A method and system for managing the entire life cycle of EAM equipment

Abstract

The present invention provides a method and system for the full life cycle management of EAM equipment, which relates to the technical field of equipment life cycle management. By constructing a full life cycle management system that integrates multi-source data fusion, physical constraint-driven intelligent prediction and root cause diagnosis, environmental condition adaptive maintenance strategy matching, multi-physics field simulation evaluation, and multi-objective optimization scheduling, the present invention achieves accurate perception of the status of key equipment parts and life prediction, improves the accuracy of fault diagnosis and the adaptability of maintenance plans, and optimizes maintenance resource allocation and inventory management. The system has dynamic adaptive optimization capabilities and can continuously improve maintenance strategies and prediction models based on actual operation feedback, thereby significantly improving the reliability and maintenance efficiency of equipment operation, reducing overall maintenance costs, and meeting the high standards of intelligent equipment management in complex industrial environments.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment lifecycle management methods, and in particular to an EAM equipment full lifecycle management method and system. Background Art

[0002] As equipment management continues to evolve toward intelligent and digital capabilities, the health monitoring and maintenance of key equipment components, particularly bearings, have become crucial for ensuring safe and efficient operation. Traditional bearing maintenance relies heavily on periodic inspections or empirical judgment, lacking accurate perception of actual operating conditions and lifespan prediction, making it difficult to make precise maintenance decisions. Furthermore, existing technologies for root cause analysis, maintenance solution matching, and maintenance effectiveness evaluation suffer from a single-dimensional approach and lack system integration and dynamic optimization, making them incapable of meeting the demands of intelligent management across multiple operating conditions and variables in complex industrial environments.

[0003] Prior art, publication number CN118014060A discloses a method and system for constructing a knowledge graph and generating maintenance recommendations for a water pump. The system obtains fault diagnosis results for the water pump; retrieves the applicable fault type from the knowledge graph; obtains applicable maintenance solutions from the knowledge graph based on the applicable fault type; and generates and outputs applicable maintenance recommendations for the water pump based on the applicable maintenance solutions. While this system can generate maintenance recommendations based on the knowledge graph, it focuses on static association rule matching and lacks dynamic prediction and adaptive optimization based on multi-source data fusion and physical constraints. The resulting maintenance rules lack real-world physical relevance and cannot meet the requirements for intelligent, real-time equipment lifecycle management in complex environments.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for managing the entire life cycle of EAM equipment to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for managing the entire life cycle of EAM equipment, including the following specific steps:

[0008] S1: Collect the operating data of the key bearings during the operation of the equipment and extract the time-space aligned feature vectors based on the topological map;

[0009] S2: Build a hybrid prediction model with real physical constraints. Use the hybrid prediction model and feature vectors to predict the remaining life of the bearing. When the predicted remaining life falls below a preset threshold, use the knowledge graph to perform root cause diagnosis.

[0010] S3: Generate an environmental condition label based on the operating data, and retrieve a maintenance method set that matches the environmental condition label and the root cause diagnosis result from a preset maintenance strategy library based on the root cause diagnosis result and the environmental condition label;

[0011] S4: Perform multi-physics simulation on bearings in key locations to simulate the impact of different maintenance methods on bearing performance, as well as their impact on production resources and inventory.

[0012] S5: Generate the optimal maintenance method based on the multi-objective optimization algorithm, and dynamically update the knowledge graph and maintenance strategy library based on the reinforcement learning mechanism.

[0013] Preferably, the operating data includes bearing operating condition data and environmental operating condition data, wherein:

[0014] Bearing operating condition data includes bearing vibration signals, surface temperature, surface image, and load pressure;

[0015] Environmental condition data include ambient temperature and ambient humidity;

[0016] The nodes of the topological graph include bearing components, sensors, and fault modes, and the edge relationships represent physical connections, thermal conduction, and fault propagation chains.

[0017] Preferably, the classification rules of the environmental condition labels are:

[0018] When the ambient temperature is greater than 80°C and the load pressure is greater than 50MPa, the environmental condition label is "high temperature and high pressure";

[0019] When the ambient temperature is greater than 80°C and the load pressure does not exceed 50MPa, the environmental condition label is "high temperature";

[0020] When the ambient temperature does not exceed 80°C and the load pressure is greater than 50MPa, the environmental operating condition label is "high pressure";

[0021] When the ambient temperature does not exceed 80°C and the load pressure does not exceed 50MPa, the environmental operating condition label is "normal temperature and pressure".

