Intelligent manufacturing optimization method, device, equipment and storage medium based on large model

Abstract

The present invention relates to a large-model-based intelligent manufacturing optimization method, device, equipment and storage medium, comprising the following steps: performing a fault type analysis on the manufacturing equipment based on the fault extraction feature parameters to obtain the fault type; performing digital twin mapping on the manufacturing equipment to obtain the fault location of the manufacturing equipment and the fault component corresponding to the fault location; performing a fault propagation path analysis based on the fault location and the fault component corresponding to the fault location to obtain the associated fault propagation path and the potential risk components on the associated fault propagation path; generating an optimization plan for the manufacturing equipment based on the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location and the fault type, thereby solving the technical problem that, due to the lack of an effective fault detection mechanism, it is often difficult to timely and accurately identify the essential cause of the fault when the equipment parameters exceed the normal range.

Description

Technical Field

[0001] The present invention relates to the field of intelligent manufacturing technology, and in particular to an intelligent manufacturing optimization method, device, equipment and storage medium based on a large model. Background Art

[0002] Smart manufacturing is the core of industry, aiming to realize automation and intelligence of production process by integrating advanced information technology, network technology and intelligent technology. As the manufacturing industry develops towards efficiency, personalization and customization, the traditional manufacturing model can no longer meet the needs of modern industry. In this context, the intelligent manufacturing optimization method based on large models has emerged. This method uses emerging technologies such as big data analysis, artificial intelligence algorithms and the Internet of Things to monitor the operating status of manufacturing equipment in real time and perform predictive maintenance, thereby improving production efficiency, reducing operating costs, and ensuring the consistency and stability of product quality.

[0003] However, in the existing manufacturing environment, although sensor networks and monitoring systems have been widely used for data collection, there are still many challenges. For example, due to the lack of an effective fault detection mechanism, when equipment parameters exceed the normal range, it is often difficult to promptly and accurately identify the root cause of the fault. In addition, even if the type of fault can be determined, it is a difficult task to accurately locate the specific location of the fault and its associated components for complex equipment. This not only affects the speed of troubleshooting, but may also lead to unnecessary downtime and increased maintenance costs. Therefore, a more intelligent and accurate method is needed to deal with these problems, allowing manufacturers to take preventive measures before failures occur.

[0004] In order to solve the above problems, the intelligent manufacturing optimization method based on large models provides a new solution. This method continuously collects and analyzes the operating parameters of manufacturing equipment, uses a preset fault extraction algorithm to quickly and accurately capture abnormal situations, and then uses digital twin mapping technology to map equipment failures in the physical world to a virtual environment, helping engineers to understand and diagnose problems more intuitively. At the same time, it can further analyze the fault propagation path and identify potential risk components, providing a scientific basis for formulating a comprehensive optimization plan. This series of steps helps to improve the reliability of the entire production system, reduce the number of unexpected downtimes, and ultimately promote the transformation of the manufacturing industry to intelligence. Summary of the invention

[0005] The main purpose of the present invention is to provide a large-model-based intelligent manufacturing optimization method, device, equipment and storage medium, which solves the technical problem that due to the lack of an effective fault detection mechanism, it is often difficult to timely and accurately identify the essential cause of the fault when the equipment parameters exceed the normal range.

[0006] To achieve the above object, the present invention provides a large model-based intelligent manufacturing optimization method, which is applied to manufacturing equipment and includes the following steps:

[0007] The operating status of manufacturing equipment is collected through a preset sensor network to obtain operating parameters;

[0008] By using a preset detection mechanism, it is detected whether the operating parameter is within a preset operating threshold. If not, a preset fault extraction algorithm is used to extract fault parameters from the operating parameter to obtain fault extraction feature parameters.

[0009] Performing a fault type analysis on the manufacturing equipment based on the fault extraction characteristic parameters to obtain the fault type;

[0010] Performing digital twin mapping on the manufacturing equipment based on the fault type to obtain a fault location of the manufacturing equipment and a faulty component corresponding to the fault location;

[0011] Perform fault propagation path analysis based on the fault location and the fault component corresponding to the fault location to obtain associated fault propagation paths and potential risk components on the associated fault propagation paths;

[0012] An optimization plan for manufacturing equipment is generated based on the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location, and the fault type.

[0013] Furthermore, the operation status of the manufacturing equipment is collected through a preset sensor network to obtain the operation parameters, including:

[0014] Through the preset sensor network, multi-modal data collection is performed on key parts of the manufacturing equipment to obtain the original operating status data;

[0015] Performing noise reduction and feature enhancement processing on the original operating status data to obtain pre-processed operating status data;

[0016] Compressing the pre-processed running status data to obtain compressed running status data;

[0017] The compressed operating state data is subjected to data fusion to obtain operating parameters; wherein the operating parameters include high-frequency vibration signals, temperature field distribution data, three-phase current waveforms, speed fluctuation signals and strain data of key parts.

[0018] Furthermore, the fault extraction algorithm includes a multi-branch residual feature extraction network, a hole convolution network, a depth-separable convolution, a deconvolution network, a cross-scale feature aggregation module and a dense connection module. The fault extraction algorithm is used to extract fault parameters from the operating parameters to obtain fault extraction feature parameters, including:

[0019] Performing multi-scale feature extraction on the operating parameters through the multi-branch residual feature extraction network to obtain a multi-layer feature map, and performing attention mechanism fusion on the multi-layer feature map to obtain fused feature data;

[0020] Based on the atrous convolutional network, the receptive field of the fused feature tensor is expanded to obtain an extended feature map, and the extended feature map is subjected to pyramid pooling analysis to obtain a multi-scale feature pyramid; wherein the multi-scale feature pyramid includes local detail features, meso-structure features and global semantic features;

[0021] Performing lightweight processing on the multi-scale feature pyramid through the depthwise separable convolution to obtain a lightweight feature map, and performing channel attention recalibration on the lightweight feature map to obtain a recalibrated feature map;

[0022] Based on the deconvolution network, the recalibrated feature map is upsampled and reconstructed to obtain a reconstructed feature map, and the reconstructed feature map is subjected to non-local correlation analysis to obtain feature association data; wherein the feature association data is a long-range dependency relationship between fault features;

[0023] Performing multi-level feature representation on the feature association data through the cross-scale feature aggregation module to obtain a hierarchical feature representation, and performing adaptive feature selection on the hierarchical feature representation to obtain an optimal feature set; wherein the optimal feature set includes significant features, discriminative features and complementary features;

[0024] Based on the dense connection module, feature reorganization and normalization processing are performed on the preferred feature set to obtain fault extraction feature parameters.

[0025] Further, the performing of fault type analysis on the manufacturing equipment based on the fault extraction characteristic parameters to obtain the fault type includes:

[0026] Performing feature decomposition on the fault extraction feature parameters by using a spectral clustering algorithm to obtain a frequency spectrum feature matrix, and performing wavelet packet analysis on the frequency spectrum feature matrix to obtain a fault vibration feature group; wherein the fault vibration feature group includes a periodic impact feature, a non-stationary vibration feature, and a harmonic resonance feature;

[0027] Vibration pattern recognition is performed based on the fault vibration feature group through a deep belief network to obtain a vibration pattern set, and envelope spectrum analysis is performed on the vibration pattern set to obtain mechanical fault features; wherein the mechanical fault features include bearing inner race fault, bearing outer race fault and bearing rolling element fault;

[0028] The mechanical fault features are modally separated by variational mode decomposition technology to obtain a fault mode sequence, and the fault mode sequence is empirically decomposed to obtain gear fault features; wherein the gear fault features include gear tooth breakage, gear pitting and gear cracks;

[0029] By using a preset sparse autoencoder, electrical features of the manufacturing equipment are extracted based on the gear fault features to obtain an electrical feature set, and the electrical feature set is Hilbert transformed to obtain electrical fault features; wherein the electrical fault features include stator winding short circuit, rotor bar breakage and motor bearing electrical corrosion;

[0030] Performing energy distribution analysis on the electrical fault characteristics through the Teager energy operator to obtain an energy distribution spectrum, and performing multi-scale entropy analysis on the energy distribution spectrum to obtain structural fault characteristics; wherein the structural fault characteristics include base cracking, coupling breakage and base loosening;

[0031] Through the fuzzy reasoning system, a comprehensive fault judgment is performed based on the structural fault characteristics to obtain a fault feature combination, and a fault type analysis is performed on the fault feature combination to obtain a fault type; wherein the fault types include mechanical wear faults, mechanical fracture faults, electrical short circuit faults, electrical open circuit faults, structural damage faults and structural deformation faults.

