Edge diagnosis and maintenance system and method for industrial equipment

Abstract

The invention provides an edge diagnosis and maintenance system and method for industrial equipment. The system comprises a data acquisition module, a cloud collaboration module and at least one edge calculation module, the data acquisition module is used for acquiring multi-mode real-time operation data of industrial equipment; the cloud collaboration module is used for performing incremental training on the initial diagnosis model and the initial knowledge graph based on the multi-modal real-time operation data to obtain a target diagnosis model and a target knowledge graph, and issuing the target diagnosis model and the target knowledge graph to the edge calculation module; and the edge calculation module is used for determining whether the industrial equipment has a fault based on the multi-modal real-time operation data and the target diagnosis model, and if so, determining a fault position and a maintenance suggestion of the industrial equipment based on the multi-modal real-time operation data and the target knowledge graph. According to the invention, by setting the edge-cloud architecture, the accuracy and real-time performance of fault diagnosis and maintenance are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of fault diagnosis and maintenance, and particularly to an edge diagnosis and maintenance system and method for industrial equipment. Background Art

[0002] In industrial production, the stable operation of equipment is directly related to production efficiency and economic benefits. To ensure the normal operation of equipment, the following maintenance methods and fault prediction schemes are mainly adopted in the prior art. 1. Regular maintenance: According to the usage cycle of the equipment, regular inspections and part replacements are arranged to reduce the occurrence of equipment failures. 2. Breakdown maintenance: Maintenance or replacement is carried out after the equipment fails. 3. Threshold-based monitoring: The status of the equipment (such as vibration, temperature, pressure, etc.) is monitored through sensors, and an alarm is triggered when the data exceeds the set threshold. The disadvantages of the above methods are: only passive diagnosis of industrial equipment can be carried out, lacking accuracy, unable to predict faults in advance, and there is a risk of false alarms or missed alarms.

[0003] To solve the technical problem of the inability to predict faults in advance caused by the above passive response, fault prediction and maintenance are carried out by analyzing big data in the cloud, that is, sensor data is uploaded to the cloud through the Industrial Internet of Things (IIoT), and complex big data analysis or machine learning algorithms are used to evaluate the health status of the equipment. However, this method has data transmission delays, cannot meet the requirements of real-time response, is highly dependent on network stability, and is difficult to apply to industrial scenarios with unstable networks. The data transmission and cloud processing costs are relatively high. Therefore, the prior art proposes a method of running a machine learning model on edge devices to evaluate the equipment status, but it has the following technical problems: the model is fixed, lacking the adaptive ability for different equipment or working conditions, resulting in low accuracy of fault diagnosis and maintenance.

[0004] Therefore, there is an urgent need to provide an edge diagnosis and maintenance system and method for industrial equipment to achieve the accuracy of fault diagnosis and maintenance of industrial equipment. Summary of the Invention

[0005] In view of this, it is necessary to provide an edge diagnosis and maintenance system and method for industrial equipment to solve the technical problem in the prior art that a fixed model is used to diagnose and maintain all industrial equipment, with poor pertinence, and thus resulting in low accuracy of fault diagnosis and maintenance.

[0006] In a first aspect, to solve the above technical problem, the present invention provides an edge diagnosis and maintenance system for industrial equipment, including: a data acquisition module, a cloud collaboration module, and at least one edge computing module; The data acquisition module is used to acquire multi-modal real-time operation data of industrial equipment; The cloud collaboration module is used to perform incremental training on the initial diagnostic model and the initial knowledge graph based on the multimodal real-time operation data, obtain the target diagnostic model and the target knowledge graph, and send the target diagnostic model and the target knowledge graph to the edge computing module; The edge computing module is used to determine whether an industrial device has a fault based on the multimodal real-time operation data and the target diagnostic model. If so, it determines the fault location and maintenance suggestions of the industrial device based on the multimodal real-time operation data and the target knowledge graph.

