Motor part whole-process collaborative management method and system based on digital convergence

Abstract

The application provides a motor parts whole-process collaborative management method and system based on digital and intelligent fusion, which comprises the following steps: collecting micro-physical and chemical parameters and manufacturing process stress of key parts of the motor at the supply chain end, and generating an initial health baseline of each motor through orthogonal coupling. A damage measurement model is established, the damage acceleration is calculated according to the real-time working condition, the physical time is converted into virtual damage equivalent for deduction, and the remaining virtual life is updated. The remaining virtual life is used to perform differentiated scheduling, and a predictive maintenance work order is generated when it is lower than the threshold. When the motor group appears common degradation, the congenital and acquired characteristics are traced back, the common factor combination is identified through cluster analysis, and the quality rectification strategy is generated according to the combination and fed back to the supply chain end. The application has the effect of improving the motor life prediction accuracy.

Description

Technical Field

[0001] This invention belongs to the field of motor lifecycle management technology, specifically involving a collaborative management method and system for the entire process of motor components based on digital intelligence fusion. Background Technology

[0002] In modern industrial automation systems, health management of large-scale motor clusters is a core element in ensuring production continuity and maximizing equipment asset benefits. With the development of the Industrial Internet of Things (IIoT) and predictive maintenance technologies, existing motor management systems can now achieve real-time acquisition and analysis of operating status data for in-service motors. By constructing physical models or data-driven degradation models, these systems can assess the health status of motors and predict their remaining effective lifespan, thereby achieving significant results in avoiding unplanned downtime and optimizing maintenance resources.

[0003] However, current predictive maintenance models typically begin their analysis at the moment the motor enters service, treating all motors of the same model as standard components with the same initial health condition. Subsequent lifespan predictions are based on this idealized assumption. But the physical reality is that no two motors are exactly alike. At the time of manufacture, the microscopic physicochemical properties of each motor's internal key components exhibit unavoidable dispersion due to differences in raw material batches and suppliers. Furthermore, the process stresses these components experience during manufacturing and assembly vary. These individual differences, already solidified at the manufacturing stage, constitute a unique initial damage state, fundamentally determining their different degradation trajectories under the same service conditions. These issues make it difficult for existing predictive maintenance models to improve the accuracy of predicting the lifespan of motors and motor components, thus hindering the full lifecycle optimization management of motors and motor components. Summary of the Invention

[0004] This invention provides a method and system for collaborative management of motor components throughout the entire process based on digital intelligence integration, in order to solve the above-mentioned technical problems.

[0005] In a first aspect, the present invention provides a method for collaborative management of motor components throughout the entire process based on digital intelligence fusion, the method comprising the following steps: Microscopic physicochemical parameters of key motor components are collected as innate characteristic data during the supply stage of the supply chain, and process stress data of key motor components are collected as acquired characteristic data during the motor manufacturing and assembly stage of the supply chain. The innate characteristic data and the acquired characteristic data are mapped and associated through an orthogonal coupling algorithm to generate an initial health baseline that characterizes the initial damage state of each individual motor in the motor group. A damage measurement model incorporating an accelerated aging model is established based on the initial health baseline, and operational status data is collected in real time during the service phase of the motor group. Based on the operating status data and combined with the damage measurement model, the damage acceleration of the motor under the current operating conditions relative to the initial health baseline is calculated. The physical operating time is converted into virtual damage equivalent and accumulated and deducted to update the remaining virtual life of each individual motor in real time. When allocating production tasks, differentiated collaborative scheduling is performed for motors based on their remaining virtual lifespan; When the remaining virtual life of the motor is detected to be lower than the preset safety threshold, a predictive maintenance work order is generated based on the initial damage status marked in the initial health baseline, and a spare parts procurement request is sent to the supply chain based on the predictive maintenance work order. When multiple individual motors in a motor group are found to exhibit the same degradation characteristics, the innate and acquired characteristic data of the individual motors exhibiting degradation characteristics are traced back, and cluster analysis algorithms are used to identify the common factor combinations that lead to the degradation characteristics. Quality rectification strategies for motor components generated based on combinations of common factors are fed back to the supply chain.

[0006] Optionally, the step of collecting microscopic physicochemical parameters of key motor components as innate characteristic data during the supply stage of the supply chain, and collecting process stress data of key motor components as acquired characteristic data during the motor manufacturing and assembly stage of the supply chain, and mapping and associating the innate characteristic data and acquired characteristic data through an orthogonal coupling algorithm to generate an initial health baseline characterizing the initial damage state of each individual motor in the motor group includes the following steps: During the supply stage of the supply chain, the microscopic physicochemical parameters of key motor components are collected, and multidimensional feature extraction is performed on the microscopic physicochemical parameters to construct innate feature data that characterizes the inherent physical properties. In the motor manufacturing and assembly stage of the supply chain, process stress data of key motor components are collected, and time series analysis is performed on the process stress data to construct acquired characteristic data that characterizes the distribution of acquired stress. Principal component analysis is used to reduce the dimensionality and noise of both innate and acquired feature data to obtain a standardized input vector set. A two-dimensional orthogonal coupling matrix containing material sensitivity coefficients is constructed. The input vector set is input into the two-dimensional orthogonal coupling matrix for cross tensor calculation, and the cross penalty weights of each key motor component under different stress dimensions are output. The ideal health benchmark preset by the motor is weighted and corrected based on the cross-penalty weights of each key motor component to generate an initial health baseline that characterizes the initial damage state of the individual motor.

[0007] Optionally, the step of inputting the input vector set into a two-dimensional orthogonal coupling matrix for cross tensor calculation and outputting the cross penalty weights of each key motor component under different stress dimensions includes the following steps: The input vector set is divided into a material dimension subset and a process dimension subset; Extract the orthogonal basis vectors from the two-dimensional orthogonal coupling matrix, use the orthogonal basis vectors to perform spatial projection on the material dimension subset and the process dimension subset, and calculate the nonlinear coupling coefficient between the material dimension subset and the process dimension subset in the projection space; The nonlinear coupling coefficient is compared with a preset damage sensitivity threshold to filter out strongly coupled nodes that exceed the damage sensitivity threshold. The selected strongly coupled nodes are reverse-mapped to calculate the corresponding stress penalty value, and the stress penalty value is aggregated and output as the cross-penalty weight of each key motor component under different stress dimensions.

[0008] Optionally, the step of spatially projecting the material dimension subset and the process dimension subset using orthogonal basis vectors and calculating the nonlinear coupling coefficient between the material dimension subset and the process dimension subset in the projected space includes the following steps: Calculate the projected coordinates of the material dimension subset and the process dimension subset on the orthogonal basis vectors; A high-order covariance matrix is ​​constructed based on projected coordinates, and the principal eigenvalues ​​of the high-order covariance matrix are extracted by eigenvalue decomposition. Substitute the principal eigenvalues ​​into the preset Lorentz nonlinear activation function to calculate the initial coupling values ​​of the material dimension subset and the process dimension subset in the projection space. By using an ambient temperature compensation factor to correct the initial coupling value, the nonlinear coupling coefficients of the material dimension subset and the process dimension subset in the projection space are obtained.

[0009] Optionally, the step of calculating the damage acceleration of the motor relative to the initial healthy baseline under the current operating condition based on the operating status data and combined with the damage measurement model, and converting the physical operating time into virtual damage equivalent for cumulative deduction, so as to update the remaining virtual life of each individual motor in real time, includes the following steps: Ambient temperature, stator current amplitude, and rotor vibration frequency are extracted from the operating status data as dynamic input parameters for the damage measurement model; The initial damage state in the initial healthy baseline is transformed into the initial boundary conditions of the accelerated aging model. Run the damage measurement model and calculate the transient damage rate of the motor under the current operating conditions based on dynamic input parameters and initial boundary conditions; The damage acceleration under the current working condition is obtained by time integral calculation of the transient damage rate, and the damage acceleration is multiplied by the physical running time to obtain the virtual damage equivalent. The virtual damage equivalent is cumulatively deducted from the preset total design life of the motor unit to update the remaining virtual life of the motor unit in real time.

[0010] Optionally, the step of calculating the damage acceleration under the current operating condition by integrating the transient damage rate over time, and then multiplying the damage acceleration by the physical running time to obtain the virtual damage equivalent, includes the following steps: The Runge-Kutta algorithm was used to calculate the transient damage rate by time integration, and preliminary integration results were obtained. The initial integration result is smoothed by using the Kalman filter algorithm, and the smoothed initial integration result is defined as the damage acceleration under the current working condition. The basic damage equivalent is calculated by multiplying the damage acceleration by the corresponding physical running time. The basic damage equivalent is then corrected using a nonlinear correction factor based on the fluctuation of the real-time operating condition spectrum to calculate the virtual damage equivalent.