[0032] Furthermore, the digital twin mapping of the manufacturing equipment based on the fault type to obtain the fault location of the manufacturing equipment and the fault component corresponding to the fault location includes:

[0033] Performing fault semantic decoding on the fault type by using a preset global feature embedding technology to obtain a fault semantic feature vector, and marking the fault trigger area of ​​the digital twin model of the manufacturing equipment based on the fault semantic feature vector to obtain an initial fault trigger area set;

[0034] Based on the reverse engineering modeling algorithm, geometric feature inversion is performed on the initial fault trigger area set to obtain a geometric space feature map of equipment faults, and local mesh refinement is performed on the fault geometric space feature map through a topology optimization algorithm to obtain a digital twin geometric mapping mesh;

[0035] Performing multi-physical field coupling simulation on the digital twin geometric mapping grid through dynamic physical simulation technology to obtain a multi-dimensional fault physical field distribution map, and performing fault field location on the multi-dimensional fault physical field distribution map to obtain a fault location parameter set;

[0036] By using a knowledge graph reasoning algorithm, the cascade relationship of the equipment components in the digital twin model is inferred based on the fault location parameter set to obtain a fault propagation association network, and node importance analysis is performed on the fault propagation association network to obtain a set of key fault nodes;

[0037] Based on the set of critical fault nodes, the digital twin model is three-dimensionally visualized to obtain a mapping result, and based on the mapping result, a fault location of the manufacturing equipment and a faulty component corresponding to the fault location are obtained.

[0038] Further, the performing of fault propagation path analysis based on the fault location and the fault component corresponding to the fault location to obtain the associated fault propagation path and the potential risk components on the associated fault propagation path includes:

[0039] Performing causal relationship modeling on the fault location and the fault component corresponding to the fault location through a Bayesian network to obtain a preliminary fault propagation path, and performing probability reasoning on the preliminary fault propagation path to obtain a fault propagation probability matrix, wherein the fault propagation probability matrix includes the fault propagation probability between each fault component;

[0040] By using the Markov chain Monte Carlo method, a fault path simulation is performed based on the fault propagation probability matrix to obtain a set of fault propagation paths, and path frequency statistics are performed on the set of fault propagation paths to obtain a high-frequency fault propagation path, wherein the high-frequency fault propagation path includes multiple fault propagation paths and the occurrence frequencies corresponding to the multiple fault propagation paths;

[0041] Performing risk priority sorting on the high-frequency fault propagation paths to obtain a risk priority list, and performing weighted analysis on the risk priority list to obtain a priority-weighted fault propagation path;

[0042] The priority weighted fault propagation path is subjected to a timing relationship analysis through a dynamic Bayesian network to obtain an associated fault propagation path, and the components on the associated fault propagation path are subjected to a timing relationship analysis to obtain potential risk components on the associated fault propagation path.

[0043] Further, the generating of an optimization plan for manufacturing equipment based on the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location, and the fault type includes:

[0044] Extracting parameters of the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location, and the fault type through a multi-objective optimization algorithm to obtain an extracted parameter set;

[0045] Dynamically simulating the digital twin model of the manufacturing equipment based on the extracted parameter set to obtain an operation status prediction graph of the digital twin model, and performing time series analysis on the operation status prediction graph of the digital twin model to obtain a fault prediction time series;

[0046] By using a fault tree analysis method, the fault cause analysis is performed on the fault prediction time series to obtain a set of fault causes, and the set of fault causes is hierarchically analyzed to obtain a list of fault cause priorities;

[0047] The preset manufacturing equipment operation data is called from the database, and the preset manufacturing equipment operation data is optimized based on the fault cause priority list to generate an optimization plan for the manufacturing equipment.

[0048] The present invention also provides an intelligent manufacturing optimization device based on a large model, which is applied to manufacturing equipment, comprising:

[0049] A collection module is used to collect the operating status of the manufacturing equipment through a preset sensor network to obtain operating parameters;

[0050] An extraction module, used to detect whether the operating parameters are within a preset operating threshold through a preset detection mechanism, and if not, to extract fault parameters from the operating parameters through a preset fault extraction algorithm to obtain fault extraction feature parameters;

[0051] A first analysis module, configured to perform a fault type analysis on the manufacturing equipment based on the fault extraction characteristic parameters to obtain the fault type;

[0052] A mapping module, used to perform digital twin mapping on the manufacturing equipment based on the fault type to obtain a fault location of the manufacturing equipment and a faulty component corresponding to the fault location;

[0053] A second analysis module is used to perform fault propagation path analysis based on the fault location and the fault component corresponding to the fault location, to obtain an associated fault propagation path and a potential risk component on the associated fault propagation path;

[0054] A generation module is used to generate an optimization plan for manufacturing equipment based on the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location and the fault type.

[0055] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0056] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0057] The large-model-based intelligent manufacturing optimization method provided by the present invention comprises the following steps: collecting the operating status of manufacturing equipment through a preset sensor network to obtain operating parameters; detecting whether the operating parameters are within a preset operating threshold through a preset detection mechanism, and if not, extracting fault parameters from the operating parameters through a preset fault extraction algorithm to obtain fault extraction feature parameters; analyzing the fault type of the manufacturing equipment based on the fault extraction feature parameters to obtain the fault type; performing digital twin mapping on the manufacturing equipment based on the fault type to obtain the fault location of the manufacturing equipment and the fault component corresponding to the fault location; analyzing the fault propagation path based on the fault location and the fault component corresponding to the fault location to obtain the associated fault propagation path and the potential risk component on the associated fault propagation path; generating an optimization plan for the manufacturing equipment based on the associated fault propagation path, the potential risk component on the associated fault propagation path, the fault location, the fault component corresponding to the fault location, and the fault type. Through the above-mentioned technical means, the technical problem that it is often difficult to timely and accurately identify the essential cause of the fault when the equipment parameters exceed the normal range due to the lack of an effective fault detection mechanism is solved, and it is realized that through in-depth analysis of the fault propagation path, the method can identify the potential risk component on the associated fault path. This feature allows manufacturers to take preventive measures in advance to avoid the expansion of failures or chain reactions. It also supports the formulation of more targeted maintenance plans, reduces the risk of unexpected downtime, and ensures the continuity and stability of production. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 It is a schematic diagram of the steps of a large model-based intelligent manufacturing optimization method in one embodiment of the present invention;

[0059] Figure 2 It is a structural block diagram of an intelligent manufacturing optimization device based on a large model in one embodiment of the present invention;

[0060] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0061] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0063] like Figure 1 As shown, Figure 1 It is a schematic diagram of the steps of a large-model-based intelligent manufacturing optimization method in one embodiment of the present invention;

[0064] In one embodiment of the present invention, a large model-based intelligent manufacturing optimization method is provided, which is applied to manufacturing equipment and includes the following steps:

[0065] Step S1, collecting the operating status of the manufacturing equipment through a preset sensor network to obtain operating parameters.

[0066] Specifically, in the process of collecting the operating status of manufacturing equipment, obtaining operating parameters through a preset sensor network is a crucial first step. The "preset sensor network" here refers to a system composed of multiple different types of sensors, which are deployed in key parts of the entire manufacturing equipment. Each sensor has a specific function and measurement range. They can monitor a variety of physical quantities such as temperature, pressure, vibration, displacement, and convert these data into electronic signals. In order to ensure the accuracy and reliability of the data, the selection and layout of these sensors are carefully designed according to the specific manufacturing environment and equipment characteristics. When it comes to "collecting the operating status of manufacturing equipment", it means using the above-mentioned sensor network to continuously collect various performance information of manufacturing equipment in actual work. For example, in an application on an automobile engine production line, vibration sensors may be installed to monitor the movement of the robot arm on the assembly line, and thermocouples may be equipped to detect temperature changes at welding points. These sensors will automatically record the corresponding data points according to the set time interval or trigger conditions, thereby forming a series of digital descriptions of the operating status of the manufacturing equipment - the so-called "operating parameters". The process of "obtaining operating parameters" refers to the process of collecting raw data from the sensor, and then extracting effective information that can reflect the health status of the equipment after preliminary processing and analysis. For example, on an automobile engine production line, if the vibration amplitude of a CNC machine tool exceeds the normal range, this abnormal value will be marked separately as a key operating parameter. Subsequently, the parameter will be passed to the subsequent fault detection mechanism for further evaluation. The entire collection process ensures that even the slightest changes will not be ignored, providing a solid data foundation for subsequent fault diagnosis. Therefore, through such a tightly integrated and highly sensitive sensor network, manufacturing companies can grasp the working status of equipment in real time, detect potential problems in a timely manner and take corresponding measures to ensure production continuity and product quality.

[0067] Step S2, using a preset detection mechanism to detect whether the operating parameter is within a preset operating threshold, if not, extracting fault parameters from the operating parameter using a preset fault extraction algorithm to obtain fault extraction feature parameters.

[0068] Specifically, in the process of monitoring the operating parameters of manufacturing equipment, it is crucial to evaluate whether these parameters are within the preset operating thresholds through a preset detection mechanism. This process is immediately followed by the acquisition of the operating status, where the "preset detection mechanism" refers to a set of pre-defined rules or algorithms for comparing the real-time data obtained from the sensor network with the preset safe operating range. For example, in an automobile engine production line, each CNC machine tool has its own specific operating parameter range, such as temperature, vibration frequency, etc., which are set based on manufacturer recommendations and long-term experience accumulation. Once the operating parameters are collected and transmitted to the detection system, the mechanism will be immediately activated to compare the actual measured values ​​with the preset operating thresholds. If all parameters are within the specified range, the equipment is considered to be operating normally and monitoring continues without taking additional action. However, if some parameters are found to be beyond the safety limit, this indicates that there may be potential problems or impending failures. At this time, the "preset detection mechanism" will not only trigger an alarm to notify maintenance personnel, but also automatically activate a deeper data processing process-that is, fault parameter extraction of the operating parameters through a preset fault extraction algorithm. The so-called "fault extraction algorithm" is a series of complex mathematical operations and technical means designed to identify diagnostic features from abnormal operating parameters. These algorithms can filter out noise and other irrelevant information and focus on key data points that truly indicate where the problem lies. Taking CNC machine tools as an example, when the vibration frequency is detected to be out of standard, the fault extraction algorithm may analyze the time series characteristics of the vibration mode, calculate the spectrum distribution, and combine other relevant parameters (such as temperature changes) to determine whether there is imbalance, wear or other mechanical failures. After this series of processing, the final result is the so-called "fault extraction feature parameters", which provide a solid foundation for subsequent fault type analysis, allowing engineers to more accurately locate the source of the problem and develop effective repair strategies. In this way, the entire detection and fault extraction process forms a closed-loop system, ensuring that any abnormalities in the production process can be quickly captured and responded to, thereby improving the reliability of the equipment and the continuity of production.