[0007] In a possible implementation manner, the multimodal real-time operation data includes vibration data, temperature data, sound data, power data, and image data; the edge computing module includes a data feature extraction unit, a data fusion unit, a fault diagnosis unit, and a knowledge reasoning unit; The data feature extraction unit is used to extract the multimodal features and the feature change trends of the multimodal real-time operation data; The data fusion unit is used to perform feature fusion on the multimodals to obtain fusion features; The fault diagnosis unit is used to input the fusion features and the feature change trends into the target diagnostic model to determine whether the industrial device has a fault; The knowledge reasoning unit is used to perform knowledge reasoning based on the multimodal real-time operation data and the target knowledge graph when the industrial device has a fault, and obtain the fault location and maintenance suggestions of the industrial device.

[0008] In a possible implementation manner, the edge computing module further includes a data processing unit; The data processing unit is used to perform noise reduction, normalization, and outlier processing on the multimodal real-time operation data.

[0009] In a possible implementation manner, the edge computing module further includes a resource scheduling unit; The resource scheduling unit is used to obtain the running task volume of the edge computing module. When the running task volume is greater than the preset task volume, it determines the target component of the industrial device and the target data of the target component; Then the data feature extraction unit is used to extract the data features and the feature change trends of the target data.

[0010] In a possible implementation manner, the edge computing module further includes an optimization training unit; The optimization training unit is used to obtain the multi-modal historical operation data acquired by the edge computing module within a preset time period, and optimize and train the target knowledge graph based on the multi-modal historical operation data to obtain an optimized knowledge graph; the optimized knowledge graph is used to determine the fault location and maintenance suggestions of the industrial equipment.

[0011] In a possible implementation, the edge computing module further includes a version management unit and a rollback unit; The version management unit is used to record multiple target diagnostic models and multiple target knowledge graphs with different versions issued by the cloud collaboration module; The rollback unit is used to respectively retrieve the historical diagnostic model in the multiple target diagnostic models and the historical knowledge graph in the multiple target knowledge graphs based on a rollback instruction.

[0012] In a possible implementation, the cloud collaboration module includes a knowledge graph construction unit and a knowledge graph training unit; The knowledge graph construction unit is used to construct a knowledge graph based on expert experience, the historical operation data of the industrial equipment, and fault cases, and input the knowledge graph into a graph neural network model for knowledge improvement and fusion to obtain the initial knowledge graph; The knowledge graph training unit is used to perform incremental training on the weights and relationships in the initial knowledge graph based on the multi-modal real-time operation data to obtain the target knowledge graph.

[0013] In a possible implementation, the knowledge graph construction unit specifically is used for: Determine component nodes, operation parameter nodes, fault mode nodes, and maintenance suggestion nodes based on the expert experience, the historical operation data, and the fault cases, and determine the logical relationships between the component nodes, the operation parameter nodes, the fault mode nodes, and the maintenance suggestion nodes, and construct the knowledge graph based on the component nodes, the operation parameter nodes, the fault mode nodes, the maintenance suggestion nodes, and the logical relationships.

[0014] In a possible implementation, the maintenance suggestions include multiple sub-suggestions; the system further includes a maintenance feedback module; The maintenance feedback module is used to push the multiple sub-suggestions to a mobile terminal and receive a maintenance feedback result; The cloud collaboration module is further used to optimize the target knowledge graph based on the maintenance feedback result.

[0015] Second aspect, the present invention also provides a method for edge diagnosis and maintenance of industrial equipment, applicable to the edge diagnosis and maintenance system of industrial equipment described in any of the above possible implementation manners. The method includes: Collect multi-modal real-time operation data of industrial equipment based on a data acquisition module; Control the cloud collaboration module to perform incremental training on the initial diagnosis model and the initial knowledge graph based on the multi-modal real-time operation data, obtain a target diagnosis model and a target knowledge graph, and send the target diagnosis model and the target knowledge graph to the edge computing module; Control the edge computing module to determine whether there is a fault in the industrial equipment based on the multi-modal real-time operation data and the target diagnosis model. If so, determine the fault location and maintenance suggestions of the industrial equipment based on the multi-modal real-time operation data and the target knowledge graph.

[0016] The beneficial effects of the present invention are as follows: The edge diagnosis and maintenance system of industrial equipment provided by the present invention collects multi-modal real-time operation data of industrial equipment through a data acquisition module, and sets the initial diagnosis model and the initial knowledge graph in the cloud collaboration module to be optimized according to the multi-modal real-time operation data, which can improve the learning ability of the target diagnosis model and the target knowledge graph, and further improve the accuracy of the target diagnosis model and the target knowledge graph in fault diagnosis and maintenance. Moreover, the present invention collects multi-modal real-time operation data of industrial equipment, and the potential correlation between data can be extracted through the multi-modal real-time operation data. Compared with collecting single-modal operation data, the data is more comprehensive, which can further improve the accuracy of fault diagnosis and maintenance.