[0011] Optionally, the step of identifying the common factor combinations leading to the decline characteristics using a clustering analysis algorithm includes the following steps: The inherent and acquired characteristic data of the motor group exhibiting degradation characteristics are extracted, and the inherent and acquired characteristic data are fused to construct a high-dimensional fault characteristic matrix; The density peak clustering algorithm is used to iteratively cluster the high-dimensional fault feature matrix to identify multiple local high-density clusters in the high-dimensional fault feature matrix; Calculate the intra-class similarity and inter-class dissimilarity of each local high-density cluster; Based on intra-class similarity and inter-class dissimilarity, discrete clusters with intra-class similarity below a preset similarity threshold are removed, while the core clusters are retained. Extract the combination of feature elements that appear most frequently in the core cluster as the common factor combination that leads to the decay feature.

[0012] Optionally, the step of iteratively clustering the high-dimensional fault feature matrix using the density peak clustering algorithm to identify multiple local high-density clusters in the high-dimensional fault feature matrix includes the following steps: Extract the local neighborhood set of each data point in the high-dimensional fault feature matrix, and dynamically calculate the local feature weight matrix of each data point based on the variance distribution of each feature dimension in the local neighborhood set; The local feature weight matrix is ​​used to perform anisotropic space mapping on the data points in the high-dimensional fault feature matrix, and the higher-order cutoff distance of each data point is calculated in the anisotropic space to obtain the anisotropic local density of each data point. For each data point, search for the target neighbor node in the local neighborhood set whose anisotropic local density is greater than that of each data point and whose anisotropic spatial distance is the smallest, and establish a density gradient directed edge from each data point to the target neighbor node. Traverse all data points in the high-dimensional fault feature matrix and connect all density gradient directed edges to construct a density gradient directed forest that represents the clustering trend of the data. The root node in the density gradient directed forest with an in-degree greater than zero and an out-degree of zero is identified as the density peak. By performing reverse tracing along the density gradient directed edges in the density gradient directed forest, all data points converging to the same density peak are divided into the same cluster set, and the aggregated cluster sets are output as multiple local high-density clusters in the high-dimensional fault feature matrix.

[0013] Secondly, the present invention also provides a collaborative management system for the entire process of motor components based on digital intelligence fusion, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the collaborative management method for the entire process of motor components based on digital intelligence fusion as described in any one of the first aspects.

[0014] Thirdly, the present invention also provides a computer-readable storage medium storing instructions, characterized in that, when executed by a processor, the instructions cause the processor to be configured to perform the whole-process collaborative management method for motor components based on digital intelligence fusion according to any one of the first aspects.

[0015] The beneficial effects of this invention are: This invention fundamentally solves the problem of insufficient lifespan prediction accuracy in traditional motor management due to the inability to quantify individual initial state differences by constructing an initial health baseline that integrates the microscopic physicochemical parameters of key components and the process stresses of manufacturing. This invention establishes a damage measurement model specific to each individual motor unit, based on its unique initial health baseline. This model can convert physical operating time into a nonlinear virtual damage equivalent for accumulation according to real-time operating conditions, thereby achieving personalized tracking of the remaining lifespan of each unit and significantly improving the accuracy of lifespan prediction. Based on the remaining virtual lifespan, at the operational level, differentiated collaborative scheduling and precise predictive maintenance can maximize asset utilization and maintenance costs; at the strategic level, by tracing and analyzing common factors of group decline, performance data during the service phase can be transformed into quality improvement strategies for the upstream of the supply chain. Attached Figure Description

[0016] Figure 1This is a flowchart illustrating a method for collaborative management of motor components based on digital intelligence fusion in one embodiment of this application. Detailed Implementation

[0017] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0018] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0019] Figure 1 This is a flowchart illustrating a method for collaborative management of motor components throughout the entire process based on data and intelligence fusion, as shown in one embodiment. It should be understood that, although... Figure 1 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figure 1 At least some steps in the process may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or in turn with other steps or at least a portion of the sub-steps or stages of other steps. For example Figure 1 As shown, the method for collaborative management of motor components throughout the entire process based on digital intelligence fusion disclosed in this invention specifically includes the following steps: S101. Collect the microscopic physicochemical parameters of key motor components as innate characteristic data during the supply stage of the supply chain, and collect the process stress data of key motor components as acquired characteristic data during the motor manufacturing and assembly stage of the supply chain. Map and associate the innate characteristic data and the acquired characteristic data through an orthogonal coupling algorithm to generate an initial health baseline that characterizes the initial damage state of each individual motor in the motor group.

[0020] The process involves collecting microscopic physicochemical parameters of key motor components as intrinsic characteristic data and process stress data during assembly as acquired characteristic data. An orthogonal coupling algorithm maps and correlates these two types of data to generate an initial health baseline characterizing the initial damage state. During implementation, intrinsic material properties such as silicon steel sheet grain size and copper wire purity are obtained using electron microscopy and spectral analysis. Residual stress and deformation tensor information are simultaneously collected during stamping and wiring. The orthogonal coupling algorithm combines constitutive equations from materials science with stress distribution patterns in manufacturing engineering, mapping dissimilar data to an orthogonal characteristic space. This quantifies microscopic physical differences into macroscopic digital benchmarks, objectively reflecting the manufacturing damage accumulated in components before use, and completely changing the traditional perception that motors are shipped in a perfect, zero-damage state.

[0021] To achieve deep integration of physical and digital models across the entire lifecycle of motor components, a warehousing and logistics collaboration mechanism based on unique identifiers is further introduced at the supply and manufacturing / assembly stages of the aforementioned supply chain. Specifically, components are categorized based on their physical properties and manufacturing characteristics for differentiated tracking. For electromagnetic core materials such as magnets, copper wires, and iron cores, which are used directly after warehousing, their physical and chemical compliance indicators are rigorously inspected during the procurement and warehousing process. The system automatically generates a unique digital identification code containing supplier information, batch number, and initial physical and chemical parameters, and binds this code to the physical label. Throughout the subsequent warehousing, storage, and assembly processes, the component's flow status is updated in real time via barcode scanning devices. This ensures that every part on the assembly line can be accurately traced back to its initial health baseline data in the digital twin system, thereby seamlessly integrating the algorithm model with the warehousing management system.

[0022] S102. Establish a damage measurement model that includes an accelerated aging model based on the initial health baseline, and collect operational status data in real time during the service phase of the motor group.

[0023] This implementation involves establishing a damage measurement model incorporating an accelerated aging model based on an established initial health baseline, and collecting operational status data in real time during service. Accelerated aging models, originally used primarily for ultimate life assessment of electronic components, are transformed in this implementation into continuous fatigue analysis for rotating machinery. In model construction, the initial health baseline serves as the starting point for the evolution of microcracks or insulation degradation. Once the equipment is operational on the production line, a distributed sensor network continuously captures dynamic operating parameters such as winding temperature, rotor vibration amplitude, and three-phase current imbalance at millisecond frequencies. The model incorporates multidimensional stress-coupled boundary conditions to distinguish between normal wear and abnormal deterioration. Through continuous integration, the model transforms the complex external electromechanical-thermal multiphysics stress input into a quantitative output of lattice dislocations and insulating dielectric aging within the material.

[0024] S103. Based on the operating status data and combined with the damage measurement model, calculate the damage acceleration of the motor under the current operating conditions relative to the initial health baseline, convert the physical operating time into virtual damage equivalent for cumulative deduction, and update the remaining virtual life of each individual motor in real time.

[0025] This process involves calculating the damage acceleration under the current operating conditions based on collected operating status parameters and damage measurement results. The physical operating time is then converted into a virtual damage equivalent for cumulative deduction, updating the remaining virtual lifespan in real time. In actual production scenarios, equipment experiences extremely drastic load fluctuations. The stress change rate within a specific time window is extracted, and the fatigue damage accumulation theory from mechanical dynamics is introduced into the virtual lifespan calculation. Even if two machines operate for the same amount of time, the machine operating under severe conditions will generate a virtual damage equivalent far greater than the actual operating time, leading to rapid lifespan decay. Conversely, under light-load and stable operating conditions, lifespan depletion is slowed down. This nonlinear conversion mechanism accurately characterizes the degree of erosion of the component's physical background by varying operating conditions.

[0026] S104. When allocating production tasks, perform differentiated collaborative scheduling for motors based on the remaining virtual lifespan.

[0027] S105. When the remaining virtual life of the motor is detected to be lower than the preset safety threshold, a predictive maintenance work order is generated based on the initial damage status marked in the initial health baseline, and a spare parts procurement request is sent to the supply chain based on the predictive maintenance work order.

[0028] When the remaining virtual lifespan of a device is detected to be below a preset safety threshold, a predictive maintenance work order is generated based on the initial damage state marked in the baseline, and a spare parts procurement request is sent to the supply chain. Traditional reactive maintenance often only begins investigating the cause and searching for replacement parts after a component has completely broken or burned out, resulting in extremely long unplanned downtime. By setting a multi-level safety threshold system, once the calculated lifespan value touches the red line, not only will an alarm be automatically triggered, but the initial state feature library established at the time of the device's manufacture will also be retrieved. The system identifies the location of the device's initial minor defects, such as weak points in the insulation of specific tooth grooves, and, combined with current evolution trends, generates a maintenance guide containing clear troubleshooting steps and required tools. Furthermore, not only can maintenance personnel intervene in advance, but the supply chain procurement end can also simultaneously receive spare parts orders containing precise specifications and material improvement requirements.