[0022] Preferably, the construction logic of the maintenance strategy library is:

[0023] Define standardized fault type labels based on the bearing failure mode;

[0024] Classify maintenance methods and establish effectiveness evaluation indicators for each maintenance method, including life extension rate, maintenance cost, and maintenance cycle;

[0025] Construct a two-dimensional label between the environmental condition label and the fault type label, and correspond each two-dimensional label to a set of maintenance methods.

[0026] Preferably, the hybrid prediction model is a time series prediction model, which takes the operating data of the bearing as the model input and the predicted remaining life of the bearing as the output, and introduces real physical constraints on the growth of bearing cracks to construct a loss function. The loss function expression is:

[0027] ;

[0028] In the formula represents the loss function, 、 Represent the predicted remaining life and the actual remaining life output by the model, Indicates the crack length on the bearing surface, 、 are material constants, represents the stress amplitude, represents the weight coefficient of the real physical constraint, and .

[0029] Preferably, the step S4 includes:

[0030] S401: Inputting corresponding process parameters according to the selected maintenance method, and simulating the thermodynamic and stress response of the bearing under actual working conditions based on the finite element simulation method;

[0031] S402: Collecting operating data of the bearing after maintenance to generate its predicted remaining life, and comparing it with the predicted remaining life before maintenance to evaluate the impact of different maintenance methods on bearing performance;

[0032] S403: Calculate the consumption of production resources by the selected maintenance method, and update the production inventory according to the consumption of production resources.

[0033] Preferably, the production resource consumption is calculated as follows:

[0034] ;

[0035] In the formula Indicates the The consumption of production resources by the maintenance method, 、 Respectively represent The first maintenance method The working hours and material consumption of each step, 、 Respectively represent the corresponding labor unit price and material unit price, subscript 、 The indexes representing the maintenance method and maintenance steps respectively;

[0036] The update expression of the production inventory is:

[0037] ;

[0038] In the formula 、 Respectively expressed in Production inventory before the first maintenance and Production inventory after the first repair, Indicates in First maintenance The consumption of production resources by the maintenance method, Indicates in The amount of production resources replenished during the first maintenance, represents the total number of methods in the maintenance method set, Indicates the selection judgment value of the maintenance method, Indicates not to select Maintenance methods, Indicates selection A maintenance method.

[0039] Preferably, the multi-objective optimization algorithm uses a weighted sum method to convert the multi-objective into a single-objective optimization, while adding constraints on the total number of maintenance methods and production inventory, and taking the maintenance method with the lowest single-objective optimization function value and satisfying the constraints as the optimal maintenance method. The expression of the multi-objective optimization algorithm is:

[0040] ;

[0041] In the formula represents the single-objective optimization function after weighted processing, 、 、 Both represent optimization weights greater than 0, and , 、 They represent the objective functions of production resource consumption and remaining life prediction, represents the penalty function on production inventory, Indicates the maximum limit value of the maintenance method;

[0042] The objective function expressions for production resource consumption and remaining life prediction are:

[0043] ;

[0044] ;

[0045] The penalty function expression for production inventory is:

[0046] ;

[0047] In the formula represents the predicted remaining life after repair, represents the maximum remaining lifetime for reference, Indicates the minimum threshold for production inventory.

[0048] An EAM equipment full lifecycle management system, which is used to implement the above-mentioned management method, specifically includes:

[0049] Data acquisition module, used to collect operating data of bearings in key parts of the equipment;

[0050] Feature extraction module, used to achieve spatiotemporal alignment and fusion of data according to the topological map and extract the corresponding feature vectors;

[0051] The prediction and diagnosis module is used to predict the remaining life of the bearing based on a hybrid prediction model with real physical constraints and feature vectors, and to perform root cause diagnosis using knowledge graphs when the remaining life falls below a threshold;

[0052] The maintenance strategy management module is used to retrieve a matching maintenance method set from the maintenance strategy library based on the root cause diagnosis results and environmental condition labels;

[0053] A simulation module is used to simulate the effects of different maintenance methods on the thermodynamic and mechanical responses of bearing performance, and to evaluate the impact of maintenance on production resource consumption and inventory status;

[0054] Multi-objective optimization module, which is used to build a multi-objective optimization model and generate the optimal maintenance plan based on production resource consumption, predicted remaining life and inventory status;