[0017] At the same time, when the edge computing module of the present invention performs fault diagnosis and prediction, it first determines whether there is a fault based on the target diagnosis model. When it is determined that there is a fault, it determines the fault location and maintenance suggestions based on the target knowledge graph. By combining the knowledge graph and the deep learning model, it provides accurate maintenance decision support and improves the accuracy of fault diagnosis and maintenance.

[0018] Furthermore, the present invention constructs an "edge-cloud" architecture, sends the optimized target diagnosis model and target knowledge graph of the cloud collaboration module to the edge computing module, and the lightweight deployment of the target diagnosis model and the target knowledge graph improves the real-time performance of fault diagnosis and maintenance. At the same time, the edge computing module can still independently complete the fault diagnosis and maintenance work when the network terminal or the cloud is unavailable, improving its applicability.

[0019] Furthermore, the cloud collaboration module can obtain the real-time operation data of all industrial devices. The global data obtained by it is cross-device and cross-condition, enabling the obtained target diagnostic model and target knowledge graph to have strong generalization ability, supporting accurate fault diagnosis and maintenance in scenarios with coexistence of multiple types of devices and multiple conditions, and improving the high robustness of the system. Moreover, the incremental training of the initial diagnostic model and the initial knowledge graph is only carried out in the cloud collaboration module, further reducing the computing power requirements of the edge computing module. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0021] Figure 1 FIG. is a schematic structural diagram of an embodiment of the edge diagnosis and maintenance system for industrial devices provided by the present invention; Figure 2 FIG. is a schematic structural diagram of an embodiment of the edge computing module provided by the present invention; Figure 3 FIG. is a schematic structural diagram of an embodiment of the cloud collaboration module provided by the present invention; Figure 4 FIG. is a schematic flowchart of an embodiment of the edge diagnosis and maintenance method for industrial devices provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative efforts fall within the protection scope of the present invention.

[0023] It should be understood that the schematic drawings are not drawn to scale. The flowcharts used in the present invention illustrate operations implemented according to some embodiments of the present invention. It should be understood that the operations of the flowchart may not be implemented in sequence, and steps without logical context may be reversed or implemented simultaneously. In addition, those skilled in the art can add one or more other operations to the flowchart or remove one or more operations from the flowchart under the guidance of the content of the present invention. Some of the block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor systems and / or microcontroller systems.

[0024] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in connection with the embodiments can be included in at least one embodiment of the present invention. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment mutually exclusive with other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0025] The present invention provides an edge diagnosis and maintenance system and method for industrial equipment, which will be described separately below.

[0026] Figure 1 FIG. is a schematic structural diagram of an embodiment of the edge diagnosis and maintenance system for industrial equipment provided by the present invention, as Figure 1 shown, the edge diagnosis and maintenance system 10 of the industrial equipment includes: a data acquisition module 100, a cloud collaboration module 200, and at least one edge computing module 300; The data acquisition module 100 is used to acquire multi-modal real-time operation data of the industrial equipment; The cloud collaboration module 200 is used to perform incremental training on the initial diagnosis model and the initial knowledge graph based on the multi-modal real-time operation data to obtain a target diagnosis model and a target knowledge graph, and send the target diagnosis model and the target knowledge graph to the edge computing module 300; The edge computing module 300 is used to determine whether the industrial equipment has a fault based on the multi-modal real-time operation data and the target diagnosis model. If so, it determines the fault location and maintenance suggestions of the industrial equipment based on the multi-modal real-time operation data and the target knowledge graph.

[0027] Among them, the data acquisition module 100 can be composed of multiple types of sensors, including but not limited to: vibration sensors, temperature sensors, sound sensors, and cameras.

[0028] It should be noted that: the number of edge computing modules 300 is the same as the number of industrial devices, that is, each edge computing module 300 is used to perform fault diagnosis and maintenance on the corresponding industrial device, further reducing the computing power requirements of the edge computing module 300.