[0029] Furthermore, the predictive maintenance work order not only includes simple alarm information, but is also a service implementation plan integrating repair processes, resource scheduling, and cost accounting. First, it automatically matches differentiated processing logic based on the type of damaged component: if it detects a significant decline in magnet performance, aging of copper wire insulation, or partial short circuit in the iron core, the system will automatically determine and propose a strategy based on the damage range and remaining virtual lifespan, such as partial replacement (e.g., partial re-insulation treatment, single magnet replacement) or complete component replacement. Simultaneously, it directly generates a standardized repair process operation guide and, combined with real-time inventory spare parts costs and labor hours, generates an accurate repair price list for customer confirmation. For cast structural components such as end caps and bases, the focus is on dimensional accuracy deviations and mechanical wear caused by long-term stress. Once tolerance thresholds are exceeded, the work order directly calls the warehouse management system to issue a replacement parts outbound instruction. For built-in sensors and other electronic components, the system primarily relies on electronic indicators such as signal deviation, stability, and lifespan to trigger replacement reminders before complete failure. The aforementioned maintenance work order mechanism enables predictive maintenance technology to be truly implemented and transformed into executable, standardized after-sales service products.

[0030] S106. When multiple individual motors in a motor group are found to have the same degradation characteristics, trace the innate and acquired characteristic data of the individual motors with degradation characteristics, and use cluster analysis algorithms to identify the common factor combinations that lead to the degradation characteristics.

[0031] In long-term operation, when multiple individuals within a group exhibit the same degradation characteristics, the innate and acquired data of these individuals are traced, and clustering analysis algorithms are used to identify the common factor combinations leading to the degradation. Individual failures may be accidental, but similar degradation in a group often points to deep-seated systemic defects. All cases of premature insulation aging or abnormal bearing wear are collected, and the corresponding equipment's microscopic physicochemical parameters and process stress data are extracted. Gene sequence alignment clustering algorithms from bioinformatics are transferred to industrial defect tracing. Density extrema are found in a multidimensional high-order feature space, and feature vectors from similar manufacturing batches, the same material suppliers, or those that have experienced similar abnormal processing stresses are clustered.

[0032] S107. Quality rectification strategies for motor components generated based on common factor combinations are fed back to the supply chain.

[0033] The process involves generating quality rectification strategies based on identified common factor combinations and feeding them back to the supply chain. After identifying the common pathogenic factors causing group failures, the final step in closed-loop management is to implement improvement measures at the source. Identified defect factors are mapped back to specific supply chain links. If the problem stems from material physicochemical indicators, rectification documents are generated to improve raw material inspection standards or change alloy proportions. If the problem is caused by manufacturing stress, process upgrade instructions are generated to optimize stamping die clearances or adjust the temperature curve of the weld heat-affected zone. Once the supply chain nodes receive the strategies, they can update the production line control programs online. This data feedback mechanism, which transcends enterprise boundaries, facilitates the transformation of manufacturing from unidirectional production to a self-evolving ecosystem, continuously pushing the manufacturing limits and reliability levels of the entire industrial chain.

[0034] Building upon this foundation, a responsibility tracing and two-way feedback mechanism based on lifespan decay patterns can be further constructed to achieve deep collaboration across enterprise boundaries. Specifically, when motor components fail prematurely or require replacement during their service life, maintenance personnel can directly retrieve the part's batch number, initial health baseline, design service life, and continuously stored historical service duration and operating load records by scanning a code on-site. By comparing the actual lifespan curves of components with standard theoretical curves in multiple dimensions, the main cause of failure can be inferred in reverse: If the background records show that the motor has been subjected to severe load fluctuations and abnormally high temperatures for a long time, resulting in serious wear or burnout before the actual service life is far less than the design service life, it will be judged as poor user operation. An optimization suggestion report containing historical violation operation records and equipment operation specifications will be automatically generated and fed back to the application end to guide it to improve its equipment usage habits and extend its lifespan. Conversely, if the records show that the equipment is operating under normal and stable conditions, but the actual lifespan of the components still declines sharply, it will be judged as an intrinsic material defect or manufacturing process problem. A detailed quality defect tracing report will then be generated and fed back to the supplier or manufacturer of the corresponding batch to determine the attribution of responsibility and promote substantive quality rectification in the upstream of the industrial chain.

[0035] In one embodiment, to further clarify how the data-driven intelligent fusion-based collaborative management method for the entire process of motor components, as described in this invention, transcends the boundaries of digital algorithms and forms a commercially viable closed-loop transformation of results in real industrial manufacturing, warehousing and logistics, client-side operation, and after-sales maintenance, this embodiment takes the full lifecycle management of a certain type of industrial high-performance permanent magnet synchronous motor as an example to detail the specific execution process of this invention in actual business flows. This embodiment not only covers the underlying damage measurement and clustering algorithms but also deeply integrates them into the entire physical process from supply chain warehousing to decommissioning traceability.

[0036] Regarding step S101, in the supply chain and manufacturing / assembly stages, this invention constructs a bottom-level physical tag traceability and warehousing collaboration system based on unique digital identity codes. When different types of key motor components are delivered from various suppliers to the manufacturing plant for warehousing, the system not only collects their microscopic physicochemical parameters but also deeply integrates these parameters with ERP and WMS systems, executing differentiated filing and management processes for components with different attributes. Specifically: (1) For core materials such as magnets, copper wires, and iron cores that directly determine the electromagnetic performance of motors, the inspection and tracking upon warehousing are particularly stringent. Taking magnets as an example, when purchasing and warehousing, the quality inspection terminal automatically collects the physicochemical indicators such as remanence, coercivity, and maximum energy product of the batch of magnets as inherent characteristic data through a gaussmeter and a fluxmeter. After the system determines that these indicators have reached the qualified procurement threshold, it will automatically generate a unique traceability barcode or RFID tag for each batch or even each core magnet, containing the supplier code, warehousing time, and physical property parameters. After entering the automated warehouse, the warehousing system reads the tag to achieve accurate allocation of storage locations and first-in-first-out control. When the assembly line calls for materials, the AGV trolley delivers the marked magnets to the workstation according to the outbound instruction. The assembly robot scans the tag to confirm the outbound delivery and uniquely binds the digital identity of the magnet to the serial number of the motor currently being assembled. At this point, the orthogonal coupling algorithm maps the inherent remanent magnetization parameters of the magnet to the acquired press-fit stress data applied during the rotor magnetization process, generating a unique initial health baseline for this specific motor. The same applies to copper wire and iron core. The incoming inspection of copper wire focuses on oxygen content, purity, and insulation varnish thickness, while the iron core focuses on the grain size of the silicon steel sheets and the degree of stamping burrs. These microscopic parameters are all rigorously attached to the digital twin model of the individual motor unit through a barcode scanning process for both incoming and outgoing inventory.

[0037] (2) For casting and machining parts such as end caps and bases, the collection of their microscopic physicochemical parameters focuses on the yield strength of the material, the distribution of casting pinholes, and the initial dimensional accuracy of machining (such as coaxiality and cylindricity errors). Upon warehousing, the data from the coordinate measuring machine is directly imported into the system to generate the inherent characteristic data of these structural parts, and a QR code containing dimensional tolerance records is also affixed. During the assembly stage, the system records the real-time torque of the tightening bolts as the acquired process stress data and integrates it into the initial health baseline of the motor.

[0038] (3) For built-in electronic components such as temperature sensors, vibration sensors and encoders, their inherent characteristic data are expressed as indicators such as factory-calibrated sensitivity, zero-point drift rate and theoretical electrical life. The warehousing and outbound processes also rely on barcode traceability to ensure that the performance baseline of each sensor is accurately recorded in the motor's initial state database.

[0039] Regarding steps S102 and S103, once the motor, with its detailed production history and initial health baseline, is delivered to the end customer and put into service, the collaborative management process enters the operation monitoring and lifespan assessment phase. At this point, the motor's built-in sensor network begins collecting dynamic operating status data such as ambient temperature, stator current amplitude, and rotor vibration frequency in milliseconds. The edge computing gateway transmits this data back to the cloud-based damage measurement model in real time. In this process, physical time is no longer the sole measure of lifespan. The system calculates damage acceleration based on real-time operating conditions, converting physical operating time into virtual damage equivalents. For example, if a client, in order to meet production targets, causes the motor to operate continuously for 10 hours under harsh conditions exceeding its rated load and with an ambient temperature as high as 45 degrees Celsius, the nonlinear conversion mechanism in the damage measurement model will sensitively capture the excessively high current peaks and thermal stress, calculating that these 10 hours of physical time are equivalent to 50 hours of virtual damage equivalent under normal stable operating conditions. This is then deducted from the motor's preset total design lifespan, and its remaining virtual lifespan is updated in real time.