[0069] Step S3, performing fault type analysis on the manufacturing equipment based on the fault extraction characteristic parameters to obtain the fault type.

[0070] Specifically, in the process of analyzing the fault type of manufacturing equipment based on the fault extraction characteristic parameters, this link is a crucial step in the entire intelligent manufacturing optimization method. It is directly related to whether the specific nature of the equipment fault can be accurately identified, thereby providing clear guidance for subsequent maintenance and repair work. After the fault extraction characteristic parameters are obtained through the preset fault extraction algorithm, these parameters need to be used to perform in-depth fault type analysis. At this stage, "based on the fault extraction characteristic parameters" means taking the key data points obtained in the previous step as input, and these characteristic parameters have been preliminarily screened and processed to represent the abnormal conditions in the operation state of the equipment. For example, in a CNC machine tool on an automobile engine production line, if the vibration frequency and temperature changes are confirmed as abnormal and further refined into specific patterns or spectral distributions through a fault extraction algorithm, then this information constitutes the fault extraction characteristic parameters. Then, the process of "analyzing the fault type of the manufacturing equipment" involves parsing these characteristic parameters using a pre-trained model or expert system to determine the specific fault type they indicate. This may include but is not limited to problems such as wear of mechanical parts, overheating of motors, and insufficient lubrication. In practical applications, this analysis relies on a large amount of historical data accumulation and the support of machine learning technology. For example, for CNC machine tools, once an abnormal vibration pattern is detected, the system will match it with a known fault case library to find out whether there is a similar historical record. If a similar situation is found, it can be quickly inferred that the current fault may be caused by spindle imbalance or tool damage based on the previous diagnosis results. In addition, the ultimate goal of "getting the fault type" is to clarify the nature of the fault so that targeted solutions can be taken. In the example of the above automobile engine production line, if the analysis results show that the fault type is "spindle imbalance", the maintenance team can check the installation accuracy of the spindle, the position of the balance block and other factors in a targeted manner without blindly disassembling the whole machine. Such precise positioning not only saves time costs, but also improves the repair efficiency, ensuring that production can resume normal operation as soon as possible. Therefore, through in-depth analysis of fault extraction feature parameters, manufacturing companies can more effectively manage equipment health, prevent potential problems, and ensure production continuity and product quality.

[0071] Step S4: Perform digital twin mapping on the manufacturing equipment based on the fault type to obtain the fault location of the manufacturing equipment and the fault component corresponding to the fault location.

[0072] Specifically, in the process of digital twin mapping of manufacturing equipment based on the fault type, this step aims to map the fault phenomenon in the physical world into the virtual environment so as to more intuitively and accurately determine the specific location of the fault and its corresponding faulty component. After the nature of the fault is clarified through fault type analysis, the next step is to use digital twin technology to implement this step. Digital twin mapping is not just a simple data conversion process. It involves the construction of a three-dimensional model of the manufacturing equipment, real-time data synchronization, and complex algorithm operations to ensure that the virtual model can truly reflect the state of the physical equipment. In this process, "based on the fault type" means using the identified fault features as input conditions, and these feature parameters guide the focus of the digital twin system. For example, in a CNC machine tool on an automobile engine production line, if the fault type is confirmed to be "spindle imbalance", the digital twin system will pay special attention to all components and connections related to the spindle. Then, the operation of "digital twin mapping of the manufacturing equipment" will be started, which relies on a pre-established high-precision virtual model that contains all the structural details and working principles of the manufacturing equipment and maintains continuous data exchange with the actual equipment. As the mapping process progresses, "obtaining the fault location of the manufacturing equipment and the faulty component corresponding to the fault location" has become the ultimate goal. By matching the fault extraction feature parameters with the corresponding parts in the digital twin model, the system can accurately locate the area where the fault occurs and identify the specific faulty component. For example, in the case of CNC machine tools, the digital twin system may show that there is an abnormal vibration mode at the spindle bearing, and at the same time point out that it may be caused by bearing wear or improper installation. Such visual feedback not only helps engineers quickly understand the problem, but also assists in formulating a more scientific and reasonable maintenance plan. In addition, digital twin mapping also supports dynamic simulation functions, allowing technicians to reproduce fault scenarios in a virtual environment, evaluate the effects of different repair solutions, and even predict possible problems in the future. For automobile engine production lines, this means that various maintenance activities can be rehearsed in advance through the digital twin platform without interrupting actual production, optimizing resource allocation and reducing downtime. Therefore, through this advanced digital twin mapping technology, manufacturing companies can significantly improve the speed and accuracy of fault diagnosis, thereby enhancing overall operational efficiency and service quality.

[0073] Step S5, performing fault propagation path analysis based on the fault location and the fault component corresponding to the fault location to obtain associated fault propagation paths and potential risk components on the associated fault propagation paths.

[0074] Specifically, in the process of fault propagation path analysis based on the fault location and the faulty components corresponding to the fault location, this link aims to explore the scope of influence of the fault not only limited to its direct occurrence point, but also to further track possible chain reactions and their potential impacts. When the digital twin mapping helps us accurately locate the specific location of the fault and the related components, we need to conduct in-depth analysis of this information to understand how the fault propagates within the manufacturing equipment and identify the potential risk components on the associated fault propagation path. Specifically, "based on the fault location and the faulty components corresponding to the fault location" means starting a more complex simulation and prediction process using the already identified fault location as a starting point. For example, in a CNC machine tool on an automobile engine production line, if the spindle bearing is confirmed as a faulty component, then it is necessary to consider how this fault may affect other components connected or interacting with it. Through this analysis, it can be revealed that the possibility of the fault starting from the spindle bearing and gradually expanding along the mechanical connection, hydraulic system or other conduction path can be revealed. The implementation of "fault propagation path analysis" relies on pre-established fault propagation models and algorithms, which can simulate the development trajectory of faults under different conditions. In the example of the CNC machine tool mentioned above, once an abnormality is found in the spindle bearing, the system automatically calls the corresponding fault propagation model to evaluate whether the fault may cause tool damage, reduced machining accuracy, or a wider range of mechanical structure imbalances. At the same time, operating environment factors such as temperature changes and load fluctuations are also taken into account, which may aggravate the impact of the fault. As the analysis deepens, "obtaining the associated fault propagation path and the potential risk components on the associated fault propagation path" has become a key achievement. This means not only finding the directly affected components, but also identifying other components that are at risk due to chain effects. For example, in a CNC machine tool, if the vibration caused by the spindle imbalance exceeds a certain limit, it may cause damage to the adjacent support structure and even affect the stability of the entire machine tool frame. Therefore, through fault propagation path analysis, engineers can foresee these problems in advance and take preventive maintenance measures, such as increasing the inspection frequency of the support structure and adjusting the working parameters to reduce vibration transmission. In addition, this forward-looking analysis helps to develop a more comprehensive optimization plan to ensure that while dealing with the initial fault, it can also prevent subsequent risks. For automobile engine production lines, this is equivalent to building a line of defense, so that even if unexpected problems occur during the production process, they can be responded to and effectively controlled quickly to avoid greater losses.

[0075] Step S6, generating an optimization plan for manufacturing equipment based on the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location, and the fault type.