[0055] The dynamic update module is used to dynamically adjust the knowledge graph and maintenance strategy library based on maintenance execution feedback and combined with the reinforcement learning mechanism to achieve online adaptive optimization of the system.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] By constructing a full lifecycle management system that integrates multi-source data fusion, physical constraint-driven intelligent prediction and root cause diagnosis, adaptive maintenance strategy matching for environmental conditions, multi-physics field simulation and evaluation, and multi-objective optimization and scheduling, this system achieves precise sensing of the status of key equipment components and lifespan prediction, improving the accuracy of fault diagnosis and the adaptability of maintenance plans, and optimizing maintenance resource allocation and inventory management. The system possesses dynamic adaptive optimization capabilities and can continuously refine maintenance strategies and prediction models based on actual operational feedback, significantly improving equipment reliability and maintenance efficiency, reducing overall maintenance costs, and meeting the high standards required for intelligent equipment management in complex industrial environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 Schematic diagram of the overall method flow of the present invention;

[0059] Figure 2 It is a point-line graph of the hybrid prediction model data of the present invention;

[0060] Figure 3 It is a schematic diagram of the module structure of the present invention. DETAILED DESCRIPTION

[0061] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.

[0062] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0063] Example:

[0064] See also Figures 1 to 2 , the present invention provides a technical solution:

[0065] A method for managing the entire life cycle of EAM equipment, including the following specific steps:

[0066] S1: Collects operational data from bearings in key areas of the equipment during operation and extracts time-space aligned feature vectors based on the topological map. This data can be from a variety of mechanical equipment containing key bearings or rotating components. For example, in petrochemical drilling rigs, key components include main drive bearings, slewing table bearings, drill stem bearings, gearbox bearings, hydraulic cylinders, and piston rod bearings. Specific settings are determined by the type of equipment being monitored and expert experience.

[0067] Operation data includes bearing operating condition data and environmental operating condition data, including:

[0068] Bearing operating condition data includes bearing vibration signals, surface temperature, surface image, and load pressure;

[0069] Environmental condition data include ambient temperature and ambient humidity;

[0070] The nodes of the topological graph include bearing components, sensors, and fault modes, and the edge relationships represent physical connections, thermal conduction, and fault propagation chains.

[0071] A topological graph refers to a graph structure that expresses positions in space and their interconnectedness. It focuses on the connectivity and relative positions between nodes (positions), and on the network topological properties and spatiotemporal geometric features between nodes, rather than emphasizing specific geometric distances or coordinates. Compared with subsequent knowledge graphs, it emphasizes the physical connection structure and spatial relationships between entities, highlighting the assembly relationship between equipment components, sensor layout, and physical conduction paths (such as thermal and mechanical transfer). Since this step requires accurate reflection of the physical connection relationship between the bearing and its key components, the spatial correspondence between the sensor and the measured object, and the physical path of fault propagation, in order to facilitate the spatiotemporal synchronization and feature fusion of multi-source data, a topological graph suitable for expressing this type of physical space structure and topological properties is used.

[0072] Specifically, when constructing a topological graph, a graph database (such as Neo4j) or a dedicated knowledge graph platform can be used to store the topological graph structure. The first step is to define its node and edge relationships. Nodes can be categorized into bearing component nodes, sensor nodes, and fault mode nodes. Bearing component nodes represent the physical components of a bearing, such as the inner ring, outer ring, rollers, and cage. Sensor nodes represent various sensors installed in and near the bearing, such as vibration sensors, temperature sensors, image acquisition devices, and pressure sensors. Fault mode nodes represent various types of bearing faults, such as cracks, wear, and pitting. Edge relationships can be categorized into physical connection edges, thermal conduction edges, and fault propagation chain edges. Physical connection edges connect bearing component nodes, reflecting the assembly relationships and interactions between components in a real mechanical structure. Thermal conduction edges represent the heat transfer paths between components, describing the propagation patterns of temperature changes. Fault propagation chain edges reflect the causal relationships in the evolution and propagation of faults from one component to another. After the construction is completed, the collected bearing operating condition data (vibration, temperature, image, load pressure) and environmental operating condition data (ambient temperature and humidity) are associated with the corresponding physical nodes according to the sensor nodes to achieve spatial mapping of the data, and multi-source data is synchronized using timestamps to complete timing alignment to ensure the temporal and spatial consistency of the data.