[0029] In some other embodiments, an edge computing module 300 may also correspond to multiple industrial devices, but the number of corresponding industrial devices should not be too large to avoid a decrease in the diagnostic efficiency of the edge computing module 300.

[0030] Among them, when the cloud collaboration module 200 distributes the target diagnostic model and the target knowledge graph to the edge computing module 300, it is distributed through the OTA (Over-The-Air) mechanism.

[0031] It should be noted that: to ensure the accuracy of fault diagnosis and maintenance, when performing fault diagnosis and maintenance, the edge computing module 300 can use historical maintenance record data in addition to multi-modal real-time operation data. For example, when the determined fault location is "bearing wear", the historical maintenance record data can be called. When it is determined through the historical maintenance record data that the bearing has been running continuously for more than the recommended replacement period, it is confirmed that the fault location is accurate. That is, by using other parameter data such as historical maintenance record data, the accuracy of diagnosis and maintenance is further improved.

[0032] Compared with the prior art, the edge diagnosis and maintenance system 10 of industrial devices provided by the embodiments of the present invention can improve the learning ability of the target diagnostic model and the target knowledge graph by setting the data acquisition module 100 to collect multi-modal real-time operation data of industrial devices and setting the initial diagnostic model and the initial knowledge graph in the cloud collaboration module 200 to be optimized according to the multi-modal real-time operation data, and then improve the accuracy of the target diagnostic model and the target knowledge graph when performing fault diagnosis and maintenance. Moreover, the multi-modal real-time operation data of industrial devices collected by the embodiments of the present invention can extract the potential correlation between data. Compared with collecting single-modal operation data, the data is more comprehensive, which can further improve the accuracy of fault diagnosis and maintenance.

[0033] At the same time, when the edge computing module 300 of the embodiments of the present invention performs fault diagnosis and prediction, it first determines whether there is a fault based on the target diagnostic model. When it is determined that there is a fault, it determines the fault location and maintenance suggestions based on the target knowledge graph, combines the knowledge graph and the deep learning model, provides accurate maintenance decision support, and improves the accuracy of fault diagnosis and maintenance.

[0034] Furthermore, the embodiment of the present invention constructs an "edge-cloud" architecture, and distributes the optimized target diagnosis model and target knowledge graph of the cloud collaboration module 200 to the edge computing module 300. The lightweight deployment of the target diagnosis model and target knowledge graph improves the real-time performance of fault diagnosis and maintenance. At the same time, the edge computing module 300 can still independently complete fault diagnosis and maintenance work when the network terminal or the cloud is unavailable, improving its applicability.

[0035] Furthermore, the cloud collaboration module 200 can obtain the real-time operation data of all industrial devices. The global data obtained by it is cross-device and cross-condition, so that the obtained target diagnosis model and target knowledge graph have strong generalization ability, support accurate fault diagnosis and maintenance in scenarios where multiple types of devices and multiple conditions coexist, and improve the high robustness of the system. In addition, the incremental training of the initial diagnosis model and the initial knowledge graph is only carried out in the cloud collaboration module 200, further reducing the computing power requirements of the edge computing module 300.

[0036] In some embodiments of the present invention, the multi-modal real-time operation data includes vibration data, temperature data, sound data, power data and image data; as shown in Figure 2 Figure, the edge computing module 300 includes a data feature extraction unit 310, a data fusion unit 320, a fault diagnosis unit 330 and a knowledge reasoning unit 340; The data feature extraction unit 310 is used to extract the multi-modal features and feature change trends of the multi-modal real-time operation data; The data fusion unit 320 is used to fuse the multi-modal features to obtain the fused features; The fault diagnosis unit 330 is used to input the fused features and feature change trends into the target diagnosis model to determine whether there is a fault in the industrial device; The knowledge reasoning unit 340 is used to perform knowledge reasoning based on the multi-modal real-time operation data and the target knowledge graph when the industrial device has a fault, and obtain the fault location and maintenance suggestions of the industrial device.

[0037] The embodiment of the present invention not only extracts multi-modal features, but also extracts feature change trends. By expressing the multi-modal real-time operation data with two different dimensions of features, the accuracy of fault diagnosis and maintenance can be further improved.