[0076] Specifically, in the process of generating an optimization plan for manufacturing equipment based on the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location, and the fault type, this link is a key step in the intelligent manufacturing optimization method. It integrates the results of all previous analyses and aims to provide a comprehensive and specific solution for manufacturing enterprises. After the preliminary work of fault detection, fault extraction feature parameter analysis, digital twin mapping, and fault propagation path analysis is completed, the next task is to use this information to develop a set of optimization plans that can effectively solve current problems and prevent future hidden dangers. Specifically, "based on the associated fault propagation path" means taking into account that the fault is not limited to the impact of its initial occurrence point, but also includes the risk of gradually expanding along the mechanical connection, hydraulic system or other conduction path. For example, in a CNC machine tool on an automobile engine production line, if the failure of the spindle bearing may cause tool damage or reduced machining accuracy, then the optimization plan needs to consider how to repair the spindle bearing and its impact on related components at the same time. "Potential risk components on the associated fault propagation path" further clarifies which other components are at risk due to the chain effect, thereby ensuring that the optimization measures cover all affected areas. At the same time, "the fault location and the faulty component corresponding to the fault location" provide specific information about the direct fault source, which is the core focus of the entire optimization plan. Taking the CNC machine tool as an example, once the spindle bearing is determined to be the faulty component, the optimization plan will first propose repair or replacement suggestions for this component and plan the implementation steps in detail. In addition, "the fault type" provides directional guidance for the formulation of the optimization plan, and different fault types require different types of repair and improvement measures. For example, if the fault is caused by wear, the optimization plan may include more frequent maintenance inspections, the use of higher-quality lubricants, or improved cooling system design. By combining all of the above factors, the process of "generating an optimization plan for manufacturing equipment" is not limited to solving the problem itself, but focuses on improving the reliability and efficiency of the overall system. For an automobile engine production line, this means not only fixing the current problems found, such as spindle bearing failures, but also evaluating and improving the weak links in the entire production process. The optimization plan may suggest introducing new monitoring technologies to monitor key parameters in real time, adjusting production plans to reduce high-load operations, and even redesigning certain process steps to improve durability and stability. Ultimately, such an optimization plan not only helps to quickly restore normal production order, but also brings long-term cost savings and technological progress to the company. In summary, by comprehensively considering the associated fault propagation paths, potential risk components, fault locations, fault components and fault types, manufacturing companies can develop more scientific and reasonable optimization plans to ensure that while dealing with existing problems, they can also prevent potential risks in the future, thereby significantly improving the reliability and competitiveness of the production system.

[0077] In a specific embodiment, the operation status of the manufacturing equipment is collected through a preset sensor network to obtain the operation parameters, including:

[0078] Through the preset sensor network, multi-modal data collection is performed on key parts of the manufacturing equipment to obtain the original operating status data;

[0079] Performing noise reduction and feature enhancement processing on the original operating status data to obtain pre-processed operating status data;

[0080] Compressing the pre-processed running status data to obtain compressed running status data;

[0081] The compressed operating state data is subjected to data fusion to obtain operating parameters; wherein the operating parameters include high-frequency vibration signals, temperature field distribution data, three-phase current waveforms, speed fluctuation signals and strain data of key parts.

[0082] Specifically, first of all, "multimodal data collection of key parts in manufacturing equipment through a preset sensor network" means using a carefully designed sensor system to capture different types of physical quantity changes inside the equipment. These sensors are deployed in locations that are critical to equipment performance, such as the spindle, tool interface, hydraulic system, etc. of CNC machine tools. For example, in CNC machine tools on automobile engine production lines, vibration sensors can monitor the operation of the spindle, temperature sensors are used to measure the temperature distribution in the processing area, current transformers are responsible for recording the three-phase current waveform of the motor, and strain gauges are installed at the structural support points to sense stress changes. In this way, the system can obtain original operating status data including high-frequency vibration signals, temperature field distribution data, three-phase current waveforms, speed fluctuation signals, and strain data of key parts. However, the "original operating status data" obtained directly from the sensor often contains a lot of noise and other irrelevant information, which will interfere with subsequent fault detection and diagnosis. Therefore, "noise reduction and feature enhancement processing of the original operating status data" is an essential step. At this stage, engineers apply advanced signal processing techniques, such as filtering algorithms and wavelet transforms, to remove random noise from the data and highlight important features that truly represent the health of the equipment. For example, in the case of CNC machine tools, if the original vibration signal is mixed with environmental noise or other non-periodic disturbances, the noise reduction algorithm will effectively filter out these unnecessary components while retaining and enhancing specific spectral components related to the spindle rotation frequency. Such processing not only improves the quality of the data, but also lays a good foundation for subsequent operations. Next, "data compression of the pre-processed operating status data" is to cope with the storage and transmission challenges brought by massive data. In modern industrial environments, sensor networks may generate thousands of data points per second. If not compressed, it will take up a lot of storage space and increase the pressure on communication bandwidth. To this end, the use of efficient compression algorithms, such as model-based predictive coding or vector quantization technology, can significantly reduce the amount of data without losing important information. Continuing with the example of CNC machine tools, the high-frequency vibration signal after noise reduction and feature enhancement is still huge, but through appropriate compression methods, it can be converted into a more compact form, which saves resources and does not affect the accuracy of subsequent analysis. Finally, “the compressed operating status data is fused to obtain operating parameters”. This is the most critical link in the entire data processing chain, because it determines whether the final output information has practical value. Data fusion is not just a simple splicing of different types of data together, but a comprehensive consideration of the relationship between various factors to build a multi-dimensional view that fully describes the operating status of the equipment.For example, in the application scenario of CNC machine tools, compressed high-frequency vibration signals, temperature field distribution data, three-phase current waveforms, speed fluctuation signals, and strain data of key parts need to be integrated to form a set of operating parameters that can fully reflect the overall health of the equipment. To achieve this, machine learning algorithms or expert systems may be used to analyze the correlation patterns between the data streams to generate more accurate and useful operating parameters. In summary, through such a process consisting of multimodal data acquisition, noise reduction and feature enhancement processing, data compression, and data fusion, manufacturing companies can obtain high-quality and easy-to-manage operating parameters. These parameters are not only the basis for fault detection and diagnosis, but also provide a reliable basis for subsequent fault type analysis, digital twin mapping, fault propagation path analysis, and the generation of the final optimization plan. In the application of CNC machine tools in automobile engine production lines, such a meticulous data processing method enables production managers to grasp the working status of the equipment in real time, discover potential problems in time, and take preventive measures, thereby improving production efficiency and product quality, reducing maintenance costs, and ensuring the continuity and stability of production.

[0083] In a specific embodiment, the fault extraction algorithm includes a multi-branch residual feature extraction network, a hole convolution network, a depth-separable convolution, a deconvolution network, a cross-scale feature aggregation module and a dense connection module. The fault extraction algorithm is used to extract fault parameters from the operating parameters to obtain fault extraction feature parameters, including:

[0084] Performing multi-scale feature extraction on the operating parameters through the multi-branch residual feature extraction network to obtain a multi-layer feature map, and performing attention mechanism fusion on the multi-layer feature map to obtain fused feature data;

[0085] Based on the atrous convolutional network, the receptive field of the fused feature tensor is expanded to obtain an extended feature map, and the extended feature map is subjected to pyramid pooling analysis to obtain a multi-scale feature pyramid; wherein the multi-scale feature pyramid includes local detail features, meso-structure features and global semantic features;

[0086] Performing lightweight processing on the multi-scale feature pyramid through the depthwise separable convolution to obtain a lightweight feature map, and performing channel attention recalibration on the lightweight feature map to obtain a recalibrated feature map;

[0087] Based on the deconvolution network, the recalibrated feature map is upsampled and reconstructed to obtain a reconstructed feature map, and the reconstructed feature map is subjected to non-local correlation analysis to obtain feature association data; wherein the feature association data is a long-range dependency relationship between fault features;

[0088] Performing multi-level feature representation on the feature association data through the cross-scale feature aggregation module to obtain a hierarchical feature representation, and performing adaptive feature selection on the hierarchical feature representation to obtain an optimal feature set; wherein the optimal feature set includes significant features, discriminative features and complementary features;

[0089] Based on the dense connection module, feature reorganization and normalization processing are performed on the preferred feature set to obtain fault extraction feature parameters.

