explosion-proof AR terminal integrates the following functions: Dynamic tightening strategy guidance: When the equipment resonance frequency deviation is detected to be ≥2%, the bolt tightening sequence is regenerated and visual guidance is provided through finite element stress cloud diagram with mesh accuracy ≤3mm; Acoustic torque verification unit, the feature extraction method thereof includes: Collect the 2kHz-8kHz frequency band signal of the wrench knocking sound; Calculate MFCC coefficients and match the template library through the dynamic time warping (DTW) algorithm; If the distance is ≤0.35, the torque is determined to be up to standard; Maintain record synchronization interface to upload AR operation logs to the workshop MES system in real time; The edge collaborative processing module transmits the vibration harmonic analysis results to the predictive maintenance decision-making module in real time through the 5G URLLC network. The maintenance instructions generated by the predictive maintenance decision-making module are sent to the explosion-proof AR terminal through the MQTT protocol. The operation verification data of the explosion-proof AR terminal is fed back to the edge collaborative processing module through the MES system.

2. According to claim 1, a data processing system for equipment intelligent operation and maintenance fault monitoring is characterized by: The dynamic resource allocation strategy is applicable to industrial scenarios where the load fluctuation range meets 0.5 ≤ Equipment clusters ≤ 2.8, including but not limited to stamping, welding and assembly workshops: Load fluctuation range of stamping workshop: 0.7 ≤ ≤ 2.1; Welding workshop load fluctuation range: 1.2 ≤ ≤ 2.8; Load fluctuation range of assembly workshop: 0.5 ≤ ≤ 1.

6.

3. The data processing system for intelligent equipment operation and maintenance fault monitoring according to claim 1 is characterized in that: The detection method of the harmonic resonance coupling fault node is: Synchronously collect vibration acceleration signals (sampling rate ≥ 10kHz) and grid current signals (accuracy 0.5 level); Harmonic resonance risk is marked when the following conditions are met: | |≤0.5Hz; The duration of the interaction was ≥10 seconds; in, is the main vibration frequency, is the grid base frequency.

4. The data processing system for intelligent equipment operation and maintenance fault monitoring according to claim 1 is characterized in that: The acoustic torque verification function is enabled when the background noise is ≤78dB. The acoustic torque verification function is suitable for torque tools with a measuring range of 5-60Nm, and is compatible with other range devices through a frequency band adaptive algorithm.

5. The data processing system for intelligent equipment operation and maintenance fault monitoring according to claim 1 is characterized in that: The detection data source of the guide rail flatness drift node is a laser interferometer, the detection cycle: a full guide rail scan is performed every 8 hours, and the alarm threshold: flatness deviation ≥ 0.1mm / m.

6. The data processing system for intelligent equipment operation and maintenance fault monitoring according to claim 1, characterized in that: The implementation effects in the welding workshop include: the downtime of robot arm failure was reduced from 12.3 hours / month to 1.5 hours / month; the maintenance false detection rate was reduced from 8.7% to 0.8%. The test conditions are: noise ≤75dB, sensor calibration cycle ≤7 days.

7. According to claim 3, a data processing system for intelligent equipment operation and maintenance fault monitoring is characterized by: The synchronization alignment method of the harmonic resonance detection is PTP protocol clock synchronization, and the synchronization error of the harmonic resonance detection is ≤1ms.

8. The data processing system for intelligent equipment operation and maintenance fault monitoring according to claim 1, characterized in that: The target strategy smoothing noise cutoff threshold of the TD3 algorithm is 0.5.

Images