[0038] Among them, the feature change trend can be extracted through a time series model (such as LSTM, GRU), and the data fusion unit can use a multi-modal neural network (such as Transformer or multi-modal LSTM) to realize the fusion of multi-modal real-time operation data.

[0039] For example, infrared thermal imaging can be combined with vibration data to identify the diagnostic result of "abnormal vibration caused by overheating of local equipment". Another example: when the characteristic change trend is "gradually increasing vibration frequency", the diagnostic result of "bearing wear" is identified.

[0040] Since the multi-modal real-time operation data directly collected by the data acquisition module 100 includes interferences such as noise, in some embodiments of the present invention, to eliminate interferences such as noise, as Figure 2 shown, the edge computing module 300 further includes a data processing unit 350; The data processing unit 350 is used to perform noise reduction, normalization, and outlier processing on the multi-modal real-time operation data.

[0041] Among them, the outlier processing is specifically: determining the range of each modal real-time operation data, and if it exceeds this range, it is regarded as an outlier and the outlier is removed.

[0042] To ensure the real-time nature of fault diagnosis and maintenance, the computing capacity that the edge computing module 300 can bear is relatively small. Therefore, when the task volume of the edge computing module 300 surges, in some embodiments of the present invention, as Figure 2 shown, the edge computing module 300 further includes a resource scheduling unit 360; The resource scheduling unit 360 is used to obtain the operation task volume of the edge computing module. When the operation task volume is greater than the preset task volume, determine the target component of the industrial equipment and the target data of the target component; Then the data feature extraction unit 310 is used to extract the data features and feature change trends of the target data.

[0043] The resource scheduling unit 360 in the embodiments of the present invention is used to, when the operation task volume is greater than the preset task volume, preferentially process the key data of the key components of the industrial equipment, that is: the target component and the target data, and delay the processing of non-key components or non-key data, taking into account the real-time nature while ensuring the accuracy of fault diagnosis and maintenance.

[0044] Among them, the key components are high-risk equipment, and the key data can be vibration data and temperature data.

[0045] Since the working conditions or fault types of different industrial equipment are not completely the same, to further improve the accuracy of fault diagnosis and maintenance for each industrial equipment, in some embodiments of the present invention, as Figure 2 shown, the edge computing module 300 further includes an optimization training unit 370; The optimization training unit 370 is used to obtain the multi-modal historical operation data acquired by the edge computing module within a preset duration, and optimize and train the target knowledge graph based on the multi-modal historical operation data to obtain an optimized knowledge graph; the optimized knowledge graph is used to determine the fault location and maintenance suggestions of industrial equipment.

[0046] In the embodiments of the present invention, by performing secondary optimization on the target knowledge graph sent by the cloud collaboration module 200 based on the multi-modal historical operation data within a preset duration, the obtained optimized knowledge graph can be more adapted to the corresponding industrial equipment, and thus the accuracy of fault diagnosis and maintenance can be further improved.

[0047] It should be noted that: the edge computing module 300 only performs secondary optimization on the target knowledge graph, rather than optimizing the target diagnosis model. This is because optimizing the target diagnosis model requires complex processes such as feature extraction and training of multi-modal operation data. Limited by the computing power of the edge computing module 300, placing the optimization process of the diagnosis model in the cloud collaboration module 200 can further ensure the lightweight deployment of the model in the edge computing module 300, and thus further improve the real-time performance of fault diagnosis and maintenance.

[0048] Among them, the parameter adjusted by the secondary optimization is the node weight of the knowledge graph.

[0049] In actual application scenarios, there are requirements such as testing that need to compare multiple different optimized versions of the target knowledge graph and the target diagnosis model. To adapt to this requirement, in some embodiments of the present invention, as Figure 2 shown, the edge computing module 300 further includes a version management unit 380 and a rollback unit 390; The version management unit 380 is used to record multiple target diagnosis models and multiple target knowledge graphs with different versions sent by the cloud collaboration module; The rollback unit 390 is used to respectively retrieve the historical diagnosis model among the multiple target diagnosis models and the historical knowledge graph among the multiple target knowledge graphs based on the rollback instruction.

[0050] In the embodiments of the present invention, by setting the version management unit 380 to record multiple versions of the target diagnosis model and the target knowledge graph for the rollback unit 390 to retrieve, the flexible use of different versions of the model can be realized. In addition to adapting to the version test requirements, when the latest version is unreliable or unstable, it can roll back to the previous version of the model to ensure the stability and reliability of fault diagnosis and maintenance.