[0073] In this step, by constructing a topological map, we achieve a comprehensive perception and deep understanding of the equipment's operating environment and status. This provides precise input features for the hybrid prediction model, supports causal reasoning for root-cause diagnosis, and effectively correlates failure modes with sensor data. Furthermore, it provides a structured knowledge foundation for precise matching of the maintenance strategy library and multi-physics simulation, ensuring the scientific and rational evaluation of maintenance plans.

[0074] S2: Build a hybrid prediction model with real physical constraints, predict the remaining life of the bearing based on the hybrid prediction model and feature vectors, and use knowledge graphs to complete root cause diagnosis when the predicted remaining life is lower than the preset threshold.

[0075] The hybrid prediction model is a time series prediction model that takes the operating data of the bearing as the model input and the predicted remaining life of the bearing as the output. It also introduces real physical constraints on the growth of bearing cracks to construct a loss function. The loss function expression is:

[0076] ;

[0077] In the formula represents the loss function, 、 Represent the predicted remaining life and the actual remaining life output by the model, Indicates the crack length on the bearing surface, which can be obtained from the surface image of the bearing. 、 are material constants, represents the stress amplitude, represents the weight coefficient of the real physical constraint, and .

[0078] It can be seen from the loss function expression of the hybrid prediction model that it consists of two parts. The first half is the traditional prediction error term, which represents the mean square error between the model's predicted remaining life and the actual remaining life. It is used to make the model output as close to the actual life data as possible to ensure prediction accuracy. It is a purely data-driven constraint, where the actual remaining life can be obtained through experiments or simulations; the second half of the loss function is based on the Paris model of bearing crack propagation, which shows that there is a specific power relationship between the crack propagation rate and the stress amplitude and crack length, thereby constraining the crack propagation dynamics predicted by the model to meet the real physical laws, thereby ensuring that the prediction process conforms to the actual crack development mechanism of the bearing.

[0079] In this embodiment, a high-intensity life test is performed on three different bearing samples to compare the ordinary pure data-driven model and the hybrid prediction model. The test conditions are as follows: the three bearing samples are subjected to a high-intensity life test with a load twice the maximum allowable load. When the crack exceeds one-third of its diameter, the life is considered to have ended. At the same time, the operating data of the three bearings are collected in real time during the test, and the data are input into the hybrid prediction model and the pure data-driven model for life prediction every 24 hours. Finally, the life of the first bearing sample ends at 486 hours, the life of the second bearing sample ends at 507 hours, and the life of the third bearing sample ends at 495 hours. The specific prediction data during the test are shown in the following table:

[0080] Table 1: Predicted remaining life data of different models

[0081]

[0082] From the data in the above table and Figure 3 It can be seen that the predicted remaining life of the pure data-driven model is higher in the early stage and becomes lower in the later stage, with a larger overall fluctuation. This is because the operating data will become worse after the rigid damage such as cracks begin to appear in the bearing sample, so its predicted remaining life will also undergo a stage-by-stage mutation; and the hybrid prediction model has better predictability for such changes in physical properties because it adds physical constraints in advance, and the predicted remaining life fluctuates less. In other words, compared with the pure data-driven model, the hybrid prediction model has no lag in prediction and has higher overall accuracy.

[0083] When the remaining bearing life predicted by the hybrid prediction model is lower than the preset threshold, it means that the equipment is at a high risk of failure. In this case, the steps for root cause diagnosis using the knowledge graph are as follows:

[0084] Map the feature vector output by the topology graph to the corresponding fault mode node in the knowledge graph, and identify the abnormal part in the feature vector (such as abnormal vibration mode, abnormal temperature, abnormal crack growth rate);

[0085] Leveraging the causal reasoning mechanisms within the knowledge graph (e.g., path search, probability inference, and rule matching), the fault propagation chain is traced step by step to identify the most likely root cause of the fault (e.g., crack propagation faults: inner race cracks, outer race cracks, etc.; pitting and corrosion faults: surface pitting, surface corrosion, etc.; poor lubrication faults: lack of lubrication, uneven lubrication, etc.).

[0086] Combining real-time data with knowledge graphs, the diagnostic conclusions are dynamically updated to improve diagnostic accuracy and interpretability.