[0040] Regarding step S104, in the production task allocation stage on the client side, the MES system performs differentiated collaborative scheduling by calling the motor's remaining virtual life interface. For motors with sufficient remaining virtual life and a good initial health baseline, the system prioritizes allocating heavy-load, long-cycle production tasks; while for motors whose remaining virtual life has been consumed by more than half and whose factory records show a slight initial coaxiality deviation in their end caps, the system schedules them to a light-load, intermittently running standby position, thereby maximizing the overall service life of the entire pump station equipment cluster without increasing additional investment.

[0041] Regarding step S105, this invention provides a predictive maintenance work order generation and spare parts supply chain linkage mechanism. When the cloud detects that the remaining virtual lifespan of a motor is lower than a preset safety threshold, the system no longer simply displays a simple alarm on the control screen as in the traditional mode, but automatically triggers a collaborative maintenance process: First, based on the motor's serial number, the system will trace back to the initial health baseline established in stage S101 and the inventory records of all key components. Combined with the output of the current damage measurement model, it will accurately locate the specific component type that is about to fail and directly generate a set of in-depth maintenance work orders that include standardized maintenance process procedures, required tooling fixtures, and maintenance pricing.

[0042] The specific response strategies are categorized as follows: If the system determines that the magnet's performance has irreversibly demagnetized due to prolonged high-temperature operation, or that the copper wire insulation layer is on the verge of breakdown, the work order will automatically assess the extent of the damage and propose a maintenance decision of "partial replacement" or "complete replacement." If it is partial demagnetization, the work order will generate a replacement process guide for a single magnet and, using the current labor rate and warehouse material unit price, automatically generate a repair price list including travel, labor, and spare parts costs, which will be sent to the customer's equipment manager. If there is a severe overall decrease in the insulation of the stator winding, the work order will recommend returning the entire machine to the factory or directly replacing the stator assembly. If it is determined that mechanical parts such as end covers and machine bases have experienced excessive wear in the bearing housing due to prolonged vibration, the work order will clearly indicate the bearing model and end cover material code that need to be replaced. If it is determined that electronic components such as sensors have reached their stability limits, the work order will instruct for plug-and-play quick replacement.

[0043] Simultaneously, upon generation, the predictive maintenance work order, through cross-enterprise penetration of the cloud platform, directly sends a spare parts procurement or outbound request to the warehouse management system on the supply chain side. Upon receiving the instruction, the warehouse system automatically verifies the corresponding batch and specifications of spare parts in inventory and pre-picks and packs them. Through barcode scanning for outbound processing, the logistics trajectory of the spare parts is updated in real-time to the maintenance work order. When the maintenance engineer arrives on-site, the required specialized parts have already been delivered. The engineer only needs to operate according to the disassembly and assembly animation customized based on the initial assembly data of the motor displayed on the system terminal, completely eliminating the risk of long-term downtime caused by blind disassembly and troubleshooting, and realizing the closed-loop implementation of predictive maintenance.

[0044] Regarding steps S106 and S107, this invention further elaborates on how to achieve cross-industry chain two-way responsibility traceability and rectification feedback using high-dimensional clustering algorithms during long-term, large-scale equipment operation. In actual business environments, the determination of responsibility for equipment damage is often the focus of disputes between suppliers and customers. This invention relies on an immutable data chain with full-process records to provide a scientific basis for responsibility determination. When multiple individual motors distributed in different workshops or even different customer sites are detected to exhibit the same degradation characteristics within a similar time period, the system immediately initiates the construction of a high-dimensional fault feature matrix and density peak clustering analysis. The algorithm automatically searches for common factor combinations that lead to this group degradation in multi-dimensional space. Once the common factors are identified, the system immediately triggers a two-way traceability determination mechanism. Maintenance engineers or back-end expert systems retrieve the full life cycle duration records and operating load records of these faulty motors by scanning codes, and compare and infer with the physicochemical parameters of the parts stored in stage S101: Client-side responsibility: If cluster analysis reveals common factors pointing to extreme ambient temperatures and frequent overload starts, and historical operating data clearly shows that these motors experiencing similar faults have been operating at the customer's site for extended periods under peak torque far exceeding the product manual's specifications, leading to an exponential increase in virtual damage equivalent and actual service life far below the design life, the system, through reverse inference, determines that this is not a problem with the material itself, but rather improper customer usage. Subsequently, the system automatically generates an "Equipment Operation Optimization Suggestions and Responsibility Determination Report," containing detailed historical violation operation curves, theoretical damage mechanisms, and correct operating procedures, and sends it directly to the client.

[0045] Supplier / Manufacturing Responsibility: If backend data proves that the client's operation is fully compliant with specifications and the working conditions are stable, but cluster analysis reveals that the common factors of these failed motors are highly concentrated in a specific batch of copper wire supplied by a specific supplier in a specific month and year, or in silicon steel sheets processed by a specific stamping machine within a specific period, the system can confirm that these prematurely failed parts all share the same numerical characteristics by calling the warehouse entry code and batch number. In this case, the system determines it to be a systemic quality event caused by intrinsic material defects or insufficient manufacturing processes.

[0046] Based on this assessment, the system immediately generates quality rectification strategies for motor components based on the common factor combination and accurately feeds these strategies back to the supply chain and manufacturing ends. For suppliers, the system automatically sends a "Raw Material Quality Rectification Notice," requiring them to improve the purity of subsequent batches of copper wire or refine the insulating varnish formula, and automatically raises the relevant thresholds in the incoming inspection standards. For the manufacturing end, the system issues instructions to the workshop MES system, requiring optimization of the stamping die clearance or reduction of the winding machine's pulling speed. This closed-loop feedback throughout the entire process ensures that every fault data point from the application end is directly transformed into feedback material that drives material upgrades and process iterations upstream in the supply chain. Ultimately, this builds a clearly defined, continuously evolving, and intelligently collaborative management ecosystem across the entire motor industry chain.

[0047] In one implementation, performing differentiated collaborative scheduling for motors based on their remaining virtual lifetime during production task allocation includes the following steps: The production tasks to be assigned are decomposed into multiple time-series load subtasks, and the basic virtual lifetime consumption of each time-series load subtask is evaluated using a damage measurement model. Each individual motor in the motor group is constructed as an independent bidding agent, and the dynamic lifetime bidding coefficient of each bidding agent is calculated based on the initial health baseline and the current remaining virtual lifetime of each individual motor. The basic virtual lifetime consumption is input into each bidding agent, and nonlinear weighting is performed in combination with the dynamic lifetime bidding coefficient to output the lifetime bidding cost of each motor unit for each time-series load subtask. Construct a global collaborative allocation function with the evolutionary objective of minimizing the total lifetime bidding cost of the motor group; The remaining virtual lifetime of each individual motor is used as the safety constraint boundary condition of the global collaborative allocation function; A multi-agent cooperative game algorithm is used to iteratively solve the global collaborative allocation function, generating an optimal task allocation matrix that maps and matches time-series load subtasks with individual motors. Based on the optimal task allocation matrix, differentiated collaborative scheduling is performed for the motors.

[0048] In this embodiment, modern industrial manufacturing faces ever-changing and complex order demands. Directly issuing macro-level production tasks can easily lead to unpredictable and severe load shocks on the underlying execution units. The macro-level production tasks to be assigned are sliced ​​along the time axis, extracting multiple time-series load sub-tasks with independent physical significance. Then, the continuous processing is deconstructed into discrete and continuous operation nodes such as transient start-up, high-frequency pulse pressurization, constant speed pressure holding, and uniform deceleration braking. Each operation node carries specific torque requirements and speed settings. Subsequently, the basic virtual lifetime consumption of each time-series load sub-task is evaluated one by one using the aforementioned damage measurement model. The peak current surges and mechanical oscillation frequency characteristics contained in the sub-tasks are extracted and injected into the calculation engine containing the accelerated aging model. Let the basic virtual lifetime consumption be... The time-series load function is The duration of the subtask is The nonlinear stress amplification factor is A local damage quantification integral is constructed for a single, extremely short time slice, and the calculation formula is expressed as follows: By using time-series calculus, the ambiguity of the forces acting on the equipment at different stages of operation was eliminated.

[0049] Conventional centralized control architectures often encounter computational bottlenecks and response delays when dealing with massive distributed devices. Therefore, each individual motor in the group can be upgraded to an independent bidding agent with autonomous decision-making capabilities. This endows each actuator with anthropomorphic self-preservation awareness, allowing each rotor to participate in task competition calculations based on its own health condition. Based on the initial health baseline value established in the pre-processing stage and the real-time updated remaining virtual lifetime, the dynamic lifetime bidding coefficient for each bidding agent is calculated. Devices in their prime without inherent defects have a strong willingness to bear loads, while devices entering their aging stage or with hidden insulation defects exhibit a strong risk-avoidance tendency. Let the dynamic lifetime bidding coefficient be... The initial health baseline value is The remaining virtual lifetime is The recession-sensitive penalty constant is An index-based valuation model is established to characterize the willingness to accept orders and the ability to withstand risks for equipment. The calculation formula is as follows: This maps the wear and tear of physical space to a price lever in virtual space, enabling each intelligent agent to output a differentiated willingness to bear the same production instruction, aligning with its current physical limits.

