Fault diagnosis method and system for intelligent manufacturing system operation and maintenance

Abstract

The present invention provides a fault diagnosis method and system for the operation and maintenance of intelligent manufacturing systems. First, an equipment association graph is constructed, which consists of equipment component nodes and component association edges. Then, operation monitoring data of each equipment component node is obtained, and node features are extracted to generate operating condition time series features. At the same time, edge features are extracted from the operation monitoring data of nodes at both ends of the component association edge to generate collaborative change features. Then, a pre-trained graph reasoning model is called to perform fault propagation analysis on the equipment association graph, operating condition time series features and collaborative change features, and generate a fault propagation result containing an abnormal originating node identifier and a propagation path sequence. Finally, based on the result, a historical fault case library is matched, the root cause component is located, and an operation and maintenance plan containing a maintenance operation sequence and parameter adjustment instructions is generated. The plan is pushed to the equipment control system to trigger a fault repair operation, thereby accurately diagnosing faults, locating root causes, and quickly generating operation and maintenance plans, thereby improving the operation and maintenance efficiency and accuracy of the intelligent manufacturing system.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing operation and maintenance technology, and in particular to a fault diagnosis method and system for intelligent manufacturing system operation and maintenance. Background Art

[0002] With the rapid development of intelligent manufacturing systems, system complexity is increasing, and the interconnectedness between devices is becoming increasingly close. Traditional intelligent manufacturing system operation and maintenance fault diagnosis methods rely primarily on manual experience and simple sensor data monitoring. Manual experience-based diagnosis methods are highly dependent on the technical expertise and accumulated experience of technicians. Differences in the skills of different technicians can lead to inaccurate and inconsistent diagnostic results. Furthermore, when faced with complex and changing system failures, manual diagnosis is inefficient and difficult to quickly locate the root cause. Simple sensor data monitoring typically only captures localized operational data from devices, lacks a comprehensive consideration of the interconnectedness between devices, and cannot accurately analyze the propagation path of the fault within the system.

[0003] Furthermore, existing methods often struggle to directly generate targeted maintenance plans after diagnosing a fault. Instead, maintenance personnel must develop action plans based on the diagnostic results. This not only increases the complexity and time-consuming nature of maintenance work, but can also lead to imperfect maintenance plans due to human factors, impacting the effectiveness and efficiency of fault repair. Therefore, how to accurately and efficiently diagnose maintenance faults in intelligent manufacturing systems and quickly generate effective maintenance plans has become a pressing technical challenge. Summary of the Invention

[0004] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a fault diagnosis method for operation and maintenance of an intelligent manufacturing system, the method comprising:

[0005] Constructing a device association graph, wherein the device association graph is composed of device component nodes and component association edges;

[0006] Acquire the operation monitoring data of each equipment component node, and perform node feature extraction processing on the operation monitoring data to generate the working condition time series features of each equipment component node;

[0007] Performing edge feature extraction processing on the operation monitoring data of the equipment component nodes at both ends of the component association edge to generate collaborative change features of the component association edge;

[0008] Calling a pre-trained graph reasoning model to perform fault propagation analysis and processing on the equipment association graph, the operating condition time series characteristics, and the coordinated change characteristics, and generating a fault propagation result including an abnormal originating node identifier and a propagation path sequence;

[0009] Based on the fault propagation results, the historical fault case library is matched to locate the root cause component that causes the equipment abnormality, and an operation and maintenance plan including a maintenance operation sequence and parameter adjustment instructions is generated according to the root cause component. The operation and maintenance plan is pushed to the equipment control system to trigger the fault repair operation.

[0010] On the other hand, an embodiment of the present invention also provides a fault diagnosis system for the operation and maintenance of an intelligent manufacturing system, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0011] Based on the above aspects, the embodiments of the present invention comprehensively depict the association relationships between equipment components by constructing an equipment association graph, obtain the operation monitoring data of each equipment component node and perform node feature extraction processing to generate operating condition time series features, which can accurately reflect the operating status of equipment components at different time points. Edge feature extraction processing is performed on the operation monitoring data of the equipment component nodes at both ends of the component association edge to generate collaborative change features, fully considering the collaborative effects between equipment components and facilitating the discovery of potential fault associations. A pre-trained graph reasoning model is called to perform fault propagation analysis and processing on the equipment association graph, operating condition time series features, and collaborative change features, and can accurately generate fault propagation results including abnormal origin node identifiers and propagation path sequences, comprehensively revealing the propagation process of faults in the system. Based on the fault propagation results, the historical fault case library is matched to quickly locate the root cause component that causes the equipment abnormality, avoiding the tediousness and uncertainty of manual troubleshooting. An operation and maintenance plan is generated based on the root cause component, including a maintenance operation sequence and parameter adjustment guidance, and the operation and maintenance plan is pushed to the equipment control system to trigger the fault repair operation. This achieves the automated connection between fault diagnosis and operation and maintenance plan generation, greatly improving the efficiency and accuracy of operation and maintenance work, reducing operation and maintenance costs, and ensuring the stable operation of the intelligent manufacturing system. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a schematic diagram of the execution flow of the fault diagnosis method for intelligent manufacturing system operation and maintenance provided by an embodiment of the present invention.

[0013] Figure 2 Schematic diagram of exemplary hardware and software components of a fault diagnosis system for intelligent manufacturing system operation and maintenance provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0014] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1This is a flow chart of a fault diagnosis method for intelligent manufacturing system operation and maintenance provided by an embodiment of the present invention. The fault diagnosis method for intelligent manufacturing system operation and maintenance is introduced in detail below.

[0015] Step S110: constructing a device association graph, wherein the device association graph is composed of device component nodes and component association edges.

[0016] In the intelligent manufacturing system, in order to construct the equipment association diagram, the automobile engine intelligent manufacturing production line is used as a specific application scenario. The automobile engine intelligent manufacturing production line is mainly used to produce different models of automobile engines and includes a variety of equipment and process links.

[0017] Step S111: Identify the core functional modules in the intelligent manufacturing system, which include a processing unit directly involved in product processing, a transmission unit responsible for power transmission, and a temperature control unit that maintains a stable equipment temperature.

[0018] For example, in an intelligent automotive engine manufacturing production line, identifying core functional modules forms the foundation for constructing an equipment association diagram. Machining units, such as cylinder block machining centers and crankshaft grinders, are directly involved in the processing of engine components. These machines use various cutting and grinding processes to transform raw materials into engine components that meet design requirements. The transmission unit is responsible for transmitting power from the power source (such as a motor) to the machining unit and other power-requiring equipment. These transmission units employ various transmission methods, including belt and gear drives, to ensure that each device operates at the desired speed and torque. The temperature control unit plays a critical role in maintaining stable equipment temperatures within the production line. Engine machining generates significant heat. If this heat is not dissipated promptly, it can lead to reduced equipment accuracy or even damage. The temperature control unit dissipates this heat through a coolant circulation system, cooling fans, and other devices, ensuring that the equipment operates within an appropriate temperature range.

[0019] Step S112: creating a device component node for each key component in the core functional module, wherein the initial attributes of the device component node include the component name, installation location, and design operation parameters.

[0020] For the machining unit, there are many key components. Taking the cylinder machining center as an example, the tool is an important key component. When creating a device component node for the tool, the component name in its initial attributes is clearly "cylinder machining center tool", and the installation position is the specific position of the tool in the machining center tool magazine. The design operating parameters include the tool speed, feed rate, cutting depth, etc. These design operating parameters are the standards for the normal operation of the tool and are used for subsequent monitoring and judgment of whether the tool has abnormalities. The spindle is also a key component of the machining unit. Its component name is "cylinder machining center spindle", and the installation position is in the spindle box of the machining center. The design operating parameters include the spindle's rated power, speed range, torque, etc. The above design operating parameters determine the working capacity and performance of the spindle.

[0021] Gears are key components in transmission units. Create a device component node for the gears, named "Transmission Gear." The installation location is determined by their specific location in the transmission system, such as a specific gear position within the gearbox. Design parameters include the gear's module, number of teeth, tooth width, and transmission ratio, all of which influence the gear's transmission efficiency and stability. The belt, also a crucial component of the transmission unit, is named "Transmission Belt." It's installed between the pulleys. Design parameters include the belt model and tension. Appropriate tension ensures proper belt transmission.

[0022] Key components of the temperature control unit, such as the coolant pump and cooling fan, also require device component nodes. The coolant pump is named "Coolant Pump" and is installed on the coolant circulation system piping. Design operating parameters include pump flow rate and head, which determine the coolant circulation speed and pressure. The cooling fan is named "Cooling Fan" and is installed near the radiator. Design operating parameters include fan speed and air volume, which affect the cooling effect.

[0023] Step S113: Analyze the physical connection structure between the processing unit and the transmission unit, and extract the specific form of the mechanical linkage relationship as the connection basis of the mechanical linkage edge.