[0090] Specifically, in order to realize the process of extracting fault parameters from the operating parameters through a preset fault extraction algorithm and obtaining fault extraction feature parameters, this link integrates multiple advanced technologies such as multi-branch residual feature extraction network, void convolution network, deep separable convolution, deconvolution network, cross-scale feature aggregation module and dense connection module. These technologies work together to ensure that key features that can indicate the existence of faults are accurately extracted from complex operating parameters. First of all, "multi-scale feature extraction of the operating parameters through the multi-branch residual feature extraction network" means using a multi-branch neural network to capture information at different scales. For example, in the application of CNC machine tools, operating parameters such as high-frequency vibration signals and temperature field distribution data contain rich details. The multi-branch residual feature extraction network can not only process multiple input modes, but also maintain the integrity of information transmission through residual connections to prevent the gradient disappearance problem in deep networks. After "obtaining a multi-layer feature map", "the multi-layer feature map is fused with an attention mechanism", that is, the attention mechanism is introduced to automatically focus on those key areas that best reflect the fault state. For example, if the spindle bearing is abnormal, the associated vibration mode changes will be given a higher weight, thus "obtaining fused feature data". This fusion not only enhances the model's characterization ability, but also improves the accuracy of subsequent analysis. Next, "the fused feature tensor is expanded based on the dilated convolutional network", that is, the network's receptive field is expanded through dilated convolution (also known as expanded convolution) without increasing the number of parameters or computational complexity. Dilated convolution can obtain a wider range of information while maintaining resolution, which is particularly important for capturing long-distance dependencies on fault propagation paths. After "obtaining the extended feature map", "pyramid pooling analysis is performed on the extended feature map", which constructs a "multi-scale feature pyramid" containing information at multiple scales. In this pyramid, "local detail features" help identify small but critical changes; "mesoscopic structural features" provide information about the interaction between components within the equipment; and "global semantic features" reflect the status of the entire system level. Taking CNC machine tools as an example, such a multi-scale analysis can simultaneously reveal the impact of spindle bearing failures on tool processing accuracy and support structure stability. Subsequently, "the multi-scale feature pyramid is lightweighted by the depthwise separable convolution", which is to reduce the computational burden of the model and improve efficiency. The depthwise separable convolution decomposes the standard convolution into two steps: spatial convolution and pointwise convolution, which reduces the number of parameters and retains the feature expression capability. After "obtaining the lightweight feature map", "recalibrating the lightweight feature map with channel attention", that is, using the channel attention mechanism to adjust the importance of each feature channel, thereby further enhancing the performance of significant features. "Obtaining the recalibrated feature map" prepares more refined data for subsequent steps.Furthermore, "up-sampling and reconstructing the recalibrated feature map based on the deconvolution network" is a process that aims to restore high-resolution feature representations in order to better understand the spatial layout of the original input. After "obtaining the reconstructed feature map", "non-local correlation analysis is performed on the reconstructed feature map", which explores the long-range dependencies between features, namely the so-called "feature association data". For example, in a CNC machine tool, a spindle bearing failure may not only affect its nearby components, but also affect the remote processing tools through mechanical transmission. This non-local analysis helps us fully understand how the fault propagates throughout the entire device. Then, "multi-level feature representation is performed on the feature association data through the cross-scale feature aggregation module", that is, feature information at different scales is integrated to form a hierarchical feature description. After "obtaining the hierarchical feature representation", "adaptive feature selection is performed on the hierarchical feature representation", which means that only the most representative features, such as "significant features", "discriminative features" and "complementary features", are retained to form a "preferred feature set". These features are carefully selected to reflect the essential characteristics of the fault to the greatest extent, thereby providing a solid foundation for the final fault diagnosis. Finally, "based on the dense connection module, the preferred feature set is reorganized and normalized", and the dense connection module ensures the efficient transmission of information between layers, and through normalization, the feature value falls within a reasonable range, thereby "obtaining the fault extraction feature parameters". This method not only improves the richness and diversity of feature representation, but also enhances the generalization ability of the model, so that fault features can be accurately extracted even under different working conditions. In summary, through such a complex process consisting of a multi-branch residual feature extraction network, a void convolution network, a deep separable convolution, a deconvolution network, a cross-scale feature aggregation module and a dense connection module, manufacturing companies can accurately extract fault features from operating parameters. In the application of CNC machine tools in automobile engine production lines, such a meticulous fault extraction method enables engineers to discover potential problems in a timely manner, take preventive measures, and thus improve production efficiency and product quality, reduce maintenance costs, and ensure the continuity and stability of production. This method not only improves the speed and accuracy of fault diagnosis, but also provides a solid foundation for subsequent fault type analysis, digital twin mapping, fault propagation path analysis, and the generation of the final optimization plan.

[0091] In a specific embodiment, the performing of fault type analysis on the manufacturing equipment based on the fault extraction characteristic parameters to obtain the fault type includes:

[0092] Performing feature decomposition on the fault extraction feature parameters by using a spectral clustering algorithm to obtain a frequency spectrum feature matrix, and performing wavelet packet analysis on the frequency spectrum feature matrix to obtain a fault vibration feature group; wherein the fault vibration feature group includes a periodic impact feature, a non-stationary vibration feature, and a harmonic resonance feature;

[0093] Vibration pattern recognition is performed based on the fault vibration feature group through a deep belief network to obtain a vibration pattern set, and envelope spectrum analysis is performed on the vibration pattern set to obtain mechanical fault features; wherein the mechanical fault features include bearing inner race fault, bearing outer race fault and bearing rolling element fault;

[0094] The mechanical fault features are modally separated by variational mode decomposition technology to obtain a fault mode sequence, and the fault mode sequence is empirically decomposed to obtain gear fault features; wherein the gear fault features include gear tooth breakage, gear pitting and gear cracks;

[0095] By using a preset sparse autoencoder, electrical features of the manufacturing equipment are extracted based on the gear fault features to obtain an electrical feature set, and the electrical feature set is Hilbert transformed to obtain electrical fault features; wherein the electrical fault features include stator winding short circuit, rotor bar breakage and motor bearing electrical corrosion;

[0096] Performing energy distribution analysis on the electrical fault characteristics through the Teager energy operator to obtain an energy distribution spectrum, and performing multi-scale entropy analysis on the energy distribution spectrum to obtain structural fault characteristics; wherein the structural fault characteristics include base cracking, coupling breakage and base loosening;

[0097] Through the fuzzy reasoning system, a comprehensive fault judgment is performed based on the structural fault characteristics to obtain a fault feature combination, and a fault type analysis is performed on the fault feature combination to obtain a fault type; wherein the fault types include mechanical wear faults, mechanical fracture faults, electrical short circuit faults, electrical open circuit faults, structural damage faults and structural deformation faults.

[0098] Specifically, in order to realize the fault type analysis of manufacturing equipment, the described method combines advanced technologies such as spectrum analysis, deep learning, modal decomposition and intelligent systems. First, "the fault extraction feature parameters are characterized by a spectral clustering algorithm", which means converting the fault extraction feature parameters into a spectral feature matrix that can reveal the internal structure of the data. In this process, the spectral clustering algorithm uses graph theory methods in mathematics to identify and group similar data points. For example, in the application of CNC machine tools, when the spindle bearing wears, its vibration signal will contain changes in specific frequency components. These changes are captured and converted into a spectral feature matrix, in which each element represents the energy intensity at a specific frequency. Next, "wavelet packet analysis is performed on the spectral feature matrix", a step aimed at further refining the spectral information to obtain a "fault vibration feature group". Wavelet packet analysis is a multi-resolution analysis technique that can provide information in both the time domain and the frequency domain, and is particularly suitable for processing non-stationary signals. In this example, "periodic impact features" may indicate mechanical impact events that occur regularly; "non-stationary vibration features" reflect random vibration patterns caused by wear or looseness; and "harmonic resonance features" are related to the natural frequencies of rotating parts, such as the speed of the spindle and the meshing frequency of the gears. Through detailed analysis of these features, the location and nature of the fault source can be more accurately located. Subsequently, "through a deep belief network, vibration pattern recognition is performed based on the fault vibration feature group", that is, an unsupervised learning artificial neural network is used to automatically learn and classify different vibration patterns. The deep belief network consists of multiple levels of restricted Boltzmann machines, which can mine complex patterns from a large number of samples. After "obtaining a set of vibration patterns", "envelope spectrum analysis is performed on the set of vibration patterns". This method can help detect weak fault signals hidden in the background of noise, thereby "obtaining mechanical fault features". For example, in CNC machine tools, if the inner ring, outer ring or rolling element of the bearing is damaged, they will produce unique vibration patterns under specific working conditions. After envelope spectrum analysis, these patterns can clearly show the specific location of the fault, whether it is a bearing inner ring fault, a bearing outer ring fault or a bearing rolling element fault. Then, "modal separation of the mechanical fault features is performed through variational modal decomposition technology." This step is to decompose the complex vibration signal into a series of intrinsic mode functions (IMFs), each of which corresponds to a specific physical process. After "obtaining the fault mode sequence," "empirical mode decomposition is performed on the fault mode sequence." This decomposition method can further reveal the vibration characteristics in different frequency ranges, especially for transmission components such as gears. Here, "gear fault features include gear tooth breakage, gear pitting and gear cracks," which are directly related to the health of the gear. Once these fault features are discovered, maintenance measures can be taken in a timely manner to avoid more extensive damage.Then, "through a preset sparse autoencoder, electrical features of the manufacturing equipment are extracted based on the gear fault features." The autoencoder is a neural network architecture for dimensionality reduction and feature learning. Here, the "sparse" attribute is particularly emphasized, which means that only a few nodes that best express the fault characteristics are retained. Therefore, after "obtaining the electrical feature set," "Hilbert transforming the electrical feature set" is a technique for analyzing instantaneous frequency, which helps to "obtain electrical fault features." For example, in CNC machine tools, motors may work abnormally due to problems such as stator winding short circuits, rotor bar breakage, or electrical corrosion of motor bearings. The results after the Hilbert transform can provide key clues about these problems. Furthermore, "energy distribution analysis of the electrical fault features is performed through the Teager energy operator." The Teager energy operator can calculate the energy density of the signal, which is very useful for understanding the development process of the fault. After "obtaining the energy distribution spectrum," "multi-scale entropy analysis of the energy distribution spectrum" is performed. This method can quantify the complexity and irregularity of the system and then "obtain structural fault features." For example, if the base of a CNC machine tool is cracked, the coupling is broken, or the base is loose, these structural problems will appear as special patterns on the energy distribution map, and multi-scale entropy analysis can help us identify these patterns and attribute them to specific "structural fault characteristics". Finally, "through the fuzzy inference system, a comprehensive fault judgment is made based on the structural fault characteristics". Fuzzy logic allows us to make decisions under uncertainty. The fuzzy inference system here makes judgments based on the known combination of fault characteristics. After "obtaining the combination of fault characteristics", "the fault type analysis is performed on the combination of fault characteristics" to finally determine the category of the fault. For example, in CNC machine tools, the fault type may be mechanical wear fault, mechanical fracture fault, electrical short circuit fault, electrical open circuit fault, structural damage fault or structural deformation fault. Through such a complete process, engineers can obtain comprehensive and accurate fault diagnosis information so as to take targeted repair measures to ensure the safety and continuity of the production process. In summary, this fault type analysis method integrates a variety of advanced analysis tools and technologies, which not only improves the accuracy of fault diagnosis, but also speeds up the response speed, so that manufacturing companies can deal with complex and changeable equipment failures more calmly and ensure efficient production operation.