[0087] In this step, there are various methods for identifying abnormalities in the feature vector, such as threshold methods, mathematical statistics, and time series detection. These are all mature existing technologies and will not be discussed in detail here. The preset threshold can be determined based on expert experience, for example, 10% to 20% of the theoretical bearing life, which can be obtained from the bearing production manual. By integrating data-driven remaining life prediction with physical crack propagation constraints based on fracture mechanics, the model's prediction accuracy and physical rationality can be improved. The knowledge graph performs root cause diagnosis, leveraging rich fault causal knowledge and real-time feature mapping mechanisms to accurately identify the root cause of the fault and support the scientific formulation of subsequent maintenance strategies.

[0088] S3: Generate an environmental condition label based on the operating data, and retrieve a maintenance method set that matches the environmental condition label and the root cause diagnosis result from a preset maintenance strategy library based on the root cause diagnosis result and the environmental condition label.

[0089] The classification rules for environmental condition labels are as follows:

[0090] When the ambient temperature is greater than 80°C and the load pressure is greater than 50MPa, the environmental condition label is "high temperature and high pressure";

[0091] When the ambient temperature is greater than 80°C and the load pressure does not exceed 50MPa, the environmental condition label is "high temperature";

[0092] When the ambient temperature does not exceed 80°C and the load pressure is greater than 50MPa, the environmental operating condition label is "high pressure";

[0093] When the ambient temperature does not exceed 80°C and the load pressure does not exceed 50MPa, the environmental operating condition label is "normal temperature and pressure".

[0094] The construction logic of the maintenance strategy library is:

[0095] Define standardized fault type labels (such as cracks, wear, pitting) based on the bearing failure mode;

[0096] Classify maintenance methods and establish effectiveness evaluation indicators for each maintenance method, including life extension rate, maintenance cost, and maintenance cycle;

[0097] A two-dimensional tag is constructed between the environmental condition tag and the fault type tag, and each two-dimensional tag is associated with a set of maintenance methods. The corresponding maintenance method set is stored using the root cause diagnosis fault type tag and the environmental condition tag as indexes. For example, if the root cause diagnosis result is "inner ring crack" and the environmental condition tag is "high temperature and high pressure," the corresponding maintenance method set in the maintenance strategy library might include: laser cladding, composite coating, precision grinding, etc.

[0098] When building a maintenance strategy library, the primary focus is on defining standardized fault type labels based on historical maintenance cases and expert experience. These labels should cover mechanical wear mechanisms, material degradation characteristics, and typical failure modes. Maintenance methods should also encompass diverse repair technologies, such as laser cladding, composite coatings, plasma spraying, lubrication upgrades, precision grinding, and dynamic balancing. These methods should be categorized and archived based on applicable fault types and environmental conditions, clearly defining the advantages and disadvantages of each method, implementation conditions, and effectiveness indicators.

[0099] In this step, the maintenance strategy library construction method ensures the scientific and systematic nature of maintenance solutions. Based on clear fault classification and environmental divisions, combined with a rich accumulation of maintenance technologies and an effectiveness evaluation system, this enables structured knowledge management. The maintenance method library's call logic enables efficient and accurate matching from diagnosis to maintenance solution, supporting dynamic adjustments and intelligent decision-making, and improving the overall effectiveness of maintenance management.

[0100] S4: Perform multi-physics simulation on bearings in key locations to simulate the impact of different maintenance methods on bearing performance, as well as their impact on production resources and inventory.

[0101] Step S4 includes:

[0102] S401: Inputting corresponding process parameters according to the selected maintenance method, and simulating the thermodynamic and stress response of the bearing under actual working conditions based on the finite element simulation method;

[0103] S402: Collecting operating data of the bearing after maintenance to generate its predicted remaining life, and comparing it with the predicted remaining life before maintenance to evaluate the impact of different maintenance methods on bearing performance;

[0104] S403: Calculate the consumption of production resources by the selected maintenance method, and update the production inventory according to the consumption of production resources.

[0105] Finite element-based thermodynamic and stress response simulations accurately replicate the physical performance changes of bearings after repair under actual operating conditions, surpassing traditional evaluation methods that rely solely on empirical evidence or single indicators. By comparing the predicted remaining life before and after repair, the positive and negative impacts of repairs on equipment performance are comprehensively quantified. Furthermore, by calculating the labor and material consumption for each step of the repair method and combining unit prices for cost accounting, accurate quantification of production resource consumption is achieved. Dynamic inventory updates ensure the timely and rational management of the material supply chain and spare parts.