[0024] In intelligent automotive engine manufacturing production lines, a close mechanical linkage exists between the machining unit and the transmission unit. For example, in a cylinder machining center and its transmission system, the transmission system transmits power to the machining center's spindle via belts or gears, enabling the spindle to rotate for cutting. When analyzing this physical connection structure, it's necessary to determine the specific path and method of power transmission. For example, when the transmission system uses a belt drive, factors such as the belt tension and the pulley diameter ratio affect the efficiency and stability of power transmission. The specific form of these mechanical linkages, such as the belt drive's transmission ratio and torque transmission characteristics, is extracted and used as a basis for connecting the mechanical linkage edges. This approach accurately reflects the interaction between the machining unit and the transmission unit.

[0025] Step S114: Analyze the heat exchange path between the processing unit and the temperature control unit, and extract the medium type of the heat conduction relationship as the connection basis of the heat conduction edge.

[0026] During operation, machining units generate a significant amount of heat, which needs to be dissipated through a temperature control unit. For example, in a cylinder machining center, friction between the tool and the workpiece during machining generates heat, which is transferred to the coolant through the machine tool's structure. The coolant, acting as a heat transfer medium, circulates through a coolant pump, carrying the heat to the radiator for dissipation. When analyzing the heat exchange path between the machining unit and the temperature control unit, it is necessary to determine the specific route and method of heat transfer. For example, factors such as the coolant's flow path within the machine tool and the radiator's heat dissipation efficiency can affect the effectiveness of heat exchange. Extracting the medium type of heat transfer relationship, such as the coolant type and specific heat capacity, as the basis for connecting heat transfer edges, accurately describes the heat exchange relationship between the machining unit and the temperature control unit, helping to promptly detect faults such as equipment overheating.

[0027] Step S115: Analyze the energy consumption allocation mechanism of the transmission unit and the temperature control unit, extract the energy conversion efficiency of the energy consumption coupling relationship, and use it as the connection basis of the energy consumption coupling edge.

[0028] Both the transmission unit and the temperature control unit consume energy during operation, creating an energy coupling relationship between them. In intelligent automotive engine manufacturing production lines, the transmission unit's motor consumes electricity to power it, while the temperature control unit's coolant pump and cooling fan also consume electricity to maintain operation. When analyzing the energy allocation mechanism between the transmission unit and the temperature control unit, it is necessary to determine the energy source and allocation method. For example, whether the transmission unit and the temperature control unit share the same power supply and what the power allocation ratio is. The energy conversion efficiency of the energy coupling relationship, such as the efficiency of the motor and the coolant pump, is extracted and used as the basis for connecting the energy coupling edge. This approach accurately reflects the energy relationship between the transmission unit and the temperature control unit, helping to optimize equipment energy management.

[0029] Step S116: creating component association edges for each pair of functionally associated device component nodes based on the connection basis, wherein the initial weights of the component association edges are determined by the tightness of the connection structure, the smoothness of the heat exchange path, and the energy conversion efficiency.

[0030] Based on the connection criteria for mechanical linkage, heat conduction, and energy consumption coupling obtained from the previous analysis, component association edges are created for each pair of functionally related equipment component nodes. For example, for the spindle of the cylinder machining center and the gear of the transmission system, a mechanical linkage edge is created based on the mechanical linkage relationship between them; for the cylinder machining center and the coolant pump, a heat conduction edge is created based on the heat conduction relationship between them; and for the transmission system motor and the cooling fan of the temperature control unit, an energy consumption coupling edge is created based on the energy consumption coupling relationship between them.

[0031] The initial weights of component-association edges are determined by the tightness of the connection structure, the smoothness of the heat exchange path, and the energy conversion efficiency. The tightness of the connection structure can be assessed by analyzing factors such as the strength of the mechanical connection and transmission efficiency. The smoothness of the heat exchange path can be assessed by factors such as the coolant flow resistance and the heat dissipation efficiency of the radiator. The energy conversion efficiency can be assessed by the efficiency of equipment such as motors and pumps. By comprehensively considering these factors, the initial weights of component-association edges are determined, ensuring that the equipment association diagram more accurately reflects the relationships between equipment components.

[0032] Step S117: Connectivity verification is performed on the device association graph so that the processing unit, transmission unit, and temperature control unit nodes form at least one functional association path through component association edges, thereby generating a complete device association graph.

[0033] After creating the device association diagram, you need to verify its connectivity. The purpose of this connectivity verification is to ensure that the machining unit, transmission unit, and temperature control unit nodes form at least one functional association path through component association edges. This means that any node in the device association diagram can be connected to other related nodes through component association edges. For example, the spindle node of a cylinder machining center can be connected to the gear node of the transmission system through a mechanical linkage edge, and then to the cooling fan node of the temperature control unit through an energy coupling edge, forming a complete functional association path.

[0034] By checking the connections between nodes in the device association graph, verify whether there are isolated nodes or disconnected subgraphs. If connectivity issues are found, re-examine the connection basis and the creation process of component association edges, making necessary adjustments and additions until connectivity requirements are met. Finally, a complete device association graph is generated.

[0035] Step S120: Acquire the operation monitoring data of each equipment component node, and perform node feature extraction processing on the operation monitoring data to generate the operating condition time series features of each equipment component node.

[0036] After constructing the equipment association diagram, it is necessary to obtain and process the operational monitoring data of each equipment component node to generate the operating condition time series characteristics of each equipment component node. In the intelligent manufacturing production line of automobile engines, sensors installed on each equipment component collect equipment operating status information in real time.

[0037] Step S121: performing time window division processing on the tool vibration signal of the machining unit to obtain a plurality of vibration data windows with continuous machining time period identifiers.

[0038] In a machining unit, tool vibration is a crucial indicator of its operating status. Vibration sensors installed on the tool collect the tool's vibration signals in real time. The collected tool vibration signals are then processed using time windowing, dividing the signals into multiple windows at set time intervals. Each window corresponds to a continuous machining period and is assigned an identifier. For example, the tool vibration signal is divided into time windows per minute, each with a unique identifier such as "T1" or "T2." This facilitates subsequent analysis and processing of vibration signals from different time periods.

[0039] Step S122: For each vibration data window, calculate the difference between the maximum and minimum amplitude values ​​of the vibration signal as the vibration amplitude parameter, count the number of changes in the vibration frequency per unit time as the vibration frequency fluctuation parameter, and identify the duration of time that the vibration signal amplitude is stable within the design threshold as the vibration stability duration parameter.

[0040] For each defined vibration data window, the relevant characteristic parameters need to be calculated. First, the difference between the maximum and minimum amplitude values ​​of the vibration signal is calculated to obtain the vibration amplitude parameter. The vibration amplitude parameter reflects the intensity of the tool's vibration during that period. For example, in a vibration data window, if the maximum amplitude value of the vibration signal is A1 and the minimum value is A2, the vibration amplitude parameter is A1-A2.

[0041] Next, count the number of vibration frequency changes per unit time to obtain the vibration frequency fluctuation parameter. This parameter reflects the stability of the tool's vibration frequency. For example, if the vibration frequency changes n times in one minute, the vibration frequency fluctuation parameter is n.

[0042] Finally, the duration during which the vibration signal amplitude remains stable within the designed threshold is identified to obtain the vibration stability duration parameter. This parameter reflects the tool's operating stability during that period. For example, if the vibration signal amplitude remains within the designed threshold for t seconds within a vibration data window, the vibration stability duration parameter is t.

[0043] Step S123: combining the vibration amplitude parameter, the vibration frequency fluctuation parameter, and the vibration stability time parameter to generate a vibration stability characteristic of the machining unit.

[0044] The calculated vibration amplitude parameters, vibration frequency fluctuation parameters, and vibration stability duration parameters are combined to generate a vibration stability signature for the machining unit. This vibration stability signature encompasses multiple aspects of tool vibration and provides a more comprehensive reflection of the machining unit's operating status. For example, the vibration stability signature can be represented as a three-dimensional vector (vibration amplitude parameter, vibration frequency fluctuation parameter, and vibration stability duration parameter). This vibration stability signature can be used to evaluate and analyze the tool's vibration stability.

[0045] Step S124: performing time window division processing on the gear torque signal of the transmission unit to obtain a plurality of torque data windows with continuous transmission period identifiers.

[0046] In a transmission unit, gear torque is a crucial indicator of its operating status. A torque sensor mounted on the gear shaft collects the gear torque signal in real time. The collected gear torque signal is then time-windowed, dividing the signal into multiple windows at set time intervals. Each window corresponds to a continuous transmission period and is assigned an identifier. For example, the gear torque signal can be divided into time windows per minute, with each window uniquely identified, such as "G1" or "G2." This facilitates subsequent analysis and processing of torque signals from different time periods.