[0099] In a specific embodiment, the digital twin mapping of the manufacturing equipment based on the fault type to obtain the fault location of the manufacturing equipment and the fault component corresponding to the fault location includes:

[0100] Performing fault semantic decoding on the fault type by using a preset global feature embedding technology to obtain a fault semantic feature vector, and marking the fault trigger area of ​​the digital twin model of the manufacturing equipment based on the fault semantic feature vector to obtain an initial fault trigger area set;

[0101] Based on the reverse engineering modeling algorithm, geometric feature inversion is performed on the initial fault trigger area set to obtain a geometric space feature map of equipment faults, and local mesh refinement is performed on the fault geometric space feature map through a topology optimization algorithm to obtain a digital twin geometric mapping mesh;

[0102] Performing multi-physical field coupling simulation on the digital twin geometric mapping grid through dynamic physical simulation technology to obtain a multi-dimensional fault physical field distribution map, and performing fault field location on the multi-dimensional fault physical field distribution map to obtain a fault location parameter set;

[0103] By using a knowledge graph reasoning algorithm, the cascade relationship of the equipment components in the digital twin model is inferred based on the fault location parameter set to obtain a fault propagation association network, and node importance analysis is performed on the fault propagation association network to obtain a set of key fault nodes;

[0104] Based on the set of critical fault nodes, the digital twin model is three-dimensionally visualized to obtain a mapping result, and based on the mapping result, a fault location of the manufacturing equipment and a faulty component corresponding to the fault location are obtained.

[0105] Specifically, in the process of fault diagnosis of manufacturing equipment, it is a crucial step to accurately locate the fault location and faulty components by mapping the manufacturing equipment digitally based on the fault type. This process not only relies on advanced computing technology, but also integrates multidisciplinary knowledge to ensure that the nature of the fault can be accurately identified and understood. First, "decoding the fault semantics of the fault type through a preset global feature embedding technology" means using machine learning or deep learning algorithms to convert the fault type into a form that can be understood and processed by a computer - that is, a "fault semantic feature vector". For example, in the application scenario of CNC machine tools, when an abnormal increase in the temperature of the spindle bearing is detected, the system will generate a corresponding semantic feature vector based on the characteristics of this fault type, combined with historical data and expert knowledge. Then, "marking the fault trigger area of ​​the digital twin model of the manufacturing equipment based on the fault semantic feature vector" means applying these feature vectors to the virtual model of the CNC machine tool to mark the initial area that may cause this fault, forming an "initial fault trigger area set". It should be noted that the digital twin model is a pre-set and trained three-dimensional visualization model, and it corresponds to the manufacturing equipment. Next, "geometric feature inversion of the initial fault trigger area set based on the reverse engineering modeling algorithm" refers to reconstructing the true geometric form of the objects in the area by analyzing the marked area, that is, the "equipment fault geometric space feature map". In this process, engineers need to consider factors such as tolerances and material properties in actual manufacturing to ensure the accuracy of the model. "And local mesh refinement of the fault geometric space feature map through the topology optimization algorithm" means further improving the accuracy of the model, especially for those key parts that are sensitive to faults, by adjusting the mesh density to better capture details, and obtain the "digital twin geometric mapping grid". This step is particularly important for CNC machine tools, because even small changes may affect the performance of the entire system. Subsequently, "multi-physics field coupling simulation of the digital twin geometric mapping grid is performed through dynamic physical simulation technology", that is, creating a virtual environment in which the interaction between multiple physical phenomena such as mechanical stress, heat conduction, and electromagnetic effects can be simulated simultaneously to "obtain a multi-dimensional fault physical field distribution map." For example, on CNC machine tools, overheating of the spindle bearing may cause a series of chain reactions, including but not limited to changing the surrounding temperature field, affecting the lubrication effect, etc. Through the comprehensive analysis of these physical fields, "and locating the fault field of the multi-dimensional fault physical field distribution map", the specific location of the fault and its impact range are finally determined to form a "fault location parameter set".Then, "through the knowledge graph reasoning algorithm, the cascade relationship of the equipment components in the digital twin model is inferred based on the fault location parameter set", that is, with the help of the pre-built knowledge base (such as the working principle of the equipment, the connection between the components, etc.), analyze how the fault spreads from one point to other parts, and establish a "fault propagation association network". Continuing with the example of CNC machine tools, if it is found that the spindle bearing problem has caused the processing accuracy to decrease, then it is necessary to investigate whether it also affects the position accuracy of the tool or the supply path of the coolant. "And perform node importance analysis on the fault propagation association network" to find out the most critical links in the entire propagation chain and form a "fault critical node set". These nodes often determine the impact of the fault and the priority of the repair work. Finally, "based on the fault critical node set, the digital twin model is three-dimensionally visualized", that is, all the above information is intuitively displayed using computer graphics methods, so that maintenance personnel can directly see the exact location of the fault and the specific components involved, which is the "mapping result". Such visualization not only helps to quickly locate the problem, but also assists in formulating maintenance plans and improving work efficiency. At the same time, "obtaining the fault location of the manufacturing equipment and the faulty component corresponding to the fault location based on the mapping results" provides clear guidance for subsequent troubleshooting. For example, in the case of CNC machine tools, technicians can quickly lock the spindle bearings based on the three-dimensional images, evaluate their impact on the surrounding structures, and then take appropriate measures to repair or replace them. In summary, the entire process is a complex and closely connected technical chain. From the initial fault signal capture to the final visualization, each step is inseparable from the support of modern information technology. In this way, not only can the precise positioning of manufacturing equipment failures be achieved, but also the reasons behind the failures can be deeply understood, providing valuable reference for future preventive maintenance. This method not only improves the speed and accuracy of fault diagnosis, but also enhances the scientificity and predictability of equipment management, which is of great significance for ensuring production continuity and product quality.

[0106] In a specific embodiment, the fault propagation path analysis based on the fault location and the fault component corresponding to the fault location to obtain the associated fault propagation path and the potential risk components on the associated fault propagation path includes:

[0107] Performing causal relationship modeling on the fault location and the fault component corresponding to the fault location through a Bayesian network to obtain a preliminary fault propagation path, and performing probability reasoning on the preliminary fault propagation path to obtain a fault propagation probability matrix, wherein the fault propagation probability matrix includes the fault propagation probability between each fault component;

[0108] By using the Markov chain Monte Carlo method, a fault path simulation is performed based on the fault propagation probability matrix to obtain a set of fault propagation paths, and path frequency statistics are performed on the set of fault propagation paths to obtain a high-frequency fault propagation path, wherein the high-frequency fault propagation path includes multiple fault propagation paths and the occurrence frequencies corresponding to the multiple fault propagation paths;

[0109] Performing risk priority sorting on the high-frequency fault propagation paths to obtain a risk priority list, and performing weighted analysis on the risk priority list to obtain a priority-weighted fault propagation path;

[0110] The priority weighted fault propagation path is subjected to a timing relationship analysis through a dynamic Bayesian network to obtain an associated fault propagation path, and the components on the associated fault propagation path are subjected to a timing relationship analysis to obtain potential risk components on the associated fault propagation path.

[0111] Specifically, first, "the causal relationship modeling of the fault location and the faulty component corresponding to the fault location is performed through the Bayesian network", which means using the probabilistic graph model of the Bayesian network to represent the causal relationship between the faulty components. The Bayesian network can capture the dependencies between the components inside the device and quantify these relationships in the form of conditional probabilities. After "obtaining the preliminary fault propagation path", "probabilistic reasoning is performed on the preliminary fault propagation path", that is, the Bayesian inference algorithm is used to calculate the possibility of fault propagation between each faulty component, and finally a "fault propagation probability matrix" is formed. For example, in a CNC machine tool, if the spindle bearing fails, the fault propagation probability between it and other components such as the tool interface and the hydraulic system will be recorded. This matrix not only reflects the relationship between directly adjacent components, but also covers a wider range of indirect effects, providing basic data for subsequent analysis. Next, "the fault path simulation is performed based on the fault propagation probability matrix through the Markov chain Monte Carlo method", that is, the random simulation method is used to explore all possible fault propagation paths. The Markov chain Monte Carlo (MCMC) method can generate a series of state transition sequences according to certain rules under a given initial state, thereby simulating the propagation process of the fault in the device. After "obtaining a set of fault propagation paths", "and performing path frequency statistics on the set of fault propagation paths", that is, counting the frequency of occurrence of each path, and screening out "high-frequency fault propagation paths". In the application of CNC machine tools, this step can help identify the most common fault propagation modes, such as spindle bearing failure leading to tool damage or reduced processing accuracy. High-frequency fault propagation paths not only reveal the development trend of common problems, but also provide a basis for preventive measures. Then, "risk priority ranking of the high-frequency fault propagation paths" is to rank them according to the risk level of each path to form a "risk priority list". The risk assessment here takes into account the possibility of failure and the severity of its consequences to ensure that the most important issues are given priority. "And weighted analysis of the risk priority list" is to assign corresponding weights based on the importance of different factors, so as to obtain a "priority weighted fault propagation path". For example, in CNC machine tools, if a path involves critical processing steps or high-cost components, its weight will be higher, which means that more attention needs to be paid to the potential risks on this path. Finally, "a dynamic Bayesian network is used to perform a temporal relationship analysis on the priority weighted fault propagation path". A dynamic Bayesian network is an extended traditional Bayesian network that can process data that changes over time and is suitable for scenarios where faults develop over time. "The associated fault propagation path is obtained", that is, to determine which components have significant temporal correlation in fault propagation. "And a temporal correlation analysis is performed on the components on the associated fault propagation path", further refining these associations and identifying "potential risk components on the associated fault propagation path".In CNC machine tools, this step helps to find components that may not show obvious signs of failure at present but are likely to be affected in the future, such as loose support structures caused by spindle bearing failures. In summary, through such a process consisting of causal relationship modeling, fault path simulation, risk priority sorting, and time series relationship analysis, manufacturing companies can accurately predict how faults propagate throughout the equipment based on the fault location and identify potential risk components. This approach not only improves the speed and accuracy of fault diagnosis, but also provides a solid foundation for formulating scientific and reasonable maintenance strategies. In the application of CNC machine tools in automobile engine production lines, such a meticulous fault propagation path analysis method enables production managers to foresee the occurrence and development of problems in advance and take preventive measures, thereby improving production efficiency and product quality, reducing maintenance costs, and ensuring the continuity and stability of production.