[0106] The calculation method for production resource consumption is:

[0107] ;

[0108] In the formula Indicates the The consumption of production resources by the maintenance method, 、 Respectively represent The first maintenance method The working hours and material consumption of each step, 、 Respectively represent the corresponding labor unit price and material unit price, subscript 、 The indexes representing the maintenance method and maintenance steps respectively;

[0109] The update expression for production inventory is:

[0110] ;

[0111] In the formula 、 Respectively expressed in Production inventory before the first maintenance and Production inventory after the first repair, Indicates in First maintenance The consumption of production resources by the maintenance method, Indicates in The amount of production resources replenished during the first maintenance, represents the total number of methods in the maintenance method set, Indicates the selection judgment value of the maintenance method, Indicates not to select Maintenance methods, Indicates selection A maintenance method.

[0112] In this step, by introducing finite element simulation and refined resource consumption calculation, it can not only ensure that the maintenance effect is quantifiable and verifiable, provide reliable performance data support for the multi-objective optimization module, and improve the rationality of maintenance plan selection, but also promote the closed-loop management of the entire life cycle of equipment. Through maintenance effect feedback and resource consumption data, it assists in the dynamic update module to optimize the maintenance knowledge base and prediction model, promote intelligent decision-making and adaptive optimization, and improve overall operation and maintenance efficiency and equipment reliability.

[0113] S5: Generate the optimal maintenance method based on the multi-objective optimization algorithm, and dynamically update the knowledge graph and maintenance strategy library based on the reinforcement learning mechanism.

[0114] The multi-objective optimization algorithm uses the weighted sum method to convert the multi-objective into a single-objective optimization. At the same time, it adds constraints on the total number of maintenance methods and production inventory. The maintenance method with the lowest single-objective optimization function value and satisfying the constraints is regarded as the optimal maintenance method. The expression of the multi-objective optimization algorithm is:

[0115] ;

[0116] In the formula represents the single-objective optimization function after weighted processing, 、 、 Both represent optimization weights greater than 0, and , 、 They represent the objective functions of production resource consumption and remaining life prediction, represents the penalty function on production inventory, Indicates the maximum limit value of the maintenance method.

[0117] The objective function expressions for production resource consumption and remaining life prediction are:

[0118] ;

[0119] ;

[0120] The penalty function expression for production inventory is:

[0121] ;

[0122] In the formula represents the predicted remaining life after repair, represents the maximum remaining lifetime for reference, Indicates the minimum threshold for production inventory. The maximum remaining life for reference can be obtained by subtracting the operating time from the theoretical life of the bearing.

[0123] From the expression of the multi-objective optimization algorithm, it can be seen that the smaller the function value of the first single-objective optimization function, the better. The smaller the function value, the more likely it is to minimize maintenance costs and resource usage while maximizing the maintenance effect as much as possible, extending the remaining life of the bearing, and thus the life of the entire equipment. At the same time, a penalty function is introduced to significantly increase the function value of the single-objective optimization function after the production inventory falls below the minimum threshold. This is intended to prevent inventory from falling below the safety threshold and ensure the supply of maintenance materials. The second term is used to limit the maximum number of maintenance methods selected in the maintenance plan, ensuring the practical feasibility and simplicity of the maintenance plan. The third term is used to ensure that the inventory after maintenance is not negative, avoiding the risk of resource shortages and production stagnation.

[0124] The logic of dynamically updating the knowledge graph and maintenance strategy library based on the reinforcement learning mechanism can be expressed as follows:

[0125] Collect current equipment operating status, fault diagnosis results, maintenance plan execution results and related environmental information to form a system status representation;

[0126] According to the current state, select or adjust the maintenance strategy from the knowledge graph and maintenance strategy library as the action output;

[0127] Implement the selected maintenance strategy, monitor the equipment's operating status and resource consumption after maintenance, and collect feedback information;

[0128] Calculate reward signals and evaluate the quality of strategy execution based on maintenance results (such as increased remaining life, reduced resource consumption, and decreased failure rate) and inventory status;

[0129] Use reinforcement learning algorithms (such as Q-learning, policy gradient, etc.) to adjust the policy selection strategy based on the reward signal, and update the association weights in the knowledge graph and the content of the maintenance policy library;

[0130] Repeat the above process, continuously optimize the knowledge graph and maintenance strategy through environmental interaction and learning, and achieve adaptive dynamic updates.