[0047] Step S125: For each torque data window, calculate the difference between the peak value and the mean value of the torque signal as the torque fluctuation parameter, count the number of times the torque signal exceeds the design rated value as the torque excess number parameter, and identify the duration of time the torque signal mean value is stable within the design range as the torque stability duration parameter.

[0048] For each defined torque data window, the relevant characteristic parameters need to be calculated. First, the difference between the peak and mean values ​​of the torque signal is calculated to obtain the torque fluctuation parameter. The torque fluctuation parameter reflects the degree of torque variation of the gear during that period. For example, in a torque data window, if the peak value of the torque signal is T1 and the mean value is T2, the torque fluctuation parameter is T1-T2.

[0049] Next, the number of times the torque signal exceeds the design rating is counted to obtain the torque overrun parameter. This parameter reflects whether the gear is overloaded during that period. For example, if the torque signal exceeds the design rating m times in one minute, the torque overrun parameter is m.

[0050] Finally, the duration during which the mean value of the torque signal remains stable within the design range is determined, yielding the torque stability duration parameter. This parameter reflects the stability of the gear during this period. For example, if the mean value of the torque signal remains within the design range for s seconds within a certain torque data window, the torque stability duration parameter is s.

[0051] Step S126: combining the torque fluctuation parameter, the torque exceeding times parameter and the torque stabilization time parameter to generate a torque smoothness characteristic of the transmission unit.

[0052] The calculated torque fluctuation parameters, torque overrun count parameters, and torque stabilization duration parameters are combined to generate the transmission unit's torque smoothness characteristic. This torque smoothness characteristic incorporates information on multiple aspects of gear torque, providing a more comprehensive reflection of the transmission unit's operating status. For example, the torque smoothness characteristic can be represented as a three-dimensional vector (torque fluctuation parameter, torque overrun count parameter, and torque stabilization duration parameter). This torque smoothness characteristic can be used to evaluate and analyze the torque smoothness of the gear.

[0053] Step S127: performing time window division processing on the heat dissipation flow signal of the temperature control unit to obtain a plurality of flow data windows with continuous temperature control period identifiers.

[0054] In a temperature control unit, heat dissipation flow is a key indicator of its operating status. A flow sensor installed on the coolant pipe collects heat dissipation flow signals in real time. The collected heat dissipation flow signals are then time-windowed, dividing the signals into multiple windows at set time intervals. Each window corresponds to a continuous temperature control period and is identified. For example, the heat dissipation flow signal is divided into time windows per minute, and each window has a unique identifier, such as "C1," "C2," and so on. This facilitates subsequent analysis and processing of flow signals from different time periods.

[0055] Step S128: For each flow data window, calculate the difference between the instantaneous value of the flow signal and the set value as the flow deviation parameter, count the number of times the flow signal fluctuation amplitude exceeds the design threshold as the flow fluctuation number parameter, and identify the duration of the flow signal being stable within the set range above and below the set value as the flow stability duration parameter.

[0056] For each defined flow data window, the relevant characteristic parameters need to be calculated. First, the difference between the instantaneous value of the flow signal and the set value is calculated to obtain the flow deviation parameter. The flow deviation parameter reflects the degree to which the heat dissipation flow rate deviates from the set value during that period. For example, in a flow data window, if the instantaneous value of the flow signal is F1 and the set value is F0, the flow deviation parameter is F1-F0.

[0057] Next, count the number of times the flow signal fluctuation amplitude exceeds the design threshold to obtain the flow fluctuation parameter. The flow fluctuation parameter reflects the stability of the heat dissipation flow. For example, if the flow signal fluctuation amplitude exceeds the design threshold k times within one minute, the flow fluctuation parameter is k.

[0058] Finally, the duration of time the flow signal remains stable within a set range above or below the set value is identified, yielding the flow stability duration parameter. This parameter reflects the stability of the temperature control unit during that period. For example, if, within a flow data window, the flow signal remains within a set range above or below the set value for u seconds, the flow stability duration parameter is u.

[0059] Step S129: The flow deviation parameter, the flow fluctuation number parameter and the flow stabilization time parameter are combined to generate a flow continuity characteristic of the temperature control unit.

[0060] The calculated flow deviation parameters, flow fluctuation frequency parameters, and flow stabilization duration parameters are combined to generate the flow continuity characteristic of the temperature control unit. This flow continuity characteristic contains information on multiple aspects of the heat dissipation flow, providing a more comprehensive reflection of the temperature control unit's operating status. For example, the flow continuity characteristic can be represented as a three-dimensional vector (flow deviation parameter, flow fluctuation frequency parameter, flow stabilization duration parameter). This flow continuity characteristic can be used to evaluate and analyze the flow continuity of the temperature control unit.

[0061] Step S130: Correlate the vibration stability characteristics, torque smoothness characteristics, and flow continuity characteristics according to the equipment component node identifiers to generate operating condition time series characteristics of each equipment component node.

[0062] After generating the vibration stability characteristics of the machining unit, the torque smoothness characteristics of the transmission unit, and the flow continuity characteristics of the temperature control unit, these characteristics need to be associated according to the equipment component node identifier. In the intelligent manufacturing production line for automobile engines, each equipment component node has a unique identifier. For example, the tool node of the cylinder machining center is identified as "M1", the gear node of the transmission system is identified as "G1", and the coolant pump node of the temperature control unit is identified as "C1".

[0063] By associating the vibration stability characteristic with the corresponding machining unit equipment component node identifier, the torque smoothness characteristic with the corresponding transmission unit equipment component node identifier, and the flow continuity characteristic with the corresponding temperature control unit equipment component node identifier, each equipment component node has a corresponding operating condition time series characteristic, which contains information about the operating status of the equipment component in different aspects. For example, the operating condition time series characteristic of the tool node of a cylinder machining center can be expressed as (vibration stability characteristic, tool node identifier). This operating condition time series characteristic allows for more comprehensive monitoring and analysis of the tool's operating status.

[0064] Step S130: performing edge feature extraction processing on the operation monitoring data of the equipment component nodes at both ends of the component association edge to generate collaborative change features of the component association edge.

[0065] After obtaining the operational monitoring data and operating condition time series characteristics of each equipment component node, it is necessary to perform edge feature extraction on the operational monitoring data of the equipment component nodes at both ends of the component-association edge to generate the coordinated change characteristics of the component-association edge. In the intelligent manufacturing production line of automobile engines, component-association edges include mechanical linkage edges, heat conduction edges, and energy consumption coupling edges.

[0066] Step S131: extracting the working condition time sequence characteristics of the processing units and the transmission units at both ends of the mechanical linkage edge, wherein the working condition time sequence characteristics include vibration stability characteristics and torque smoothness characteristics.

[0067] For mechanical linkage edges, extract the time series characteristics of the machining units and transmission units at their ends. In the previous steps, the vibration stability characteristics of the machining units and the torque smoothness characteristics of the transmission units were generated. For example, if a mechanical linkage edge connects the tool node of a cylinder machining center and the gear node of the transmission system, extract the vibration stability characteristics of the tool node and the torque smoothness characteristics of the gear node. These characteristics reflect the state changes of the machining units and transmission units during operation.

[0068] Step S132: Locate the mutation time point of the vibration amplitude parameter of the machining unit, locate the mutation time point of the torque fluctuation parameter of the transmission unit, and calculate the difference between the two mutation time points as the load transfer delay parameter of the mechanical linkage relationship.

[0069] After extracting the time-series characteristics of the machining and transmission units at both ends of the mechanical linkage, further analysis is needed to identify any sudden changes in these characteristics. For example, in an intelligent automotive engine manufacturing production line, the tool vibration signals of the machining units are monitored and analyzed for their amplitude parameters to identify the time points when these parameters suddenly change. For example, if the tool vibration amplitude is initially relatively stable but suddenly increases or decreases significantly at a certain moment, this is the time point when the vibration amplitude parameter suddenly changes. Similarly, for the gear torque signals of the transmission unit, the time points when the torque fluctuation parameters suddenly change are identified.

[0070] The difference between these two mutation time points is calculated to obtain the load transfer delay parameter of the mechanical linkage relationship. This load transfer delay parameter reflects the time delay between the load change of the machining unit and the response of the transmission unit during the mechanical linkage process. For example, if the vibration amplitude of the machining unit's tool suddenly increases due to encountering hard material, the transmission unit needs a certain amount of time to adjust the torque to this load change. The load transfer delay parameter reflects this adjustment time. This load transfer delay parameter is very important for evaluating the efficiency and stability of the mechanical linkage. If the load transfer delay is too long, it may lead to increased wear of the equipment or reduced machining accuracy.

[0071] Step S133: extracting the operating time sequence characteristics of the processing units and the temperature control unit at both ends of the heat conduction edge, wherein the operating time sequence characteristics include vibration stability characteristics and flow continuity characteristics.