[0112] In a specific embodiment, the generating of an optimization plan for manufacturing equipment based on the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location, and the fault type includes:

[0113] Extracting parameters of the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location, and the fault type through a multi-objective optimization algorithm to obtain an extracted parameter set;

[0114] Dynamically simulating the digital twin model of the manufacturing equipment based on the extracted parameter set to obtain an operation status prediction graph of the digital twin model, and performing time series analysis on the operation status prediction graph of the digital twin model to obtain a fault prediction time series;

[0115] By using a fault tree analysis method, the fault cause analysis is performed on the fault prediction time series to obtain a set of fault causes, and the set of fault causes is hierarchically analyzed to obtain a list of fault cause priorities;

[0116] The preset manufacturing equipment operation data is called from the database, and the preset manufacturing equipment operation data is optimized based on the fault cause priority list to generate an optimization plan for the manufacturing equipment.

[0117] Specifically, first, "parameters of the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location, and the fault type are extracted through a multi-objective optimization algorithm", which means using mathematical optimization methods to identify and quantify the key factors affecting the operation of the equipment. The multi-objective optimization algorithm can find the best balance between multiple conflicting objectives. For example, in CNC machine tools, it is necessary to consider both repairing the fault to resume production and taking into account maintenance costs and downtime. After "obtaining the extracted parameter set", these parameters not only include physical quantities directly related to the fault, such as vibration amplitude, temperature change, etc., but also cover operating conditions and environmental factors that may affect the performance of the equipment. Next, "dynamically simulate the digital twin model of the manufacturing equipment based on the extracted parameter set", that is, use a pre-established high-precision virtual model to conduct simulation experiments. The digital twin model can reflect the status of the actual equipment in real time and allow engineers to test different maintenance and improvement plans in a virtual environment. "Obtain the operation status prediction diagram of the digital twin model", which shows the expected performance of the equipment in the future, including the parameter change trend under normal working conditions and possible abnormal conditions. “Performing a time series analysis on the operation status prediction graph of the digital twin model”, this method can identify the timing of potential problems and form a “fault prediction time series”. For example, in a CNC machine tool, if a spindle bearing failure may cause tool damage or reduced machining accuracy, then time series analysis can help predict when these problems may occur, thereby providing a basis for preventive maintenance. Then, “performing a fault cause analysis on the fault prediction time series through a fault tree analysis method”, that is, building a logical structure to represent all possible causes of a specific fault and their interrelationships. Fault tree analysis is a systematic tool that starts from the final event (such as a fault) and gradually traces back to the root cause. “Obtaining a set of fault causes”, this set includes all factors that may cause the fault, such as design defects, material aging, improper operation, etc. “Performing a hierarchical analysis on the set of fault causes”, that is, ranking each factor according to its importance to form a “fault cause priority list”. For example, in a CNC machine tool, if the main cause of the spindle bearing failure is insufficient lubrication and the secondary cause is installation error, the former will be given a higher priority, which means that more attention needs to be paid to the maintenance of the lubrication system. Finally, the step of "calling preset manufacturing equipment operation data from the database and optimizing the preset manufacturing equipment operation data based on the fault cause priority list" aims to combine historical data and current analysis results to make specific optimization suggestions. "Generate an optimization plan for manufacturing equipment", which includes not only specific repair measures for the problems found, such as replacing worn parts and adjusting process parameters, but also long-term strategies, such as improving maintenance plans and introducing new monitoring technologies.For example, in CNC machine tools, if insufficient lubrication is found to be the main problem, the optimization plan may recommend increasing the frequency of lubricant replacement and installing smart sensors to monitor oil level and quality in real time; if installation error is also an important factor, the equipment may need to be recalibrated to ensure that all components are properly aligned. In summary, through such a process consisting of multi-objective optimization, dynamic simulation, time series analysis, fault tree analysis and hierarchical analysis, manufacturing companies can accurately predict the occurrence and development of faults based on the fault location and its associated risk components, and formulate a comprehensive and targeted optimization plan based on this. This approach not only improves the speed and accuracy of fault diagnosis, but also provides a solid foundation for formulating scientific and reasonable maintenance strategies. In the application of CNC machine tools in automobile engine production lines, such a meticulous optimization plan enables production managers to foresee the occurrence and development of problems in advance, take preventive measures, and thus improve production efficiency and product quality, reduce maintenance costs, and ensure the continuity and stability of production.

[0118] The above describes the intelligent manufacturing optimization method based on a large model in the embodiment of the present invention. The following describes the intelligent manufacturing optimization device based on a large model in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, an intelligent manufacturing optimization device based on a large model includes:

[0119] The acquisition module 21 is used to acquire the operating status of the manufacturing equipment through a preset sensor network to obtain operating parameters;

[0120] The extraction module 22 is used to detect whether the operating parameter is within a preset operating threshold through a preset detection mechanism, and if not, to extract fault parameters from the operating parameter through a preset fault extraction algorithm to obtain fault extraction feature parameters;

[0121] A first analysis module 23, configured to perform a fault type analysis on the manufacturing equipment based on the fault extraction characteristic parameters to obtain a fault type;

[0122] A mapping module 24, configured to perform digital twin mapping on the manufacturing equipment based on the fault type, and obtain a fault location of the manufacturing equipment and a fault component corresponding to the fault location;

[0123] A second analysis module 25 is used to perform fault propagation path analysis based on the fault location and the fault component corresponding to the fault location, and obtain an associated fault propagation path and a potential risk component on the associated fault propagation path;

[0124] The generation module 26 is used to generate an optimization plan for manufacturing equipment based on the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location and the fault type.

[0125] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.

[0126] Reference Figure 3 The present invention also provides a computer device in an embodiment, wherein the internal structure of the computer device can be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0127] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0128] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0129] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0130] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0131] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A large-model-based intelligent manufacturing optimization method, characterized in that: Applied to manufacturing equipment, including the following steps: The operating status of manufacturing equipment is collected through a preset sensor network to obtain operating parameters; By using a preset detection mechanism, it is detected whether the operating parameter is within a preset operating threshold. If not, a preset fault extraction algorithm is used to extract fault parameters from the operating parameter to obtain fault extraction feature parameters. Performing a fault type analysis on the manufacturing equipment based on the fault extraction characteristic parameters to obtain the fault type; Performing digital twin mapping on the manufacturing equipment based on the fault type to obtain a fault location of the manufacturing equipment and a faulty component corresponding to the fault location; Perform fault propagation path analysis based on the fault location and the fault component corresponding to the fault location to obtain associated fault propagation paths and potential risk components on the associated fault propagation paths; Generate an optimization plan for manufacturing equipment based on the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location, and the fault type; The performing digital twin mapping on the manufacturing equipment based on the fault type to obtain the fault location of the manufacturing equipment and the fault component corresponding to the fault location includes: Performing fault semantic decoding on the fault type by using a preset global feature embedding technology to obtain a fault semantic feature vector, and marking the fault trigger area of ​​the digital twin model of the manufacturing equipment based on the fault semantic feature vector to obtain an initial fault trigger area set; Based on the reverse engineering modeling algorithm, geometric feature inversion is performed on the initial fault trigger area set to obtain a geometric space feature map of equipment faults, and local mesh refinement is performed on the fault geometric space feature map through a topology optimization algorithm to obtain a digital twin geometric mapping mesh; Performing multi-physical field coupling simulation on the digital twin geometric mapping grid through dynamic physical simulation technology to obtain a multi-dimensional fault physical field distribution map, and performing fault field location on the multi-dimensional fault physical field distribution map to obtain a fault location parameter set; By using a knowledge graph reasoning algorithm, the cascade relationship of the equipment components in the digital twin model is inferred based on the fault location parameter set to obtain a fault propagation association network, and node importance analysis is performed on the fault propagation association network to obtain a set of key fault nodes; Based on the set of critical fault nodes, the digital twin model is three-dimensionally visualized to obtain a mapping result, and based on the mapping result, a fault location of the manufacturing equipment and a faulty component corresponding to the fault location are obtained.