[0131] The reward signal here can be set to the function value of the single-objective optimization function that satisfies the other two constraints in the multi-objective optimization algorithm. The smaller the function value, the higher the quality of strategy execution. Reinforcement learning algorithms are all existing technologies, and their specific calculation methods are not described in detail again.

[0132] In this step, by building an optimization model that takes into account multiple objectives and constraints, we can not only provide a scientific and reasonable basis for maintenance plan decision-making, ensuring that the maintenance plan is not only technically effective and economically feasible, but also taking into account production resources and inventory conditions. It can also enhance the overall intelligence and adaptability of the plan, achieve closed-loop optimization and continuous improvement of the equipment lifecycle management system, and enhance equipment reliability and operation and maintenance efficiency. Furthermore, through reinforcement learning mechanisms, new information can be absorbed, improving the accuracy and timeliness of fault diagnosis and maintenance recommendations.

[0133] See also Figure 3 This embodiment also provides an EAM equipment full life cycle management system, which is used to execute the above management method, specifically including:

[0134] Data acquisition module, used to collect operating data and environmental conditions data of bearings in key parts of the equipment;

[0135] Feature extraction module, used to achieve spatiotemporal alignment and fusion of data according to the topological map and extract the corresponding feature vectors;

[0136] The prediction and diagnosis module is used to predict the remaining life of the bearing based on a hybrid prediction model with real physical constraints and feature vectors, and to perform root cause diagnosis using knowledge graphs when the remaining life falls below a threshold;

[0137] The maintenance strategy management module is used to retrieve a matching maintenance method set from the maintenance strategy library based on the root cause diagnosis results and environmental condition labels;

[0138] A simulation module is used to simulate the effects of different maintenance methods on the thermodynamic and mechanical responses of bearing performance, and to evaluate the impact of maintenance on production resource consumption and inventory status;

[0139] Multi-objective optimization module, which is used to build a multi-objective optimization model and generate the optimal maintenance plan based on production resource consumption, predicted remaining life and inventory status;

[0140] The dynamic update module is used to dynamically adjust the knowledge graph and maintenance strategy library based on maintenance execution feedback and combined with the reinforcement learning mechanism to achieve online adaptive optimization of the system.

[0141] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0142] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraints of the technical solution.

[0143] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0144] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for managing the entire life cycle of EAM equipment, characterized in that: The specific steps include: S1: Collect the operating data of the key bearings during the operation of the equipment and extract the time-space aligned feature vectors based on the topological map; S2: Build a hybrid prediction model with real physical constraints. Use the hybrid prediction model and feature vectors to predict the remaining life of the bearing. When the predicted remaining life falls below a preset threshold, use the knowledge graph to perform root cause diagnosis. The hybrid prediction model is a time series prediction model that uses the operating data of the bearing as the model input and the predicted remaining life of the bearing as the output. It also introduces real physical constraints on the growth of bearing cracks to construct a loss function. The loss function expression is: In the formula represents the loss function, 、 Represent the predicted remaining life and the actual remaining life output by the model, Indicates the crack length on the bearing surface, 、 are material constants, represents the stress amplitude, represents the weight coefficient of the real physical constraint, and ; S3: Generate an environmental condition label based on the operating data, and retrieve a maintenance method set that matches the environmental condition label and the root cause diagnosis result from a preset maintenance strategy library based on the root cause diagnosis result and the environmental condition label; S4: Perform multi-physics simulation on bearings in key locations to simulate the impact of different maintenance methods on bearing performance, as well as their impact on production resources and inventory. The step S4 comprises: S401: Inputting corresponding process parameters according to the selected maintenance method, and simulating the thermodynamic and stress response of the bearing under actual working conditions based on the finite element simulation method; S402: Collecting operating data of the bearing after maintenance to generate its predicted remaining life, and comparing it with the predicted remaining life before maintenance to evaluate the impact of different maintenance methods on bearing performance; S403: Calculate the consumption of production resources by the selected maintenance method and update the production inventory according to the consumption of production resources; S5: Generate the optimal maintenance method based on the multi-objective optimization algorithm, and dynamically update the knowledge graph and maintenance strategy library based on the reinforcement learning mechanism.