[0072] For heat conduction edges, extract the time series characteristics of the operating conditions of the machining units and temperature control units at both ends. The vibration stability characteristics of the machining units and the flow continuity characteristics of the temperature control units were obtained in the previous steps. For example, the cylinder machining center and coolant circulation system in an intelligent automotive engine manufacturing production line generate heat during the machining process. Its vibration stability characteristics reflect the stability of the machining process and the heat generation. The coolant circulation system removes heat through the flow of coolant, and its flow continuity characteristics reflect the heat dissipation capacity and stability. Extracting these two characteristics allows for further analysis of the synergistic changes between the equipment components at both ends of the heat conduction edge.

[0073] Step S134: Locate the falling time point of the vibration stabilization time parameter of the processing unit, locate the falling time point of the flow stabilization time parameter of the temperature control unit, and calculate the difference between the two falling time points as the temperature response hysteresis parameter of the heat conduction relationship.

[0074] After obtaining the operating time characteristics of the processing units and temperature control units at both ends of the heat conduction edge, the key parameters are analyzed. For the vibration stability time parameter of the processing unit, its decline is monitored and the time point of decline is located. When the vibration stability time parameter of the processing unit declines, it indicates that the processing process may be unstable, which may lead to an increase in heat generation. Similarly, for the flow stability time parameter of the temperature control unit, the time point of its decline is located. When the flow stability time parameter declines, it indicates that the heat dissipation capacity of the temperature control unit may be affected.

[0075] The difference between these two drop time points is calculated to obtain the temperature response hysteresis parameter of the heat conduction relationship. This temperature response hysteresis parameter reflects the time delay from the heat generation change in the machining unit to the temperature control unit's response and adjustment of heat dissipation. For example, if the vibration stability time of the machining unit decreases due to tool wear, the temperature control unit needs a certain amount of time to adjust the coolant flow to adapt to the heat change. The temperature response hysteresis parameter reflects this adjustment time. This temperature response hysteresis parameter is very important for evaluating the efficiency of heat conduction and the thermal stability of the equipment. If the temperature response hysteresis is too long, it may cause the equipment to overheat, affecting the equipment life and processing quality.

[0076] Step S135: extracting the operating time series characteristics of the transmission unit and the temperature control unit at both ends of the energy consumption coupling edge, wherein the operating time series characteristics include torque smoothness characteristics and flow continuity characteristics.

[0077] For energy-coupling edges, the operating time series characteristics of the transmission unit and temperature control unit at either end are extracted. In intelligent automotive engine manufacturing production lines, the torque smoothness characteristics of the transmission unit reflect the stability and energy consumption of its power transmission, while the flow continuity characteristics of the temperature control unit reflect the stability and energy consumption of its heat dissipation. For example, if the transmission unit's gears experience large torque fluctuations when transmitting torque, they can consume more energy; if the temperature control unit's coolant pump maintains coolant flow and the flow is unstable, energy consumption will also increase. These two features are extracted to analyze the coordinated changes between the equipment components at both ends of the energy-coupling edge.

[0078] Step S136: Calculate the ratio of the transmission unit torque overrun times parameter to the temperature control unit flow rate fluctuation times parameter as the energy conversion deviation parameter of the energy consumption coupling relationship.

[0079] After extracting the time series characteristics of the operating conditions of the transmission unit and temperature control unit at both ends of the energy consumption coupling edge, we further calculated relevant parameters. For the transmission unit's torque signal, we counted its torque overrun parameter, which measures the number of times the torque exceeded the design rating. For the temperature control unit's flow signal, we counted its flow fluctuation parameter, which measures the number of times the flow fluctuation amplitude exceeded the design threshold.

[0080] The ratio of these two parameters is calculated to obtain the energy conversion deviation parameter for the energy consumption coupling relationship. This energy conversion deviation parameter reflects the degree of matching between the energy conversion between the transmission unit and the temperature control unit during the energy consumption coupling process. For example, if the transmission unit frequently experiences torque overruns while the temperature control unit experiences few flow fluctuations, this indicates possible energy distribution discrepancies, and the energy conversion deviation parameter reflects this degree of deviation. This energy conversion deviation parameter is crucial for assessing the efficiency and energy-saving potential of energy consumption coupling. Excessive energy conversion deviation can lead to energy waste and increased equipment operating costs.

[0081] Step S137: Normalize the load transfer delay parameter, temperature response lag parameter, and energy conversion deviation parameter, and perform weighted fusion processing on the normalized parameters and the initial weights of the component association edges to generate dynamically updated collaborative change features. The collaborative change features establish a binding relationship with the corresponding component association edges in the device association graph through the association edge identifier.

[0082] After obtaining the load transfer delay parameter for the mechanical linkage relationship, the temperature response hysteresis parameter for the heat conduction relationship, and the energy conversion deviation parameter for the energy consumption coupling relationship, these parameters need to be normalized. The purpose of normalization is to map these parameters to a common scale for subsequent weighted fusion. For example, the load transfer delay parameter, temperature response hysteresis parameter, and energy conversion deviation parameter may have different value ranges and dimensions. Normalization can eliminate these differences and make them comparable.

[0083] Normalization can be performed using a common linear normalization method to map the value of each parameter to the range of 0 to 1. Specifically, for each parameter, first find its maximum and minimum values, then subtract the minimum value from each value, and then divide by the difference between the maximum and minimum values ​​to obtain the normalized parameter.

[0084] The normalized load transfer delay parameter, temperature response hysteresis parameter, and energy conversion deviation parameter are weighted and fused with the initial weights of the component-association edges. The initial weights of the component-association edges are determined earlier when constructing the device association graph and reflect the initial strength of the associations between device components. Weighted fusion involves assigning different weights to each parameter based on its importance and then fusing them with the initial weights. For example, different weights can be assigned to the load transfer delay parameter, temperature response hysteresis parameter, and energy conversion deviation parameter, and then weighted summed with the initial weights to produce a dynamically updated collaborative variation feature.

[0085] Finally, the generated collaborative change signature is bound to the corresponding component edge in the device association graph through the edge identifier. In the device association graph, each component edge has a unique identifier. Associating the collaborative change signature with the edge identifier allows accurate acquisition of the collaborative change information for each component edge in subsequent fault propagation analysis.

[0086] Step S140: Call the pre-trained graph reasoning model to perform fault propagation analysis and processing on the equipment association graph, the operating condition timing characteristics and the coordinated change characteristics, and generate a fault propagation result including the abnormal originating node identifier and the propagation path sequence.

[0087] After constructing the device association graph, extracting operating condition time series features, and generating coordinated change features, a pre-trained graph reasoning model is used to analyze and process this information for fault propagation. A graph reasoning model is a deep learning model based on graph structures that effectively processes node and edge information in the device association graph and discovers patterns in fault propagation between devices.

[0088] Step S141: Input the node structure information of the equipment association graph, the operating condition timing characteristics of each equipment component node, and the collaborative change characteristics of each component association edge into the input layer of the graph reasoning model, perform feature dimension standardization processing, and generate an initial graph representation that meets the model input requirements.

[0089] The node structure information of the equipment association graph, the operating time series characteristics of each equipment component node, and the coordinated change characteristics of each component's associated edges are input into the input layer of the graph reasoning model. In the intelligent manufacturing production line for automobile engines, the node structure information of the equipment association graph includes the number of nodes and the connection relationships between nodes. The operating time series characteristics of each equipment component node contain information about the equipment's operating status in different aspects, and the coordinated change characteristics of each component's associated edges reflect the collaborative working status of the equipment.

[0090] Before inputting into the graph inference model, the aforementioned features need to be dimensionally normalized. Because different features may have different dimensions and value ranges, their dimensions need to be unified to enable the model to better process them. For example, for operating condition time series features and coordinated change features, some feature vectors may have higher dimensions, while others may have lower dimensions. Dimensional normalization can bring these dimensions into line.

[0091] After dimension normalization, an initial graph representation that meets the model input requirements is generated. This initial graph representation integrates the structural information of the equipment association graph, the operating condition timing characteristics of the nodes, and the coordinated change characteristics of the edges.

[0092] Step S142: Perform information fusion processing on the initial graph representation through the neighborhood aggregation layer of the graph reasoning model to obtain neighborhood fusion features.

[0093] The neighborhood aggregation layer of the graph reasoning model performs information fusion on the initial graph representation. In the device association graph, each node has its neighboring nodes, and the information about these neighboring nodes is crucial for understanding the current node's state. In the neighborhood aggregation layer, the operating time series features of the current device component node are first obtained as the central node features. Simultaneously, the operating time series features of all neighboring nodes of the current node are obtained as the neighbor node feature set.

[0094] For example, in the equipment association diagram of an intelligent manufacturing production line for an automobile engine, the adjacent nodes of the tool node of a cylinder machining center might include the gear node of a transmission unit and the coolant pump node of a temperature control unit. The operating time series features of the tool node are obtained as the central node features, while the operating time series features of the gear node and the coolant pump node are obtained as the neighbor node feature set.