[0051] Since the comprehensiveness and accuracy of the target knowledge graph are crucial for the rationality and accuracy of diagnosis and maintenance, therefore, in some embodiments of the present invention, as Figure 3 shown, the cloud collaboration module 200 includes a knowledge graph construction unit 210 and a knowledge graph training unit 220; The knowledge graph construction unit 210 is used to construct a knowledge graph based on expert experience, historical operation data, and fault cases of industrial equipment, and input the knowledge graph into the graph neural network model for knowledge improvement and fusion to obtain an initial knowledge graph; The knowledge graph training unit 220 is used to perform incremental training on the weights and relationships in the initial knowledge graph based on multi-modal real-time operation data to obtain a target knowledge graph.

[0052] In the embodiment of the present invention, during the construction process of the initial knowledge graph, the knowledge graph is improved and fused based on the graph neural network model, so that the constructed initial knowledge graph can realize fault diagnosis in complex scenarios and improve the scenario adaptability of fault diagnosis. Moreover, through the graph neural network model, the diagnosis of unknown fault modes can be realized, and further the accuracy of fault diagnosis can be improved.

[0053] In a specific embodiment of the present invention, the knowledge graph construction unit 210 is specifically used for: Determine component nodes, operation parameter nodes, fault mode nodes, and maintenance suggestion nodes based on expert experience, historical operation data, and fault cases, and determine the logical relationships between the component nodes, operation parameter nodes, fault mode nodes, and maintenance suggestion nodes, and construct a knowledge graph based on the component nodes, operation parameter nodes, fault mode nodes, maintenance suggestion nodes, and logical relationships.

[0054] In a specific embodiment of the present invention, the component nodes include but are not limited to nodes such as bearings, gears, and transmission devices, the operation parameter nodes include but are not limited to nodes such as temperature, vibration, pressure, and power, the fault mode nodes include but are not limited to nodes such as looseness, overheating, wear, and cracks, and the maintenance suggestion nodes include but are not limited to nodes such as repair, replacement, and lubricating oil addition.

[0055] In a specific embodiment of the present invention, the lightweight deep learning model deployed on the edge computing module 300 performs spectrum analysis on the vibration signal and finds that the vibration amplitude is mainly concentrated at a specific frequency (consistent with the historical "bearing wear" case). At the same time, the temperature, sound, and power data further indicate abnormal operation of the main shaft, and the possibility of a fault in the cutting tool is preliminarily excluded. Then, the edge computing module 300 combines real-time data and historical cases through the knowledge graph inference engine to further lock in the specific fault mode: the rule chain "increased vibration frequency + abnormal bearing temperature + impact sound = bearing wear" in the knowledge graph is triggered; the historical maintenance record shows that the bearing has been continuously operated for more than the recommended replacement period, further deepening the possibility of "bearing wear" as a fault.

[0056] Through the real-time analysis and knowledge reasoning of the edge computing module 300, the system completes the diagnosis within milliseconds after the vibration anomaly appears, generating a preliminary diagnosis result: "The main shaft bearing is severely worn and needs to be replaced as soon as possible." This avoids the cascading impact of further bearing damage on the main shaft or the entire machine tool.

[0057] In some embodiments of the present invention, the maintenance suggestions include multiple sub-suggestions; as Figure 1 shown, the edge diagnosis and maintenance system 10 of industrial equipment further includes a maintenance feedback module 400; The maintenance feedback module 400 is used to push multiple sub-suggestions to the mobile terminal and receive the maintenance feedback result; The cloud collaboration module 200 is further used to optimize the target knowledge graph based on the maintenance feedback result.

[0058] In the embodiment of the present invention, the feedback result received by the maintenance feedback module 400 is used as the optimization parameter data of the target knowledge graph to optimize the target knowledge graph, which can improve the accuracy of the maintenance suggestions determined by the subsequent target knowledge graph.