2. The large model-based intelligent manufacturing optimization method according to claim 1 is characterized in that: The operation status of the manufacturing equipment is collected through a preset sensor network to obtain the operation parameters, including: Through the preset sensor network, multi-modal data collection is performed on key parts of the manufacturing equipment to obtain the original operating status data; Performing noise reduction and feature enhancement processing on the original operating status data to obtain pre-processed operating status data; Compressing the pre-processed running status data to obtain compressed running status data; The compressed operating state data is subjected to data fusion to obtain operating parameters; wherein the operating parameters include high-frequency vibration signals, temperature field distribution data, three-phase current waveforms, speed fluctuation signals and strain data of key parts.

3. The large model-based intelligent manufacturing optimization method according to claim 1 is characterized in that: The fault extraction algorithm includes a multi-branch residual feature extraction network, a hole convolution network, a depth-separable convolution, a deconvolution network, a cross-scale feature aggregation module and a dense connection module. The fault extraction algorithm is used to extract fault parameters from the operating parameters to obtain fault extraction feature parameters, including: Performing multi-scale feature extraction on the operating parameters through the multi-branch residual feature extraction network to obtain a multi-layer feature map, and performing attention mechanism fusion on the multi-layer feature map to obtain fused feature data; Based on the atrous convolutional network, the receptive field of the fused feature tensor is expanded to obtain an extended feature map, and the extended feature map is subjected to pyramid pooling analysis to obtain a multi-scale feature pyramid; wherein the multi-scale feature pyramid includes local detail features, meso-structure features and global semantic features; Performing lightweight processing on the multi-scale feature pyramid through the depthwise separable convolution to obtain a lightweight feature map, and performing channel attention recalibration on the lightweight feature map to obtain a recalibrated feature map; Based on the deconvolution network, the recalibrated feature map is upsampled and reconstructed to obtain a reconstructed feature map, and the reconstructed feature map is subjected to non-local correlation analysis to obtain feature association data; wherein the feature association data is a long-range dependency relationship between fault features; Performing multi-level feature representation on the feature association data through the cross-scale feature aggregation module to obtain a hierarchical feature representation, and performing adaptive feature selection on the hierarchical feature representation to obtain an optimal feature set; wherein the optimal feature set includes significant features, discriminative features and complementary features; Based on the dense connection module, feature reorganization and normalization processing are performed on the preferred feature set to obtain fault extraction feature parameters.

4. The large model-based intelligent manufacturing optimization method according to claim 1 is characterized in that: The performing fault type analysis on the manufacturing equipment based on the fault extraction characteristic parameter to obtain the fault type includes: Performing feature decomposition on the fault extraction feature parameters by using a spectral clustering algorithm to obtain a frequency spectrum feature matrix, and performing wavelet packet analysis on the frequency spectrum feature matrix to obtain a fault vibration feature group; wherein the fault vibration feature group includes a periodic impact feature, a non-stationary vibration feature, and a harmonic resonance feature; Vibration pattern recognition is performed based on the fault vibration feature group through a deep belief network to obtain a vibration pattern set, and envelope spectrum analysis is performed on the vibration pattern set to obtain mechanical fault features; wherein the mechanical fault features include bearing inner race fault, bearing outer race fault and bearing rolling element fault; The mechanical fault features are modally separated by variational mode decomposition technology to obtain a fault mode sequence, and the fault mode sequence is empirically decomposed to obtain gear fault features; wherein the gear fault features include gear tooth breakage, gear pitting and gear cracks; By using a preset sparse autoencoder, electrical features of the manufacturing equipment are extracted based on the gear fault features to obtain an electrical feature set, and the electrical feature set is Hilbert transformed to obtain electrical fault features; wherein the electrical fault features include stator winding short circuit, rotor bar breakage and motor bearing electrical corrosion; Performing energy distribution analysis on the electrical fault characteristics through the Teager energy operator to obtain an energy distribution spectrum, and performing multi-scale entropy analysis on the energy distribution spectrum to obtain structural fault characteristics; wherein the structural fault characteristics include base cracking, coupling breakage and base loosening; Through the fuzzy reasoning system, a comprehensive fault judgment is performed based on the structural fault characteristics to obtain a fault feature combination, and a fault type analysis is performed on the fault feature combination to obtain a fault type; wherein the fault types include mechanical wear faults, mechanical fracture faults, electrical short circuit faults, electrical open circuit faults, structural damage faults and structural deformation faults.

5. The large model-based intelligent manufacturing optimization method according to claim 1 is characterized in that: The performing of fault propagation path analysis based on the fault location and the fault component corresponding to the fault location to obtain the associated fault propagation path and the potential risk components on the associated fault propagation path includes: Performing causal relationship modeling on the fault location and the fault component corresponding to the fault location through a Bayesian network to obtain a preliminary fault propagation path, and performing probability reasoning on the preliminary fault propagation path to obtain a fault propagation probability matrix, wherein the fault propagation probability matrix includes the fault propagation probability between each fault component; By using the Markov chain Monte Carlo method, a fault path simulation is performed based on the fault propagation probability matrix to obtain a set of fault propagation paths, and path frequency statistics are performed on the set of fault propagation paths to obtain a high-frequency fault propagation path, wherein the high-frequency fault propagation path includes multiple fault propagation paths and the occurrence frequencies corresponding to the multiple fault propagation paths; Performing risk priority sorting on the high-frequency fault propagation paths to obtain a risk priority list, and performing weighted analysis on the risk priority list to obtain a priority-weighted fault propagation path; The priority weighted fault propagation path is subjected to a timing relationship analysis through a dynamic Bayesian network to obtain an associated fault propagation path, and the components on the associated fault propagation path are subjected to a timing relationship analysis to obtain potential risk components on the associated fault propagation path.

6. The large model-based intelligent manufacturing optimization method according to claim 5 is characterized in that: The generating of an optimization plan for manufacturing equipment based on the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location, and the fault type comprises: Extracting parameters of the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location, and the fault type through a multi-objective optimization algorithm to obtain an extracted parameter set; Dynamically simulating the digital twin model of the manufacturing equipment based on the extracted parameter set to obtain an operation status prediction graph of the digital twin model, and performing time series analysis on the operation status prediction graph of the digital twin model to obtain a fault prediction time series; By using a fault tree analysis method, the fault cause analysis is performed on the fault prediction time series to obtain a set of fault causes, and the set of fault causes is hierarchically analyzed to obtain a list of fault cause priorities; The preset manufacturing equipment operation data is called from the database, and the preset manufacturing equipment operation data is optimized based on the fault cause priority list to generate an optimization plan for the manufacturing equipment.

7. An intelligent manufacturing optimization device based on a large model, characterized in that: Applied to manufacturing equipment, including: A collection module is used to collect the operating status of the manufacturing equipment through a preset sensor network to obtain operating parameters; An extraction module, used to detect whether the operating parameters are within a preset operating threshold through a preset detection mechanism, and if not, to extract fault parameters from the operating parameters through a preset fault extraction algorithm to obtain fault extraction feature parameters; A first analysis module, configured to perform a fault type analysis on the manufacturing equipment based on the fault extraction characteristic parameters to obtain the fault type; A mapping module, used to perform digital twin mapping on the manufacturing equipment based on the fault type to obtain a fault location of the manufacturing equipment and a faulty component corresponding to the fault location; A second analysis module is used to perform fault propagation path analysis based on the fault location and the fault component corresponding to the fault location, to obtain an associated fault propagation path and a potential risk component on the associated fault propagation path; A generation module, configured to generate an optimization plan for manufacturing equipment based on the associated fault propagation path, the potential risk components on the associated fault propagation path, the fault location, the fault component corresponding to the fault location, and the fault type; The performing digital twin mapping on the manufacturing equipment based on the fault type to obtain the fault location of the manufacturing equipment and the fault component corresponding to the fault location includes: Performing fault semantic decoding on the fault type by using a preset global feature embedding technology to obtain a fault semantic feature vector, and marking the fault trigger area of ​​the digital twin model of the manufacturing equipment based on the fault semantic feature vector to obtain an initial fault trigger area set; Based on the reverse engineering modeling algorithm, geometric feature inversion is performed on the initial fault trigger area set to obtain a geometric space feature map of equipment faults, and local mesh refinement is performed on the fault geometric space feature map through a topology optimization algorithm to obtain a digital twin geometric mapping mesh; Performing multi-physical field coupling simulation on the digital twin geometric mapping grid through dynamic physical simulation technology to obtain a multi-dimensional fault physical field distribution map, and performing fault field location on the multi-dimensional fault physical field distribution map to obtain a fault location parameter set; By using a knowledge graph reasoning algorithm, the cascade relationship of the equipment components in the digital twin model is inferred based on the fault location parameter set to obtain a fault propagation association network, and node importance analysis is performed on the fault propagation association network to obtain a set of key fault nodes; Based on the set of critical fault nodes, the digital twin model is three-dimensionally visualized to obtain a mapping result, and based on the mapping result, a fault location of the manufacturing equipment and a faulty component corresponding to the fault location are obtained.

8. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images