[0095] Perform a linear transformation on the central node features to generate central node transformed features. The purpose of linear transformation is to adjust and transform the features to make them more suitable for subsequent calculations and fusion. Similarly, perform a linear transformation on each neighbor node feature in the neighbor node feature set to generate a neighbor node transformed feature set.

[0096] The collaborative change features of the component-association edges between the current node and each adjacent node are obtained as an edge feature set. Each edge feature in the edge feature set is linearly transformed to generate an edge transformation feature set. The neighbor node transformation feature set is then multiplied by the corresponding element of the edge transformation feature set to generate a weighted neighbor node transformation feature set. This process takes into account the strength of the association between adjacent nodes. By weighting the edge features, the information of the neighbor nodes is appropriately reflected in the fusion process.

[0097] Aggregate the weighted transformed features of neighboring nodes to generate neighborhood aggregate features. Aggregation can be performed using methods such as summation and averaging to integrate information from adjacent nodes. Finally, add the central node's transformed features to the neighborhood aggregate features and apply an activation function to generate updated node features for the current device component node.

[0098] Traversing all device component nodes in the device association graph, we repeatedly perform the aforementioned central node feature processing, neighbor node feature processing, and feature aggregation operations to generate a graph representation that completes neighborhood information fusion. Through the neighborhood aggregation layer, each node incorporates information from its neighboring nodes, resulting in richer and more accurate neighborhood fusion features.

[0099] Step S143: The attention enhancement layer of the graph reasoning model is used to perform importance evaluation processing on the neighborhood fusion features to generate attention-enhanced node features.

[0100] After obtaining the neighborhood fusion features, the graph reasoning model's attention enhancement layer performs an importance assessment on these features. The role of the attention enhancement layer is to identify which information in the neighborhood fusion features is more important for fault diagnosis of the current node, thereby highlighting this important information.

[0101] In the equipment association graph of an intelligent automotive engine manufacturing production line, different neighborhood information may have varying degrees of influence on each node. For example, for a tool node in a cylinder machining center, changes in the transmission unit's torque may have a significant impact on its operating state, while changes in certain minor parameters of the temperature control unit may have a smaller impact. The attention enhancement layer analyzes and evaluates neighborhood fusion features and assigns an attention weight to each feature.

[0102] Specifically, the attention enhancement layer calculates the importance score of each feature based on the relationship between the feature vectors of the neighboring fused features. This importance score reflects the importance of the feature to the current node. Based on this importance score, each feature is then assigned an attention weight, with a higher weight indicating a more important feature.

[0103] The attention weight is multiplied by the neighborhood fusion feature to obtain the attention-enhanced node feature. This process highlights important information and suppresses secondary information, allowing the model to focus more on key features and improve the accuracy of fault diagnosis.

[0104] Step S144: Dynamically analyzing and processing the graph representation of the continuous time periods through the temporal modeling layer of the graph reasoning model to generate a temporal context feature reflecting the state evolution of the device association graph.

[0105] The temporal modeling layer of the graph reasoning model is used to analyze the dynamic evolution of graph representations of continuous time periods. In the intelligent manufacturing production line of automobile engines, the operating status of equipment changes over time, so information in the time dimension needs to be considered.

[0106] First, the graph representations of consecutive time periods are arranged in chronological order to generate a graph representation time series. Each graph representation contains the structural information, node features, and edge features of the device association graph at a specific moment. For example, during a day's production process, equipment operating data is collected at regular intervals to generate corresponding graph representations. These graph representations are then arranged in chronological order to form a graph representation time series.

[0107] Feed the graph representation of the time series into the input gate of a long short-term memory (LSTM) network. LSTM is a commonly used time series model that effectively handles long-term dependencies in sequential data. The input gate controls the input level of the graph representation for the current time period. By calculating the activation value of the input gate, we determine which information is allowed to enter the LSTM network for processing.

[0108] Calculate the activation value of the LSTM's forget gate. The forget gate controls the degree of retention of the graph representation of historical periods. When processing the graph representation of the current period's time series, the LSTM needs to decide which historical information to retain and which to forget. The forget gate activation value is calculated based on the current period's input and the historical hidden state.

[0109] Calculate the candidate memory cell value of LSTM. The candidate memory cell value is determined by the current period graph representation and the historical hidden state, which represents the new information that may need to be stored in the current period.

[0110] Update the LSTM cell state, which is the part of the LSTM network used to store long-term information. The cell state is updated by multiplying the forget gate activation value by the historical cell state, plus the input gate activation value by the candidate memory cell value. This process takes into account both the preservation of historical information and the addition of new information.

[0111] Calculate the activation value of the LSTM output gate. The output gate controls the output level of the current cell state. Based on the output gate activation value, determine which information can be output as the hidden state of the current period.

[0112] Generate the hidden state of the current period. The hidden state is obtained by multiplying the output gate activation value and the hyperbolic tangent function value of the cell state. This hidden state contains the comprehensive information of the graph representation of the current period and historical information.

[0113] By traversing all time periods of the time series represented by the graph, the aforementioned input gate calculation, forget gate calculation, and hidden state update operations are repeated to generate a temporal context feature containing temporal dimension information. This temporal context feature reflects the state evolution of the device association graph over consecutive time periods.

[0114] Step S145: The output layer of the graph reasoning model performs abnormality probability prediction processing on the time context features to generate abnormality probability values ​​of each device component node and abnormality propagation probability values ​​between nodes.

[0115] After obtaining temporal context features reflecting the state evolution of the device association graph, the output layer of the graph reasoning model processes these features to predict anomaly probabilities. The output layer is the final layer of the graph reasoning model. Based on the information processed by the previous layers, it predicts the anomaly probability of each device component node and the probability of anomaly propagation between nodes.

[0116] In an intelligent automotive engine manufacturing production line, for each equipment component node, the output layer predicts the probability of failure based on information from the temporal context features. For example, for a tool node in a cylinder machining center, the output layer comprehensively considers the tool's vibration stability characteristics, neighborhood information, and temporal evolution to predict its abnormality probability.

[0117] Similarly, for the relationships between nodes in the device association graph, the output layer predicts the probability of an anomaly propagating from one node to another. For example, it predicts the probability of an anomaly propagating from the gear node of the transmission unit to the tool node of the machining unit. This anomaly propagation probability reflects the likelihood of a fault propagating within the device association graph.

[0118] The output layer analyzes and processes the temporal context features, and uses the built-in neural network structure and trained parameters to calculate the abnormality probability value of each equipment component node and the abnormality propagation probability value between nodes.

[0119] Step S146: Based on the abnormal probability value, identify the equipment component node whose abnormal probability exceeds the preset threshold as a candidate abnormal originating node, construct a propagation path tree of the candidate abnormal originating node based on the abnormal propagation probability value, select the path with the highest confidence as the final propagation path sequence through path confidence evaluation, and generate a fault propagation result including the abnormal originating node identifier and the propagation path sequence.

[0120] After obtaining the anomaly probability values ​​for each device component node and the anomaly propagation probability values ​​between nodes, candidate anomaly originating nodes are identified and propagation paths analyzed. First, a preset threshold is set. Device component nodes with an anomaly probability exceeding this threshold are identified as candidate anomaly originating nodes. This threshold, determined based on extensive historical data and experience, is used to screen out nodes that may be at risk of failure.

[0121] In the equipment association diagram of the automobile engine intelligent manufacturing production line, for example, if the abnormal probability of the tool node of the cylinder machining center exceeds the preset threshold, the tool node is identified as a candidate abnormal originating node.

[0122] Based on the anomaly propagation probability, a propagation path tree is constructed for each candidate anomaly originating node. The propagation path tree is a tree-like structure that starts from the candidate anomaly originating node and expands outward to possible propagation paths based on the anomaly propagation probability between nodes. For example, if the tool node is a candidate anomaly originating node, based on the anomaly propagation probability, it may propagate the fault to the gear node of the transmission unit or the coolant pump node of the temperature control unit, and so on.

[0123] A path confidence assessment is performed on each path in the constructed propagation path tree. This assessment takes into account factors such as the probability of anomalies at each node on the path and the probability of anomaly propagation between nodes. For example, if the probability of anomalies at all nodes on a path is high, and the probability of anomaly propagation between nodes is also high, then the confidence of this path is high.

[0124] By comparing the confidence levels of each path, the path with the highest confidence level is selected as the final propagation path sequence. This ultimately generates a fault propagation result that includes the abnormal originating node identifier and the propagation path sequence. This fault propagation result identifies the possible starting point and propagation direction of the fault.

[0125] Step S150: Based on the fault propagation result, the historical fault case library is matched to locate the root cause component that causes the equipment abnormality, and an operation and maintenance plan including a maintenance operation sequence and parameter adjustment instructions is generated according to the root cause component, and the operation and maintenance plan is pushed to the equipment control system to trigger the fault repair operation.