[0059] In a specific embodiment of the present invention, when the fault diagnosis result is "the main shaft bearing is worn", the maintenance suggestions include four items, namely, stop the equipment operation to avoid further damage to the main shaft; check and replace the main shaft bearing; check the lubrication system to confirm the lubricating oil quality and smooth oil supply; prompt the purchasing department to stock up in advance to ensure sufficient spare parts for subsequent replacement. The feedback result after the operator of the mobile terminal performs maintenance according to the maintenance suggestions is: "The bearing replacement is completed, and the lubrication system is checked without abnormality", then this data is fed back to the target knowledge graph for optimizing future diagnosis and suggestions.

[0060] In summary, the edge diagnosis and maintenance system for industrial equipment proposed in the embodiments of the present invention: 1. When constructing the knowledge graph, it combines a graph neural network model to achieve dynamic expansion and reasoning of knowledge, enabling the system to not only identify known faults but also have the ability to autonomously evolve and adapt to unknown fault patterns by learning new rules in real time. 2. It innovatively integrates heterogeneous data sources (such as vibration, temperature, sound, images, power, etc.), extracts the potential correlations between data through a multi-modal deep learning model, and transcends the single data analysis mode of traditional methods, enabling the system to support panoramic perception of the operating state of complex equipment and accurate fault location, especially for diagnosis under the coexistence of multiple faults or complex working conditions. 3. It integrates a lightweight deep learning model and a knowledge graph reasoning engine on the edge computing module, with millisecond-level real-time diagnosis and autonomous decision-making capabilities. And even in the case of network interruption or unavailability of the cloud, the system can still independently complete the monitoring of the equipment operating state and fault diagnosis. 4. The cloud collaboration module is responsible for global data integration and model optimization, and the edge computing module realizes rapid adaptation to different working conditions. 5. Based on the knowledge graph and equipment historical records, personalized and actionable maintenance suggestions are generated, providing full-process closed-loop support from fault diagnosis - predictive analysis - maintenance suggestion generation. The maintenance suggestions combine the actual usage scenarios of the equipment (such as environmental conditions, historical fault characteristics), significantly improving the accuracy and practicality of maintenance.

[0061] Correspondingly, the embodiments of the present invention also provide an edge diagnosis and maintenance method for industrial equipment, which is applicable to the edge diagnosis and maintenance system for industrial equipment in any of the above embodiments. As Figure 4 shown, the edge diagnosis and maintenance method for industrial equipment includes: S401. Collect multi-modal real-time operation data of the industrial equipment based on the data acquisition module; S402. Control the cloud collaboration module to perform incremental training on the initial diagnosis model and the initial knowledge graph based on the multi-modal real-time operation data, obtain the target diagnosis model and the target knowledge graph, and send the target diagnosis model and the target knowledge graph to the edge computing module; S403. Control the edge computing module to determine whether there is a fault in the industrial equipment based on the multi-modal real-time operation data and the target diagnosis model. If so, determine the fault location and maintenance suggestions of the industrial equipment based on the multi-modal real-time operation data and the target knowledge graph.

[0062] For the specific implementation processes and details of each step of the edge diagnosis and maintenance method for industrial equipment provided in the above embodiments, reference can be made to the technical solutions described in the embodiments of the edge diagnosis and maintenance system for industrial equipment above, which will not be elaborated here.

[0063] The above has introduced in detail a system and method for edge diagnosis and maintenance of an industrial device provided by the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those skilled in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. An edge diagnosis and maintenance system for industrial equipment, characterized in that: include: A data collection module, a cloud collaboration module, and at least one edge computing module; The data acquisition module is used to collect multi-modal real-time operation data of industrial equipment; The cloud collaboration module is used to perform incremental training on the initial diagnosis model and the initial knowledge graph based on the multimodal real-time operation data, obtain the target diagnosis model and the target knowledge graph, and send the target diagnosis model and the target knowledge graph to the edge computing module; The edge computing module is used to determine whether the industrial equipment has a fault based on the multimodal real-time operation data and the target diagnostic model. If so, the fault location and maintenance suggestions of the industrial equipment are determined based on the multimodal real-time operation data and the target knowledge graph.