[0050] After establishing the individual valuation intentions of each execution unit, the actual damage that may result from the allocation action is further quantified. The basic virtual lifetime consumption is precisely fed into the data processing center of each bidding agent as an input signal. Within the agent's internal logic unit, the received external load cost and internal dynamic lifetime bidding coefficient are deeply nonlinearly weighted and fused to output the absolute lifetime bidding cost for each device for each time-series load subtask. A single load value cannot truly reflect the degree of damage to devices in different health conditions, just as the destructive force of the same heavy object pressing on an intact structure is vastly different from that of a micro-cracked structure. By introducing a geometrically multiplicative nonlinear amplification mechanism, the calculated bidding cost of a device in a critical state of near-scrap condition explodes when receiving a heavy load task, thus automatically disqualifying it from winning the bid in subsequent competitions. Let the lifetime bidding cost be... The basic virtual lifetime consumption is The dynamic lifespan pricing coefficient is The load degradation exacerbation index is A mathematical mapping relationship for individualized injury discount is constructed, and the calculation formula is expressed as follows: This mechanism rigorously cross-maps the objective difficulty of the task with the subjective capacity of the equipment, ensuring that the final bidding cost is a customized depreciation quote that matches the current fatigue state of each physical entity.

[0051] Having grasped the differentiated capabilities of all independent intelligent agents, a comprehensive approach from a macro-cluster perspective is necessary to prevent overprotection of individuals from causing overall production line stagnation. Therefore, a global collaborative allocation function is constructed with the evolutionary objective of minimizing the total lifecycle bidding cost of the motor group. Specifically, a large-scale evaluation function is designed, encompassing all sub-task nodes and all available hardware resources. This function must not only ensure that currently issued production orders are fully and completely accepted, but also ensure that the total physical losses at the entire workshop level are minimized. This transforms the original single-dimensional assessment of maximizing production capacity into a multi-dimensional green production assessment that pursues minimizing lifecycle costs. Let the global total cost be... The lifetime bidding cost of a single device is The task allocation Boolean matrix is ​​as follows: The cluster collaboration scale coefficient is Construct a global cost optimization equation that iterates through all matching possibilities. The calculation formula is expressed as follows: This allocation function constructs a vast virtual game field, forcing all agents to find the combination path with the least destructive impact on the entire hardware ecosystem while meeting the rigid demands of production. This fundamentally eliminates the persistent problem of industrial scheduling where local optima lead to global collapse, and maximizes the value of equipment throughout its entire lifecycle.

[0052] In the mathematical calculations aimed at minimizing the global evolution goal, it is easy to encounter extreme situations where devices in a marginal state are sacrificed in pursuit of an extreme value. To prevent algorithmic loss of control, the remaining virtual lifetime updated in real time by each execution unit is forcibly set as a safety constraint boundary condition of the global collaborative allocation function. Regardless of how much total cost can be saved globally by assigning a specific task to a critically endangered device, as long as the estimated consumption of that task reaches the current device's endurance limit, a penalty mechanism will be immediately triggered and the allocation will be rejected. Let the remaining virtual lifetime be... The lower limit of the safety constraint is Lifetime bidding cost is The maximum protection margin is To establish an insurmountable nonlinear defense line, the constraint condition formula is expressed as follows: After establishing a complete evolutionary objective and stringent constraints, a multi-agent cooperative game theory algorithm is employed to perform a deep iterative solution to the global collaborative allocation function. Within an extremely short computation cycle, hundreds or even thousands of agents representing different motors frequently exchange bidding information and compromise with each other in virtual space. Multiple iterative screenings driven by high computing power isolate inferior combinations, ultimately converging and generating an optimal task allocation matrix that precisely maps and matches time-series load subtasks to individual motors. Based on this matrix, differentiated collaborative scheduling instructions are issued to the underlying execution mechanism. Let the optimal task allocation matrix be... The algorithm's iterative convergence factor is The total global cost is The equilibrium state value of the game is The final output of the extreme value optimization is expressed by the formula as follows: Ultimately, without any additional hardware investment, the equipment aging curve of the entire workshop cluster can be smoothed out, achieving a perfect balance between production efficiency and capital depreciation.

[0053] In one implementation, the microscopic physicochemical parameters of key motor components are collected as innate characteristic data during the supply stage of the supply chain, and the process stress data of key motor components during the motor manufacturing and assembly stage of the supply chain are collected as acquired characteristic data. The innate characteristic data and the acquired characteristic data are mapped and correlated through an orthogonal coupling algorithm to generate an initial health baseline characterizing the initial damage state of each individual motor in the motor group, including the following steps: During the supply stage of the supply chain, the microscopic physicochemical parameters of key motor components are collected, and multidimensional feature extraction is performed on the microscopic physicochemical parameters to construct innate feature data that characterizes the inherent physical properties. In the motor manufacturing and assembly stage of the supply chain, process stress data of key motor components are collected, and time series analysis is performed on the process stress data to construct acquired characteristic data that characterizes the distribution of acquired stress. Principal component analysis is used to reduce the dimensionality and noise of both innate and acquired feature data to obtain a standardized input vector set. A two-dimensional orthogonal coupling matrix containing material sensitivity coefficients is constructed. The input vector set is input into the two-dimensional orthogonal coupling matrix for cross tensor calculation, and the cross penalty weights of each key motor component under different stress dimensions are output. The ideal health benchmark preset by the motor is weighted and corrected based on the cross-penalty weights of each key motor component to generate an initial health baseline that characterizes the initial damage state of the individual motor.

[0054] In this embodiment, starting from the supply chain stage, scanning electron microscopy and X-ray fluorescence spectrometry are used to collect in-depth microscopic physicochemical parameters such as grain boundary density of silicon steel sheets, impurity content of winding copper wires, and polymerization degree of insulating resin. After completing the basic data collection, kernel function mapping is used to extract multidimensional features from the complex microscopic physicochemical parameters, eliminate redundant environmental noise, and accurately identify the core factors that directly affect the fatigue strength and permeability of the equipment, thus constructing inherent characteristic data that characterizes the inherent physical properties of the components. Let the microscopic physicochemical parameter matrix be... The feature extraction operator is The feature dimension weight coefficient is The innate eigenvectors are A nonlinear feature extraction equation is established, and the calculation formula is expressed as follows: After the supply stage, components inevitably bear various mechanical and thermal loads when entering the actual assembly stage, leaving permanent marks on the materials. During the motor manufacturing and assembly stage, distributed fiber optic grating sensors and high-frequency torque meters are used to comprehensively collect process stress data on key motor components during stator stamping, winding, and rotor heat fitting processes. Because the manufacturing load exhibits extremely strong high-frequency pulse and nonlinear fluctuation characteristics, extreme value records alone cannot reflect the true cumulative damage. Using a mathematical tool combining Fourier transform and wavelet packet analysis, time-series analysis is performed on the complex process stress data, decoupling the time-domain signal into a frequency-domain energy spectrum containing frequency, amplitude, and duration, constructing acquired characteristic data characterizing the acquired stress distribution of components. Let the process stress data array be... The time window truncation function is The wavelet energy attenuation coefficient is The acquired feature vector is The time-frequency conversion analytical relationship is constructed, and the calculation formula is expressed as follows: .

[0055] Next, deep dimensionality reduction and denoising processing is needed for both innate and acquired feature data. Specifically, firstly, highly collinear and redundant mapping dimensions in both types of data are identified. High-frequency noise caused by sensor thermal drift or signal transmission attenuation is removed. The orthogonal projection axes with the largest variance contribution are extracted, projecting the original discrete data points with different dimensions into a unified low-dimensional space, thus obtaining a highly condensed and standardized input vector set. Let the standardized input vector set be... The principal component projection transformation matrix is The eigenvalue diagonal matrix is The joint extended feature matrix containing both innate and acquired eigenvectors is: A high-dimensional feature dimensionality reduction mapping equation is constructed, and the calculation formula is expressed as follows: The relationship between the damage resistance of equipment materials and the mechanical stress applied during the manufacturing stage is not a simple linear superposition. Therefore, it is necessary to construct a two-dimensional orthogonal coupling matrix with specific material sensitivity coefficients. The extracted standardized input vector set is directly fed into the two-dimensional orthogonal coupling matrix to perform high-order cross tensor calculations. If a batch of copper wire is found to have low purity and is subjected to high-frequency pulse tension during the wire embedding process, the tensor engine will instantly capture the fatal internal and external factor superposition resonance phenomenon and exponentially amplify the risk assessment, ultimately outputting the cross penalty weights of key motor components under different stress dimensions. Let the standardized input vector set be... The material sensitivity coefficient is The two-dimensional orthogonal coupling matrix is The cross-penalty weight is The tensor product penalty mapping logic is established, and the calculation formula is expressed as follows: .