[0126] Step S151: matching a historical fault case library based on the fault propagation result to locate the root cause component that causes the equipment abnormality.

[0127] After obtaining the fault propagation results, which include the abnormal originating node identifier and propagation path sequence, the results must be matched against a historical fault case database to locate the root cause of the equipment abnormality. The historical fault case database stores information on various types of past equipment failures, including equipment component type, relationship type, fault symptoms, and the ultimately confirmed root cause component.

[0128] Step S1511: extract the abnormal originating node identifier in the fault propagation result, and obtain the device component type corresponding to the abnormal originating node.

[0129] In the context of an intelligent automotive engine manufacturing production line, the abnormal origin node identifier is extracted from the fault propagation results. For example, if the abnormal origin node identifier indicates that the node is a tool node in a cylinder machining center, then the corresponding equipment component type is the tool of the machining unit. This information is one of the important bases for subsequent matching historical fault cases, as the causes and manifestations of failures of different equipment components may vary.

[0130] Step S1512: extract the first associated edge identifier in the propagation path sequence, and obtain the associated relationship type corresponding to the edge.

[0131] At the same time, the first edge identifier in the propagation path sequence is extracted. In the device association graph, edges represent different relationships between device components, such as mechanical linkage edges, thermal conduction edges, and energy coupling edges. If the first edge identifier corresponds to a mechanical linkage edge, then the corresponding relationship type is a mechanical linkage relationship. This relationship type is also key information for matching historical fault cases, helping to narrow the search scope and more accurately locate the root cause component.

[0132] Step S1513: traverse the historical fault case library to search for historical fault cases containing the same equipment component type and association relationship type, and count the root cause components finally confirmed in the historical cases.

[0133] Based on the extracted equipment component type corresponding to the abnormal originating node and the association type corresponding to the first edge of the propagation path, the historical fault case library is traversed. During the traversal process, all historical fault cases containing the same equipment component type and association type are searched. For each historical fault case found, the root cause component that was ultimately confirmed is counted. For example, when the abnormal originating node is a tool in a machining unit and the first association edge is a mechanical linkage edge, the historical fault case library is searched for historical fault cases involving machining unit tools and mechanical linkage relationships, and the components ultimately determined to be the root cause components in these cases are counted.

[0134] Step S1514: Calculate the matching degree between the current fault propagation result and the historical fault case. The matching degree is determined by the consistency of the abnormal originating node type, the consistency of the propagation path head edge type, and the closeness of the coordinated change characteristics.

[0135] After finding historical fault cases that meet the requirements, the degree of match between the current fault propagation results and these historical fault cases needs to be calculated. The matching calculation takes into account multiple factors. First, the consistency of the anomaly originating node type. Specifically, is the type of the anomaly originating node of the current fault the same as that of the anomaly originating node in the historical fault case? If so, it is given a higher weight in the matching calculation. Second, the consistency of the leading edge type of the propagation path. Similarly, if the leading edge type of the current fault propagation path is consistent with that of the leading edge in the historical fault case, the matching degree is increased. Finally, the degree of similarity between the co-variation characteristics is considered. The co-variation characteristics of the component-related edges have been generated in the previous step. The degree of similarity between the co-variation characteristics of the current fault and those of the historical fault cases is compared. For example, the load transfer delay parameter of the mechanical linkage relationship, the temperature response hysteresis parameter of the heat conduction relationship, and the energy conversion deviation parameter of the energy consumption coupling relationship can be calculated to determine the degree of similarity and factor them into the matching calculation. Combining these three factors, the matching degree between the current fault propagation results and each historical fault case is calculated.

[0136] Step S1515: Select the root cause component recorded in the historical fault case with the highest matching degree as the candidate root cause component.

[0137] Based on the calculated matching degree, the case with the highest matching degree is selected from the historical fault cases. The root cause component recorded in this case is selected as the candidate root cause component. For example, if the matching degree calculation finds that a historical fault case has the highest matching degree with the current fault propagation results, and the root cause component ultimately confirmed in this case is a gear in the transmission unit, then the gear becomes a candidate root cause component.

[0138] Step S1516: Verify the degree of deviation between the design operating parameters of the candidate root cause component and the current operating monitoring data, where the degree of deviation is calculated by the machining unit vibration amplitude parameter deviation, the transmission unit torque fluctuation parameter deviation, and the temperature control unit flow deviation parameter.

[0139] After the candidate root cause components are determined, they need to be verified. The verification method is to calculate the degree of deviation between the design operating parameters of the candidate root cause components and the current operating monitoring data. In the intelligent manufacturing production line of automobile engines, different types of equipment components have different operating parameters. For the processing unit, pay attention to the deviation of the vibration amplitude parameter; for the transmission unit, pay attention to the deviation of the torque fluctuation parameter; for the temperature control unit, pay attention to the flow deviation parameter. These deviation parameters are combined and calculated to obtain the degree of deviation of the candidate root cause components. For example, if the candidate root cause component is the gear of the transmission unit, it is necessary to calculate the deviation between its design torque and the torque fluctuation parameter of the current operating monitoring, and at the same time combine the deviation of the relevant parameters of the processing unit and the temperature control unit to comprehensively evaluate the degree of deviation.

[0140] Step S1517: If the degree of deviation exceeds the preset verification threshold, the candidate root cause component is confirmed to be the root cause component; if not, the historical failure case with the second highest matching degree is selected to repeat the verification operation until a root cause component that meets the verification threshold is found.

[0141] A preset verification threshold is set to determine whether the candidate root cause component is the true root cause component. If the calculated deviation exceeds the preset verification threshold, it indicates that there is a significant difference between the operating status of the candidate root cause component and the designed status, and it is likely the root cause component that caused the equipment abnormality. At this time, the candidate root cause component is confirmed as the root cause component. If the deviation does not exceed the preset verification threshold, it indicates that the candidate root cause component may not be the true root cause component. It is necessary to select the historical failure case with the next highest match and repeat the above verification operation, that is, recalculate the deviation between the design operating parameters of the root cause component in this case and the current operating monitoring data, until a root cause component that meets the verification threshold is found.

[0142] Step S152: Generate an operation and maintenance plan including a maintenance operation sequence and parameter adjustment instructions based on the root cause component, and push the operation and maintenance plan to the equipment control system to trigger a fault repair operation.

[0143] After locating the root cause, an O&M plan is generated based on that component and pushed to the equipment control system to initiate fault repair. The O&M plan includes a maintenance operation sequence and parameter adjustment instructions to help maintenance personnel efficiently and accurately repair the fault.

[0144] Step S1521: Obtain maintenance history records of the root cause component, and calculate the historical failure frequency and average repair time of the root cause component as a basis for evaluating the maintenance operation sequence.

[0145] In an intelligent automotive engine manufacturing production line, maintenance history records are collected for identified root-cause components. These records include information such as the time of past failures and repairs. Statistical analysis of these records reveals the historical failure frequency of the root-cause component—the number of times the component has failed within a certain period of time. Furthermore, the repair time for each component failure is calculated, and the average repair time is calculated. These two metrics serve as important criteria for evaluating the maintenance operation sequence. For example, if a root-cause component has a high historical failure frequency and a long average repair time, it may warrant prioritization when scheduling maintenance operations.

[0146] Step S1522: Calculate the expansion rate of the fault impact range based on the propagation speed of the propagation path sequence.

[0147] Combined with the propagation path sequence in the fault propagation results, its propagation speed is analyzed. The propagation speed reflects how quickly the fault propagates from the fault-originating node to other nodes. In the device association graph, each node on the propagation path is associated with other nodes, and the fault gradually spreads along the associated edges. Based on the propagation speed of the propagation path sequence, the expansion rate of the fault's impact range is calculated. For example, a faster propagation speed will also increase the expansion rate of the fault's impact range, potentially affecting more equipment components in a shorter period of time. This expansion rate is important for determining the urgency of maintenance operations.

[0148] Step S1523: Determine the maintenance priority of the root cause component according to the historical fault frequency, average repair time, and fault impact range expansion rate using a preset priority evaluation rule.

[0149] Pre-set priority assessment rules are used to determine the maintenance priority of a root-cause component based on its historical failure frequency, mean time to repair, and rate of expansion of the fault's impact area. These rules are developed based on extensive historical data and experience, comprehensively considering the impact of these three factors on equipment operation. For example, a root-cause component with a high historical failure frequency, a long mean time to repair, and a rapid rate of expansion of the fault's impact area may be given a higher maintenance priority to expedite repair and minimize the impact on production.

[0150] Step S1524: extract the design and maintenance manual of the root cause component, obtain the standard maintenance process of the root cause component, the list of required replacement parts and tool preparation requirements as the basic content of the parameter adjustment guide.