2. The edge diagnosis and maintenance system for industrial equipment according to claim 1, characterized in that: The multimodal real-time operation data includes vibration data, temperature data, sound data, power data and image data; the edge computing module includes a data feature extraction unit, a data fusion unit, a fault diagnosis unit and a knowledge reasoning unit; The data feature extraction unit is used to extract the multimodal features and feature change trends of the multimodal real-time operation data; The data fusion unit is used to fuse the multi-modal features to obtain fusion features; The fault diagnosis unit is used to input the fusion feature and the feature change trend into the target diagnosis model to determine whether the industrial equipment has a fault; The knowledge reasoning unit is used to perform knowledge reasoning based on the multimodal real-time operation data and the target knowledge graph when the industrial equipment fails, so as to obtain the fault location and maintenance suggestions of the industrial equipment.

3. The edge diagnosis and maintenance system for industrial equipment according to claim 2, characterized in that: The edge computing module also includes a data processing unit; The data processing unit is used to perform noise reduction, normalization and outlier processing on the multimodal real-time operation data.

4. The edge diagnosis and maintenance system for industrial equipment according to claim 2, characterized in that: The edge computing module also includes a resource scheduling unit; The resource scheduling unit is used to obtain the running task amount of the edge computing module, and when the running task amount is greater than the preset task amount, determine the target component of the industrial equipment and the target data of the target component; The data feature extraction unit is used to extract the data features and feature change trends of the target data.

5. The edge diagnosis and maintenance system for industrial equipment according to claim 1, characterized in that: The edge computing module also includes an optimization training unit; The optimization training unit is used to obtain the multimodal historical operation data obtained by the edge computing module within a preset time period, and optimize the target knowledge graph based on the multimodal historical operation data to obtain an optimized knowledge graph; The optimized knowledge graph is used to determine the fault location and maintenance suggestions of the industrial equipment.

6. The edge diagnosis and maintenance system for industrial equipment according to claim 1, characterized in that: The edge computing module also includes a version management unit and a rollback unit; The version management unit is used to record multiple target diagnosis models and multiple target knowledge graphs of different versions issued by the cloud collaboration module; The rollback unit is used to call the historical diagnostic models in the multiple target diagnostic models and the historical knowledge graphs in the multiple target knowledge graphs respectively based on the rollback instruction.

7. The edge diagnosis and maintenance system for industrial equipment according to claim 1, characterized in that: The cloud collaboration module includes a knowledge graph construction unit and a knowledge graph training unit; The knowledge graph construction unit is used to construct a knowledge graph based on expert experience, historical operation data and fault cases of the industrial equipment, and input the knowledge graph into the graph neural network model for knowledge improvement and fusion to obtain the initial knowledge graph; The knowledge graph training unit is used to perform incremental training on the weights and relationships in the initial knowledge graph based on the multimodal real-time operation data to obtain the target knowledge graph.

8. The edge diagnosis and maintenance system for industrial equipment according to claim 7, characterized in that: The knowledge graph construction unit is specifically used for: Based on the expert experience, the historical operation data and the failure cases, component nodes, operation parameter nodes, failure mode nodes and maintenance suggestion nodes are determined, and the logical relationships among the component nodes, the operation parameter nodes, the failure mode nodes and the maintenance suggestion nodes are determined, and the knowledge graph is constructed based on the component nodes, the operation parameter nodes, the failure mode nodes, the maintenance suggestion nodes and the logical relationships.

9. The edge diagnosis and maintenance system for industrial equipment according to claim 1, characterized in that: The maintenance suggestion includes a plurality of sub-suggestions; the system also includes a maintenance feedback module; The maintenance feedback module is used to push the multiple sub-suggestions to the mobile terminal and receive maintenance feedback results; The cloud collaboration module is also used to optimize the target knowledge graph based on the maintenance feedback results.

10. An edge diagnosis and maintenance method for industrial equipment, characterized in that: The edge diagnosis and maintenance system for industrial equipment according to any one of claims 1 to 9, the method comprising: Collect multi-modal real-time operation data of industrial equipment based on data acquisition module; Control the cloud-based collaborative module to perform incremental training on the initial diagnosis model and the initial knowledge graph based on the multimodal real-time operation data, obtain the target diagnosis model and the target knowledge graph, and send the target diagnosis model and the target knowledge graph to the edge computing module; The control edge computing module determines whether the industrial equipment has a fault based on the multimodal real-time operation data and the target diagnosis model. If so, the fault location and maintenance suggestions of the industrial equipment are determined based on the multimodal real-time operation data and the target knowledge graph.

Images