[0056] Based on the cross-penalty weights of key motor components output from the upstream module, a weighted correction operation is performed on the ideal health benchmark preset before the equipment leaves the factory. First, the ideal full-score state is considered the reference plane. Then, based on the accumulated penalty weights of each component in dimensions such as material purity, stamping microcracks, and weak points in winding insulation, the reference plane is reduced, much like deducting credit points. For rotors equipped with micro-lattice dislocations and uneven stress during the heat-fitting process, the initial baseline value will be significantly lowered. Let the preset ideal health benchmark be... The health decline compensation factor is The cross-penalty weight is The initial health baseline value characterizing the initial damage state is A nonlinear loss reduction function is established, and the calculation formula is expressed as follows: After corrections, an initial digital image is finally generated that accurately reflects the actual factory condition of each physical device.

[0057] In one implementation, the input vector set is fed into a two-dimensional orthogonal coupling matrix for cross tensor calculation, and the cross penalty weights of each key motor component under different stress dimensions are output, including the following steps: The input vector set is divided into a material dimension subset and a process dimension subset; Extract the orthogonal basis vectors from the two-dimensional orthogonal coupling matrix, use the orthogonal basis vectors to perform spatial projection on the material dimension subset and the process dimension subset, and calculate the nonlinear coupling coefficient between the material dimension subset and the process dimension subset in the projection space; The nonlinear coupling coefficient is compared with a preset damage sensitivity threshold to filter out strongly coupled nodes that exceed the damage sensitivity threshold. The selected strongly coupled nodes are reverse-mapped to calculate the corresponding stress penalty value, and the stress penalty value is aggregated and output as the cross-penalty weight of each key motor component under different stress dimensions.

[0058] In this embodiment, the acquired high-dimensional hybrid feature data contains information on different physical properties, and direct global computation can easily mask the salience of local features. Therefore, a data decoupling operation is performed to divide the fused standardized input vector set into a material dimension subset and a process dimension subset. Elements with intrinsic material attribute labels are identified in the data processing architecture, such as feature components characterizing the resistance to magnetic domain wall movement in silicon steel sheets, and these are grouped into the material dimension. Simultaneously, elements with externally applied load labels, such as the tensile stress peak component during stator winding, are stripped to the process dimension. Let the standardized input vector set be... The material dimension subset is The process dimension subset is The two-dimensional mask assignment matrix is A data-directed distribution and dimensional isolation mechanism is constructed, and the calculation formula is as follows: as well as After separating the data dimensions, the orthogonal basis vectors implicit within the two-dimensional orthogonal coupling matrix are extracted. Then, using these extracted orthogonal basis vectors, a high-order spatial projection operation is performed on the material dimension subset and the process dimension subset. Within the orthogonal projection space, the originally isolated lattice dislocation features and the externally applied mechanical strain features exhibit strong spatial overlap. By calculating the degree of coincidence and phase angle difference between the two on the same-dimensional orthogonal coordinate axes, the nonlinear coupling coefficient between the material dimension subset and the process dimension subset in the projection space can be determined.

[0059] Next, the calculated massive nonlinear coupling coefficients are compared one by one with the preset damage sensitivity threshold. In the field of industrial product quality control, not all the accumulation of minor defects will lead to premature equipment failure. Only when the resonance intensity of material defects and processing stress exceeds the physical tolerance limit of the material will substantial damage occur. Therefore, we can traverse every intersecting feature point in the projection space. Once a nonlinear coupling coefficient at a certain point is found to exceed the set damage sensitivity threshold, it is immediately marked and selected as a strongly coupled node. Let the set of strongly coupled nodes be... The preset damage sensitivity threshold is The nonlinear step activation function is The nonlinear coupling coefficient is A discretized Boolean screening equation based on threshold judgment is constructed, and the calculation formula is expressed as follows: This completely shields weakly coupled noise that is insufficient to cause a serious quality crisis, leaving only the intersection of core risk factors that have a decisive impact on the lifespan of components.

[0060] For each selected strongly coupled node, a reverse mapping procedure is initiated, reversing the original orthogonal mapping path to precisely project the coupling strength value in the high-dimensional space back to the original physical stress dimension. During this reverse analysis, the stress penalty value corresponding to each node is calculated. A strongly coupled node resulting from a tiny material pore encountering stamping stress concentration is decoded into an extremely high mechanical stress penalty component. Finally, the stress penalty values ​​across all dimensions are weighted and aggregated to output the cross-penalty weights for each key motor component under different stress dimensions. Let the stress penalty value be... The reverse mapping analytic operator is The set of strongly coupled nodes is The cross-penalty weight is The aggregation mathematical relationship from the abstract feature space to the physical penalty weights is established, and the calculation formula is expressed as follows: and aggregation formulas The closed-loop mapping operation completely establishes a digital channel from the underlying microscopic physicochemical data to the upper-level macroscopic lifetime prediction. The output cross-penalty weight is no longer an empty mathematical concept, but truly reflects the additional physical wear and tear costs that the component must bear during its future service life.

[0061] In one implementation, spatial projection of the material dimension subset and the process dimension subset using orthogonal basis vectors, and calculation of the nonlinear coupling coefficient between the material dimension subset and the process dimension subset in the projection space, includes the following steps: Calculate the projected coordinates of the material dimension subset and the process dimension subset on the orthogonal basis vectors; A high-order covariance matrix is ​​constructed based on projected coordinates, and the principal eigenvalues ​​of the high-order covariance matrix are extracted by eigenvalue decomposition. Substitute the principal eigenvalues ​​into the preset Lorentz nonlinear activation function to calculate the initial coupling values ​​of the material dimension subset and the process dimension subset in the projection space. By using an ambient temperature compensation factor to correct the initial coupling value, the nonlinear coupling coefficients of the material dimension subset and the process dimension subset in the projection space are obtained.

[0062] In this embodiment, the orthogonal basis vectors constructed in the preceding steps are extracted, pulling the inherent material properties and post-manufacturing stresses, which were originally discrete and could not be directly compared, onto an absolutely level mathematical observation plane. In the multidimensional tensor engine, the projection length and direction of the material characteristic components and process strain components on each independent orthogonal basis axis are calculated. This projection is not a simple numerical multiplication, but rather involves an inner product operation involving physical dimension transformation. Let the material dimension projection coordinate matrix be... The material dimension subset is The orthogonal basis vectors are A coordinate mapping equation based on the inner product measure is constructed, and the calculation formula is expressed as follows: Through projection transformation, material defects originally hidden in microscopic lattice dislocations and mechanical tension caused by macroscopic stamping are uniformly measured as a set of mathematical coordinate points with absolute directions. This transformation eliminates the cognitive barriers caused by the inconsistencies in measurement standards for different physical fields, allowing for a quantitative comparison of microscopic cracks and violent vibrations on a completely equal scale.

[0063] After obtaining the precise projection coordinates, a high-order covariance matrix capable of accommodating a massive number of cross-variables is constructed based on the projection coordinate data set. Within the mathematical computation unit, a deep matrix multiplication and expectation extraction are performed between the material projection coordinates and the process projection coordinates to capture their joint variation trend in spatial distribution. Subsequently, a high-precision eigenvalue decomposition operation is performed on the constructed high-order covariance matrix. The principal eigenvalues ​​representing the highest energy concentration of the matrix are extracted. In a physical sense, the principal eigenvalues ​​are directly equivalent to the maximum probability amplitude of a fatal overlap between microscopic material weakness points and macroscopic processing stress points. Let the high-order covariance matrix be... The material dimension projection coordinate matrix is The process dimension projection coordinate matrix is The extracted principal eigenvalues ​​are A model for extracting covariance and eigenvalues ​​to characterize multidimensional dispersion is established, and the calculation formula is expressed as follows: And the formula for solving the maximum eigenvalue. After extracting the principal eigenvalues ​​representing the maximum deterioration resonance trend, these eigenvalues ​​are substituted into a pre-defined Lorentz nonlinear activation function for evolution. In real motor components, when subjected to superimposed dual stresses, the damage explodes exponentially once a certain safety threshold is exceeded. The bell-shaped curve of the Lorentz function perfectly matches this physical evolution. When the principal eigenvalues ​​approach the central pole of the function, the output value exhibits a sharp nonlinear amplification. Let the initial coupling value be... The principal characteristic roots are The preset Lorentz linewidth parameter is: The Lorentz center offset is A nonlinear physical resonance excitation equation is constructed, and the calculation formula is expressed as follows: This allows it to be exceptionally sensitive in detecting subtle combinations of hidden dangers that are masked by normal daily operation. For example, if a tiny silicon steel sheet has excessive carbon content, it will be instantly amplified to an extremely high initial coupling risk value under the action of the activation function once it encounters cutting vibrations at a specific frequency.

[0064] The complex and variable thermal environment of industrial sites has a profound impact on the fatigue limit of materials. High temperatures not only accelerate the aging of insulating media but also significantly reduce the yield strength of metal lattices. Real-time thermodynamic data transmitted from external monitoring equipment is extracted, and an environmental temperature compensation factor is used to comprehensively correct the initial coupling value calculated in the previous step. When the equipment is in non-standard operating temperature ranges such as extreme cold or heat, the temperature compensation mechanism adaptively adjusts the pre-calculated risk assessment. During continuous high-temperature operation, even the smallest initial physical coupling hazards can be significantly amplified by thermal effects. Let the environmental temperature compensation factor be... The initial coupling value is The nonlinear coupling coefficient is A nonlinear compensation correction formula incorporating external thermal field effects is established, and the calculation formula is expressed as follows: After final modification of thermodynamic variables, the resulting nonlinear coupling coefficient is a warning scale that dynamically changes with fluctuations in the workshop ambient temperature.