[0151] Relevant information is extracted from the design and maintenance manual of the root-causing component. This manual provides detailed instructions for component maintenance, including standard maintenance procedures, a list of required replacement parts, and tool preparation requirements. This information forms the basis for parameter adjustment instructions and provides basic operational guidance for maintenance personnel. For example, for the gear root-causing component of a transmission unit, the design and maintenance manual will detail the gear removal and installation procedures, the model and quantity of gears to be replaced, and the tools required for repair.

[0152] Step S1525: Adaptively adjust the standard maintenance process based on the abnormal parameter characteristics in the current operation monitoring data to obtain an adjusted maintenance process.

[0153] Adaptively adjust the standard maintenance process based on the abnormal parameter characteristics in the current operation monitoring data. The current operation monitoring data contains real-time operating status information of equipment components, such as the vibration amplitude parameters of the processing unit, the torque fluctuation parameters of the transmission unit, the flow deviation parameters of the temperature control unit, etc. Based on these abnormal parameter characteristics, the specific fault conditions of the root cause components can be determined, so that targeted adjustments can be made to the standard maintenance process. For example, if the monitoring data shows that the torque fluctuation parameters of the transmission unit have increased abnormally, it may mean that the gears are worn or damaged. On the basis of the standard maintenance process, it may be necessary to add detailed inspection and measurement steps for the gears, or prepare in advance the gear spare parts that may need to be replaced.

[0154] Step S1526: Combining the maintenance priority, the adjusted maintenance process, the required replacement parts list and the tool preparation requirements to generate an operation and maintenance plan.

[0155] The determined maintenance priority, adjusted maintenance procedures, a list of required replacement parts, and tool preparation requirements are combined to generate a complete maintenance plan. This plan details the requirements and steps for maintaining the root-cause component, providing clear operational guidance for maintenance personnel. For example, the maintenance plan will clearly indicate that the root-cause component has a high maintenance priority, provide a detailed adjusted maintenance procedure, and list the required replacement parts and tool preparation requirements, enabling maintenance personnel to carry out targeted maintenance preparations and operations.

[0156] Step S1527: Send the operation and maintenance plan to the equipment control system, so that the equipment control system allocates maintenance resources according to maintenance priorities and performs fault isolation and component repair operations according to the adjusted maintenance process.

[0157] The generated O&M plan is sent to the equipment control system. Upon receiving the plan, the control system allocates maintenance resources based on maintenance priorities. Root-causing components with high maintenance priorities are prioritized with more maintenance personnel, tools, spare parts, and other resources to ensure timely repairs. Simultaneously, the control system performs fault isolation and component repair according to the adjusted maintenance process. For example, the faulty equipment is first isolated to prevent further spread of the fault. The root-causing component is then repaired and replaced according to the adjusted maintenance process, ultimately repairing the equipment and restoring normal production line operation.

[0158] Figure 2 A schematic diagram illustrates exemplary hardware and software components of a fault diagnosis system 100 for intelligent manufacturing system operation and maintenance, provided in some embodiments of the present application, that can implement the concepts of the present application. For example, a processor 120 can be used in the fault diagnosis system 100 for intelligent manufacturing system operation and maintenance and perform the functions of the present application.

[0159] The fault diagnosis system 100 for intelligent manufacturing system operation and maintenance can be a general-purpose server or a special-purpose server, both of which can be used to implement the fault diagnosis method for intelligent manufacturing system operation and maintenance of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0160] For example, the fault diagnosis system 100 for intelligent manufacturing system operation and maintenance may include a network port 110 connected to the network, one or more processors 120 for executing program instructions, a communication bus 130, and different forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the fault diagnosis system 100 for intelligent manufacturing system operation and maintenance may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application can be implemented according to these program instructions. The fault diagnosis system 100 for intelligent manufacturing system operation and maintenance also includes an I / O interface 150 between the computer and other input and output devices.

[0161] For ease of explanation, only one processor is described in the fault diagnosis system 100 for intelligent manufacturing system operation and maintenance. However, it should be noted that the fault diagnosis system 100 for intelligent manufacturing system operation and maintenance in the present application may also include multiple processors, so the steps performed by one processor described in the present application may also be performed jointly or individually by multiple processors. For example, if the processor of the fault diagnosis system 100 for intelligent manufacturing system operation and maintenance executes step A and step B, it should be understood that step A and step B may also be executed jointly by two different processors or individually in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor execute steps A and B together.

[0162] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer-executable instructions are preset. When a processor executes the computer-executable instructions, the above-mentioned fault diagnosis method for operation and maintenance of an intelligent manufacturing system is implemented.

[0163] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, multiple features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A fault diagnosis method for intelligent manufacturing system operation and maintenance, characterized in that: The method comprises: Constructing a device association graph, wherein the device association graph is composed of device component nodes and component association edges; Acquire the operation monitoring data of each equipment component node, and perform node feature extraction processing on the operation monitoring data to generate the working condition time series features of each equipment component node; Performing edge feature extraction processing on the operation monitoring data of the equipment component nodes at both ends of the component association edge to generate collaborative change features of the component association edge; Calling a pre-trained graph reasoning model to perform fault propagation analysis and processing on the equipment association graph, the operating condition time series characteristics, and the coordinated change characteristics, and generating a fault propagation result including an abnormal originating node identifier and a propagation path sequence; Based on the fault propagation results, the historical fault case library is matched to locate the root cause component that caused the equipment abnormality, and an operation and maintenance plan including a maintenance operation sequence and parameter adjustment instructions is generated according to the root cause component, and the operation and maintenance plan is pushed to the equipment control system to trigger the fault repair operation; The calling of the pre-trained graph reasoning model performs fault propagation analysis and processing on the equipment association graph, the operating condition time series characteristics, and the coordinated change characteristics to generate a fault propagation result including an abnormal originating node identifier and a propagation path sequence, including: Input the node structure information of the equipment association graph, the operating condition time series characteristics of each equipment component node, and the coordinated change characteristics of the associated edges of each component into the input layer of the graph reasoning model, perform feature dimension standardization processing, and generate an initial graph representation that meets the model input requirements; The initial graph representation is processed through the neighborhood aggregation layer of the graph reasoning model to obtain the neighborhood fusion feature; The attention enhancement layer of the graph reasoning model performs importance evaluation on the neighborhood fusion features to generate attention-enhanced node features; The time series modeling layer of the graph reasoning model dynamically analyzes and processes the graph representation of continuous time periods to generate time context features that reflect the state evolution of the device association graph. The output layer of the graph reasoning model performs abnormality probability prediction processing on the time context features to generate abnormality probability values ​​of each device component node and abnormality propagation probability values ​​between nodes; Based on the abnormality probability value, equipment component nodes whose abnormality probability exceeds a preset threshold are identified as candidate abnormality originating nodes; based on the abnormality propagation probability value, a propagation path tree of the candidate abnormality originating nodes is constructed; through path confidence evaluation, the path with the highest confidence is selected as the final propagation path sequence; and a fault propagation result including the abnormality originating node identifier and the propagation path sequence is generated; The matching of the historical fault case library based on the fault propagation result to locate the root cause component causing the equipment abnormality includes: Extracting the abnormal originating node identifier from the fault propagation result, and obtaining the device component type corresponding to the abnormal originating node; Extracting the first associated edge identifier in the propagation path sequence and obtaining the associated relationship type corresponding to the edge; Traverse the historical fault case library to find historical fault cases containing the same equipment component type and association type, and count the root cause components finally confirmed in the historical cases; Calculate the matching degree between the current fault propagation result and the historical fault cases. The matching degree is determined by the consistency of the abnormal originating node type, the consistency of the propagation path head edge type, and the closeness of the coordinated change characteristics. Select the root cause component recorded in the historical fault case with the highest matching degree as the candidate root cause component; Verify the degree of deviation between the design operating parameters of the candidate root cause component and the current operating monitoring data, where the degree of deviation is calculated by the vibration amplitude parameter deviation of the machining unit, the torque fluctuation parameter deviation of the transmission unit, and the flow deviation parameter of the temperature control unit; If the degree of deviation exceeds a preset verification threshold, the candidate root cause component is confirmed to be the root cause component; if not, the historical failure case with the second highest matching degree is selected to repeat the verification operation until a root cause component that meets the verification threshold is found.