2. The EAM equipment full life cycle management method according to claim 1, characterized in that: The operating data includes bearing operating condition data and environmental operating condition data, wherein: Bearing operating condition data includes bearing vibration signals, surface temperature, surface image, and load pressure; Environmental condition data include ambient temperature and ambient humidity; The nodes of the topological graph include bearing components, sensors, and fault modes, and the edge relationships represent physical connections, thermal conduction, and fault propagation chains.

3. The EAM equipment full life cycle management method according to claim 2, characterized in that: The classification rules of the environmental condition labels are as follows: When the ambient temperature is greater than 80°C and the load pressure is greater than 50MPa, the environmental condition label is "high temperature and high pressure"; When the ambient temperature is greater than 80°C and the load pressure does not exceed 50MPa, the environmental condition label is "high temperature"; When the ambient temperature does not exceed 80°C and the load pressure is greater than 50MPa, the environmental operating condition label is "high pressure"; When the ambient temperature does not exceed 80°C and the load pressure does not exceed 50MPa, the environmental operating condition label is "normal temperature and pressure".

4. The EAM equipment full life cycle management method according to claim 3, characterized in that: The construction logic of the maintenance strategy library is: Define standardized fault type labels based on the bearing failure mode; Classify maintenance methods and establish effectiveness evaluation indicators for each maintenance method, including life extension rate, maintenance cost, and maintenance cycle; Construct a two-dimensional label between the environmental condition label and the fault type label, and correspond each two-dimensional label to a set of maintenance methods.

5. The EAM equipment full life cycle management method according to claim 1, characterized in that: The calculation method of the production resource consumption is: In the formula Indicates the The consumption of production resources by the maintenance method, 、 Respectively represent The first maintenance method The working hours and material consumption of each step, 、 Respectively represent the corresponding labor unit price and material unit price, subscript 、 The indexes representing the maintenance method and maintenance steps respectively; The update expression of the production inventory is: In the formula 、 Respectively expressed in Production inventory before the first maintenance and Production inventory after the first repair, Indicates in First maintenance The consumption of production resources by the maintenance method, Indicates in The amount of production resources replenished during the first maintenance, represents the total number of methods in the maintenance method set, Indicates the selection judgment value of the maintenance method, Indicates not to select Maintenance methods, Indicates selection A maintenance method.

6. The EAM equipment full life cycle management method according to claim 5, characterized in that: The multi-objective optimization algorithm uses the weighted sum method to convert multiple objectives into single-objective optimization, while adding constraints on the total number of maintenance methods and production inventory, and takes the maintenance method with the lowest single-objective optimization function value and satisfying the constraints as the optimal maintenance method. The expression of the multi-objective optimization algorithm is: In the formula represents the single-objective optimization function after weighted processing, 、 、 Both represent optimization weights greater than 0, and , 、 They represent the objective functions of production resource consumption and remaining life prediction, represents the penalty function on production inventory, Indicates the maximum limit value of the maintenance method; The objective function expressions for production resource consumption and remaining life prediction are: The penalty function expression for production inventory is: In the formula represents the predicted remaining life after repair, represents the maximum remaining lifetime for reference, Indicates the minimum threshold for production inventory.

7. An EAM equipment full life cycle management system, characterized by: The management system is used to execute the management method according to any one of claims 1 to 6, specifically comprising: Data acquisition module, used to collect operating data of bearings in key parts of the equipment; Feature extraction module, used to achieve spatiotemporal alignment and fusion of data according to the topological map and extract the corresponding feature vectors; The prediction and diagnosis module is used to predict the remaining life of the bearing based on a hybrid prediction model with real physical constraints and feature vectors, and to perform root cause diagnosis using knowledge graphs when the remaining life falls below a threshold; The maintenance strategy management module is used to retrieve a matching maintenance method set from the maintenance strategy library based on the root cause diagnosis results and environmental condition labels; A simulation module is used to simulate the effects of different maintenance methods on the thermodynamic and mechanical responses of bearing performance, and to evaluate the impact of maintenance on production resource consumption and inventory status; Multi-objective optimization module, which is used to build a multi-objective optimization model and generate the optimal maintenance plan based on production resource consumption, predicted remaining life and inventory status; The dynamic update module is used to dynamically adjust the knowledge graph and maintenance strategy library based on maintenance execution feedback and combined with the reinforcement learning mechanism to achieve online adaptive optimization of the system.

Images