[0065] In one implementation, the damage acceleration of the motor relative to the initial healthy baseline under the current operating condition is calculated based on the operating status data and combined with the damage measurement model. The physical operating time is converted into virtual damage equivalent and cumulatively deducted to update the remaining virtual life of each individual motor in real time. The steps include the following: Ambient temperature, stator current amplitude, and rotor vibration frequency are extracted from the operating status data as dynamic input parameters for the damage measurement model; The initial damage state in the initial healthy baseline is transformed into the initial boundary conditions of the accelerated aging model. Run the damage measurement model and calculate the transient damage rate of the motor under the current operating conditions based on dynamic input parameters and initial boundary conditions; The damage acceleration under the current working condition is obtained by time integral calculation of the transient damage rate, and the damage acceleration is multiplied by the physical running time to obtain the virtual damage equivalent. The virtual damage equivalent is cumulatively deducted from the preset total design life of the motor unit to update the remaining virtual life of the motor unit in real time.

[0066] In this embodiment, when the electric motor operates in a complex and ever-changing industrial environment, the drastic fluctuations in external load and environment directly accelerate the fatigue degradation process within the materials. The ambient temperature (thermodynamic level), stator current amplitude (electromagnetic level), and rotor vibration frequency (dynamic level) are extracted from the operating data. These three core parameters respectively map the most direct driving forces of insulation thermal aging, winding electrodynamic impact, and mechanical bearing wear. Let the ambient temperature be... The stator current amplitude is The rotor vibration frequency is The environmental heat effect weighting factor is The current impact weighting factor is The oscillation fatigue weighting factor is The dynamic input parameters are A linear fusion and normalization extraction matrix for multiphysics monitoring parameters is established, and the calculation formula is expressed as follows: This allows external sensing signals to be purified into a set of structured input tensors that can directly reflect transient severe operating conditions.

[0067] Next, the initial health baseline value characterizing the initial damage state is extracted. Combining this with the microscopic lattice stress relaxation effect, its mathematical properties are reshaped into the first initial value for continuous integral operations. Let the initial health baseline value characterizing the initial damage state be... The lattice relaxation coefficient is The initial boundary conditions are A mapping assignment formula reflecting the baseline fatigue strength is constructed, and the calculation formula is expressed as follows: Then, the damage measurement model in the underlying architecture is run. Based on the instantaneous thermal radiation and mechanical torque intensity experienced by the equipment, combined with the equipment's initial compressive strength, the instantaneous rate at which the material's microstructure is destroyed within this extremely short time slice is calculated. Let the dynamic input parameters be... The initial boundary conditions are The material damage sensitivity gain constant is The transient damage rate is A logarithmic mapping relationship representing the severity of immediate degradation is established, and the calculation formula is expressed as follows: After obtaining the transient damage rate, it needs to be converted into a fatigue scale that can be directly measured at the macroscopic level. Specifically, the discrete rate scatter plot can be fitted into a continuous smooth curve and the area can be calculated to obtain the comprehensive deterioration trend coefficient under the current operating condition. Then, the abstract deterioration trend coefficient is multiplied by the actual operating time experienced by the equipment, transforming the invisible microscopic material fracture into a visualized virtual loss index. Let the transient damage rate be... Damage acceleration is The physical runtime is The virtual damage equivalent is Establish the integration rule and equivalent conversion equation for time series data. The integration formula is expressed as follows: The multiplication conversion formula is expressed as follows This allows for a highly fair quantification of the actual damage caused by various extreme operating conditions. The physical destructive force generated in one hour of overload operation will be amplified geometrically through the formula, potentially equivalent to the wear and tear of tens of hours under stable operating conditions.

[0068] Finally, the ideal maximum lifespan value set at the initial design and development stage of the equipment is retrieved. With each start, stop, and frequency conversion speed adjustment of the motor, the virtual computing center continuously receives the loss equivalent data generated in the previous stage and performs subtraction operations to eliminate the imaginary fraction that has been actually consumed, thus displaying the remaining healthy margin of the equipment on the display terminal. Let the virtual damage equivalent be... The preset total design life is The remaining virtual lifetime is An irreversible, unidirectionally decreasing lifecycle function is constructed, and its calculation formula is expressed as follows: .

[0069] In one embodiment, the damage acceleration under the current operating condition is obtained by time integration of the transient damage rate, and the virtual damage equivalent is obtained by multiplying the damage acceleration by the physical running time, including the following steps: The Runge-Kutta algorithm was used to calculate the transient damage rate by time integration, and preliminary integration results were obtained. The initial integration result is smoothed by using the Kalman filter algorithm, and the smoothed initial integration result is defined as the damage acceleration under the current working condition. The basic damage equivalent is calculated by multiplying the damage acceleration by the corresponding physical running time. The basic damage equivalent is then corrected using a nonlinear correction factor based on the fluctuation of the real-time operating condition spectrum to calculate the virtual damage equivalent.

[0070] In this embodiment, the degradation rate of electromechanical equipment under complex operating conditions often manifests as a discrete data sequence containing high-frequency abrupt changes. Using a simple Euler's ordinary differential law for linear accumulation will cause the estimated life trajectory to deviate from the true physical state due to the continuous amplification of truncation errors. Therefore, four different detection nodes can be set within each small time interval to collect the degradation slope at the starting point, two midpoints, and the ending point. These four slopes represent the instantaneous evolutionary tendency of microstructural degradation at different times. Subsequently, specific weights are assigned to these four slopes, and a high-order weighted average is performed. Let the transient damage rate be... The preliminary integration result is The integral value at the previous moment was The tiny time step is The slope variables of the four detection nodes are respectively , , , A high-precision numerical integration derivation formula is constructed, and the calculation formula is expressed as follows: Numerical integration overcomes the computational divergence problem caused by fluctuations in external operating conditions, fitting discrete failure rates into smooth curves that characterize the cumulative fatigue trend of materials, and outputting preliminary integration results with extremely high mathematical fidelity.

[0071] After obtaining the initial cumulative damage trend, the data sequence is usually mixed with high-frequency spurious fluctuations caused by thermal drift of the underlying sensor network and electromagnetic interference of the transmission link. Directly using integral data with random spikes for lifetime calculation can lead to misjudgments of the equipment's health baseline by the macro-scheduling center. Therefore, a recursive observation and state update mechanism with dynamic correction capabilities can be constructed. Based on the physical inertia exhibited by mechanical components at historical moments, the theoretical degradation baseline value of the current time slice is extrapolated and estimated, and this theoretical baseline value is compared with the covariance of the input initial integral result. The computational logic calculates the proportional relationship between external observation noise and internal model uncertainty, dynamically generating gain coefficients to adjust the trust weights. Let the initial integral result be... The estimated value of the previous state is The dynamic Kalman gain is Damage acceleration is The optimal state estimation formula for integrating multi-source trust assessment is established, and the calculation formula is expressed as follows: The recursive filtering operation removes interference signals caused by accidental external vibrations, preserving the pure trend reflecting the intrinsic lattice collapse of the material. Finally, the cleaned data is established as the damage acceleration under the current operating conditions.

[0072] After obtaining the smoothed and denoised attenuation acceleration index, it needs to be mapped into an intuitive wear scale that can be directly read by the macro asset management center. First, linear spatiotemporal conversion is performed, multiplying and coupling the instantaneous destructive force in the digital space with the actual operating time experienced by the equipment in the real workshop to preliminarily calculate the basic damage consumption of the equipment. However, the real industrial manufacturing environment has phenomena such as heavy-load start-stop and sudden overload, and the fatigue superposition damage effect triggered by alternating stress exceeds the scope of single linear time accumulation. Therefore, it is possible to mine the fluctuation characteristics of operating status data within a very short time window, extract the variance distribution of load stress amplitude at high frequency, and construct a nonlinear correction factor that can sense the degree of external extreme impact. Let the damage acceleration be... The physical runtime is The basic damage equivalent is The nonlinear correction factor based on the fluctuation of the operating status data is: The virtual damage equivalent is An evolution equation incorporating linear cross-limit conversion and nonlinear operating condition penalty amplification is established, and the calculation formula is expressed as follows: and the corrected conversion formula .

[0073] In one implementation, identifying common factor combinations that lead to the decline characteristics using a clustering analysis algorithm includes the following steps: The inherent and acquired characteristic data of the motor group exhibiting degradation characteristics are extracted, and the inherent and acquired characteristic data are fused to construct a high-dimensional fault characteristic matrix; The density peak clustering algorithm is used to iteratively cluster the high-dimensional fault feature matrix to identify multiple local high-density clusters in the high-dimensional fault feature matrix; Calculate the intra-class similarity and inter-class dissimilarity of each local high-density cluster; Based on intra-class similarity and inter-class dissimilarity, discrete clusters with intra-class similarity below a preset similarity threshold are removed, while core clusters are retained; Extract the combination of feature elements that appear most frequently in the core cluster as the common factor combination that leads to the decay feature.