2. The fault diagnosis method for intelligent manufacturing system operation and maintenance according to claim 1, characterized in that: The constructing of the device association diagram includes: Identify the core functional modules in the intelligent manufacturing system, including the processing unit directly involved in product processing, the transmission unit responsible for power transmission, and the temperature control unit that maintains the stable temperature of the equipment; Create a device component node for each key component in the core functional module, where the initial attributes of the device component node include component name, installation location, and design operation parameters; Analyze the physical connection structure between the processing unit and the transmission unit, and extract the specific form of the mechanical linkage relationship as the connection basis of the mechanical linkage edge; Analyze the heat exchange path between the processing unit and the temperature control unit, and extract the medium type of the heat conduction relationship as the basis for connecting the heat conduction edge; Analyze the energy consumption distribution mechanism of the transmission unit and the temperature control unit, and extract the energy conversion efficiency of the energy consumption coupling relationship as the connection basis of the energy consumption coupling edge; Creating component association edges for each pair of functionally associated device component nodes based on the connection basis, wherein the initial weights of the component association edges are determined by the tightness of the connection structure, the smoothness of the heat exchange path, and the energy conversion efficiency; The connectivity of the device association graph is verified so that the processing unit, transmission unit and temperature control unit nodes form at least one functional association path through the component association edges, thereby generating a complete device association graph.

3. The fault diagnosis method for intelligent manufacturing system operation and maintenance according to claim 1, characterized in that: The performing node feature extraction processing on the operation monitoring data to generate the operating condition time series features of each equipment component node includes: Performing time window division processing on the tool vibration signal of the machining unit to obtain multiple vibration data windows with continuous machining time period identification; For each vibration data window, the difference between the maximum and minimum amplitude values ​​of the vibration signal is calculated as the vibration amplitude parameter, the number of vibration frequency changes per unit time is counted as the vibration frequency fluctuation parameter, and the duration of the vibration signal amplitude being stable within the design threshold is identified as the vibration stability duration parameter; Combining the vibration amplitude parameter, the vibration frequency fluctuation parameter, and the vibration stability time parameter to generate a vibration stability characteristic of the machining unit; Performing time window division processing on the gear torque signal of the transmission unit to obtain multiple torque data windows with continuous transmission period identifiers; For each torque data window, the difference between the peak and mean values ​​of the torque signal is calculated as the torque fluctuation parameter. The number of times the torque signal exceeds the design rated value is counted as the torque overlimit number parameter. The duration during which the mean value of the torque signal remains stable within the design range is identified as the torque stability duration parameter. Combining the torque fluctuation parameter, the torque exceeding times parameter, and the torque stabilization time parameter to generate a torque smoothness characteristic of the transmission unit; Performing time window division processing on the heat dissipation flow signal of the temperature control unit to obtain multiple flow data windows with continuous temperature control period identification; For each flow data window, the difference between the instantaneous value of the flow signal and the set value is calculated as the flow deviation parameter, the number of times the flow signal fluctuation amplitude exceeds the design threshold is counted as the flow fluctuation number parameter, and the duration of the flow signal being stable within the set range above and below the set value is identified as the flow stability duration parameter; The flow deviation parameter, the flow fluctuation number parameter and the flow stabilization time parameter are combined to generate a flow continuity characteristic of the temperature control unit; The vibration stability characteristics, torque smoothness characteristics and flow continuity characteristics are associated according to the equipment component node identification to generate the working condition time series characteristics of each equipment component node.

4. The fault diagnosis method for intelligent manufacturing system operation and maintenance according to claim 1, characterized in that: The performing edge feature extraction processing on the operation monitoring data of the equipment component nodes at both ends of the component association edge to generate the collaborative change feature of the component association edge includes: Extracting the working condition time series characteristics of the processing unit and the transmission unit at both ends of the mechanical linkage edge, wherein the working condition time series characteristics include vibration stability characteristics and torque smoothness characteristics; Locate the mutation time point of the vibration amplitude parameter of the machining unit, locate the mutation time point of the torque fluctuation parameter of the transmission unit, and calculate the difference between the two mutation time points as the load transfer delay parameter of the mechanical linkage relationship; Extracting the operating time series characteristics of the processing units and the temperature control units at both ends of the heat conduction edge, wherein the operating time series characteristics include vibration stability characteristics and flow continuity characteristics; Locate the falling time point of the vibration stabilization time parameter of the machining unit, locate the falling time point of the flow stabilization time parameter of the temperature control unit, and calculate the difference between the two falling time points as the temperature response hysteresis parameter of the heat conduction relationship; Extracting the operating time series characteristics of the transmission unit and the temperature control unit at both ends of the energy consumption coupling edge, wherein the operating time series characteristics include torque smoothness characteristics and flow continuity characteristics; Calculate the ratio of the transmission unit torque overrun times parameter to the temperature control unit flow fluctuation times parameter as the energy conversion deviation parameter of the energy consumption coupling relationship; The load transfer delay parameter, temperature response lag parameter and energy conversion deviation parameter are normalized, and the normalized parameters are weightedly fused with the initial weights of the component association edges to generate dynamically updated collaborative change features. The collaborative change features establish a binding relationship with the corresponding component association edges in the device association graph through association edge identifiers.

5. The fault diagnosis method for intelligent manufacturing system operation and maintenance according to claim 1, characterized in that: The neighborhood aggregation layer of the graph reasoning model performs information fusion processing on the initial graph representation to obtain neighborhood fusion features, including: Obtain the working condition time series characteristics of the current equipment component node as the central node characteristics, and obtain the working condition time series characteristics of all adjacent nodes of the current node as the neighbor node feature set; Acquire the collaborative change features of the component association edges between the current node and each adjacent node as an edge feature set, perform linear transformation processing on the central node feature to generate a central node transformation feature, and perform linear transformation processing on each neighbor node feature in the neighbor node feature set to generate a neighbor node transformation feature set; Performing linear transformation processing on each edge feature in the edge feature set to generate an edge transformation feature set, and multiplying the neighbor node transformation feature set by corresponding elements of the edge transformation feature set to generate a neighbor node weighted transformation feature set; Aggregating the weighted transformation feature set of the neighboring nodes to generate a neighborhood aggregation feature, adding the center node transformation feature to the neighborhood aggregation feature, and generating an updated node feature of the current device component node through activation function processing; All device component nodes in the device association graph are traversed, and the above-mentioned center node feature processing, neighbor node feature processing and feature aggregation operations are repeatedly performed to generate a graph representation that completes the neighborhood information fusion.

6. The fault diagnosis method for intelligent manufacturing system operation and maintenance according to claim 1, characterized in that: The method of performing dynamic evolution analysis and processing on the graph representation of continuous time periods through the temporal modeling layer of the graph reasoning model to generate time context features reflecting the state evolution of the device association graph includes: Arrange the graphical representations of consecutive time periods in chronological order to generate a graphical representation time series; Inputting the graph representation time series into the input gate of the long short-term memory network, calculating the activation value of the input gate, wherein the activation value of the input gate is used to control the input degree of the graph representation in the current period; Calculating a forget gate activation value of a long short-term memory network, wherein the forget gate activation value is used to control the degree of retention of the historical period graph representation; Calculating candidate memory cell values ​​of the long short-term memory network, where the candidate memory cell values ​​are determined by the current time period graph representation and the historical hidden state; Updating the cell state of the long short-term memory network, where the cell state is obtained by multiplying the forget gate activation value by the historical cell state plus the input gate activation value by the candidate memory unit value; Calculating an output gate activation value of a long short-term memory network, wherein the output gate activation value is used to control the output degree of the current cell state; Generate a hidden state for the current period, where the hidden state is obtained by multiplying the output gate activation value by the hyperbolic tangent function value of the cell state; Traverse all time periods of the time series represented by the graph, repeatedly perform the above-mentioned input gate calculation, forget gate calculation and hidden state update operations, and generate a time context feature containing time dimension information.

7. The fault diagnosis method for intelligent manufacturing system operation and maintenance according to claim 1, characterized in that: Generating an operation and maintenance plan including a maintenance operation sequence and parameter adjustment instructions based on the root cause component, and pushing the operation and maintenance plan to the equipment control system to trigger a fault repair operation, includes: Obtain maintenance history records of the root cause component, and calculate the historical failure frequency and average repair time of the root cause component as a basis for evaluating the maintenance operation sequence; Calculate the expansion rate of the fault impact range based on the propagation speed of the propagation path sequence; Determine the maintenance priority of the root cause component using a preset priority evaluation rule based on the historical failure frequency, mean repair time, and failure impact range expansion rate; Extract the design and maintenance manual of the root cause component to obtain the standard maintenance process, the list of required replacement parts, and the tool preparation requirements for the root cause component as the basis for the parameter adjustment guide; Adaptively adjust the standard maintenance process based on the abnormal parameter characteristics in the current operation monitoring data to obtain an adjusted maintenance process; Combining the maintenance priorities, adjusted maintenance procedures, required replacement parts list, and tool preparation requirements to generate an operation and maintenance plan; The operation and maintenance plan is sent to the equipment control system so that the equipment control system allocates maintenance resources according to maintenance priorities and performs fault isolation and component repair operations according to the adjusted maintenance process.

8. A fault diagnosis system for intelligent manufacturing system operation and maintenance, characterized in that: It includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the fault diagnosis method for intelligent manufacturing system operation and maintenance as described in any one of claims 1 to 7 above.

Images