[0074] In this embodiment, for a group of controlled electromechanical equipment exhibiting the same failure behavior, the inherent characteristic data pre-recorded during the manufacturing stage of the equipment showing degradation characteristics is retrieved. This data details microscopic physicochemical parameters such as the grain size of silicon steel sheets and the oxygen content of copper wires. Simultaneously, process stress data accumulated during the manufacturing, assembly, and subsequent service stages of the aforementioned equipment is extracted, covering acquired parameters such as stamping deformation tensor and operating thermal load indices. A data fusion channel spanning physical characteristics is established, performing high-order splicing and alignment of inherent microscopic factors with completely different dimensions and physical meanings with acquired macroscopic loads. Let the inherent characteristic vector be... The acquired feature vector is The heterogeneous data fusion weight allocation operator is The high-dimensional fault feature matrix is Establish the formula for fusion splicing and dimensional expansion, and the calculation formula is expressed as follows: The matrix construction operation described above will structurally aggregate various potential pathogenic factors that are discretely distributed across the upstream and downstream of the supply chain and at the application end, forming a tensor map containing multi-physics information.

[0075] After fusing and reconstructing multi-source heterogeneous fault data, it is necessary to locate the highly clustered local spatial regions in the multidimensional matrix. Specifically, within the abstract feature vector space, there is no need to rely on preset classification labels; instead, all fused feature data points are automatically optimized and aggregated based on their relative positions in the coordinate system. The calculation process uses each feature data point as a scanning reference, measuring the number of adjacent nodes within a specific spatial radius to define the local information density of that spatial location; simultaneously, the spatial distance between each high-density node and other higher-density peak nodes is calculated. When a specific batch of material defects overlaps with a specific process peak, causing similar failures, the corresponding features will exhibit a dense distribution in the high-dimensional space. Let the local information density be... The Euclidean distance between high-dimensional feature data points is The cluster cutoff distance is The high-dimensional fault feature matrix is A nonlinear spatial density measurement function is established, and the calculation formula is expressed as follows: Clustering algorithms iteratively calculate and traverse all nodes, automatically identifying and delineating multiple spatially tightly clustered local high-density clusters, thus achieving precise spatial location of pathogenic factor combinations.

[0076] After identifying high-density feature clusters suspected of causing population failure, a quantitative verification of the cluster's internal compactness and external independence is performed. Specifically, for each independently partitioned local high-density cluster, the computational logic first performs a centripetal scan, calculating the spatial variance of the distance between edge feature nodes and core nodes within the cluster. This measures the tightness of clustering within the cluster, i.e., intra-class similarity; a higher value indicates a more generalized pathogenic logic. Subsequently, an exclusivity measure is performed, calculating the absolute spatial span between the current cluster core coordinates and other cluster cores, assessing the essential differences in pathogenic mechanisms between different feature combinations, i.e., inter-class dissimilarity. Let the intra-class similarity be... The dissimilarity between classes is The set of local cluster feature points is The coordinates of the cluster center are The center distance between different clusters is A mathematical model for jointly measuring spatial dispersion and exclusivity is constructed, and the calculation formula is expressed as follows: and the formula for difference The two-way spatial measurement mechanism quantifies the cohesion and exclusivity of feature clusters, providing a quantitative basis for eliminating invalid feature clusters caused by random sensor drift.

[0077] Based on the bidirectional verification results of cluster purity and specificity, a noise reduction and invalid cluster removal procedure is performed in the digital feature space. In real operating environments, individual equipment damage often occurs due to occasional operational deviations or transient power grid fluctuations. Such damage also forms scattered micro-clusters in the feature space, but it does not have universal quality rectification value. Based on the bidirectional measurement values ​​calculated and output by the pre-processing steps, a mathematical cleaning defense line based on threshold comparison is established. The extracted intra-class similarity values ​​are compared with the lower similarity threshold set at the bottom layer. If the internal element arrangement of a local cluster is too loose, that is, the intra-class similarity fails to reach the lower threshold, the cluster is determined to be a discrete cluster caused by random factors, and it is removed from the high-dimensional matrix. Let the effective core cluster determination label be denoted as . Intra-class similarity is The dissimilarity between classes is The preset similarity threshold is Establish a filtering decision equation for coupling exclusivity, and the calculation formula is expressed as follows: The cleaning process effectively removed interfering noise artifacts while preserving the dense core cluster.

[0078] After removing random noise and retaining pure core clusters, the final feature decoding stage for the causes of equipment cluster degradation begins. A deep scan is performed on each retained core cluster, conducting high-frequency exhaustive enumeration and cross-comparison of the massive number of feature elements contained within. Through decoupling and projection of the previous dimension, each scan can extract clear physical labels. Specific factor pairs exhibiting high co-occurrence frequency in the current core cluster are calculated, such as the combination of high carbon content in a specific batch of steel and the lower limit of a specific stamping pressure of a certain piece of equipment. Let the frequency of occurrence of feature element combinations be... The Boolean matrix matching specific feature elements is The total number of feature samples in the core cluster is The common pathogenic factors are: A factor purification equation based on probability density is constructed, and the calculation formula is expressed as follows: Subsequently, combinations of elements with a frequency higher than a set limit are extracted and output as pathogenic common factor combinations. This decoding operation transforms the common causes of decline hidden in complex manufacturing processes into specific parameter combinations.

[0079] In one embodiment, iteratively clustering a high-dimensional fault feature matrix using a density peak clustering algorithm to identify multiple local high-density clusters in the high-dimensional fault feature matrix includes the following steps: Extract the local neighborhood set of each data point in the high-dimensional fault feature matrix, and dynamically calculate the local feature weight matrix of each data point based on the variance distribution of each feature dimension in the local neighborhood set; The local feature weight matrix is ​​used to perform anisotropic space mapping on the data points in the high-dimensional fault feature matrix, and the higher-order cutoff distance of each data point is calculated in the anisotropic space to obtain the anisotropic local density of each data point. For each data point, search for the target neighbor node in the local neighborhood set whose anisotropic local density is greater than that of each data point and whose anisotropic spatial distance is the smallest, and establish a density gradient directed edge from each data point to the target neighbor node. Traverse all data points in the high-dimensional fault feature matrix and connect all density gradient directed edges to construct a density gradient directed forest that represents the clustering trend of the data. The root node in the density gradient directed forest with an in-degree greater than zero and an out-degree of zero is identified as the density peak. By performing reverse tracing along the density gradient directed edges in the density gradient directed forest, all data points converging to the same density peak are divided into the same cluster set, and the aggregated cluster sets are output as multiple local high-density clusters in the high-dimensional fault feature matrix.

[0080] In this embodiment, for each independent data point within the high-dimensional fault feature matrix, a radial scan is performed outward using a preset Euclidean space radius as the detection benchmark to extract a local neighborhood set surrounding the central node. Subsequently, a precise high-frequency statistical analysis is conducted within this set for each independent dimension constituting the feature vector, calculating and extracting the variance distribution of each feature dimension. A larger variance indicates that the local region contains richer pathogenic differences. Based on the extracted variance distribution, a local feature weight matrix specific to the central node is dynamically constructed. Let the high-dimensional fault feature matrix be... The local neighborhood set is The variance distribution vector is The local feature weight matrix is An adaptive weight allocation equation based on diagonalized variance is constructed, and the calculation formula is expressed as follows: Then, using the generated local feature weight matrix, a depth-based anisotropic space mapping is performed on each original data point in the high-dimensional fault feature matrix. Mathematically, this process manifests as a nonlinear stretching and compression of specific feature coordinate axes, making nodes that appear sparse in Euclidean space but are arranged along key degradation dimensions more compact after mapping. Subsequently, the higher-order cutoff distance of each data point is calculated in the newly constructed anisotropic space. Based on a high-dimensional relative metric, the anisotropic local density of each data point in the distorted space is accurately obtained. Let the local feature weight matrix be... The high-dimensional fault feature matrix is The mapped spatial coordinates are The higher-order cutoff distance is The anisotropic local density is Establish a formula for spatial nonlinear twisted projection. and density measurement models The spatial reconstruction of depth completely releases the deep physical correlations that are masked by linear global distance, allowing the high-order resonance tendency generated by tiny silicon steel sheet defects when encountering specific stamping stresses to stand out in the deformed feature space with extremely high local density, greatly enhancing the signal-to-noise ratio and detectability of weak failure signals.

[0081] Next, a short-range radar scanning mechanism is initiated for each independent data point in the virtual digital space, conducting a rigorous conditional search within a pre-defined local neighborhood set. Scan targets are strictly limited to those neighboring nodes that not only surpass the current node in anisotropic local density but also have the closest anisotropic spatial distance to the current node. After locking onto a target, a directed edge with density gradient is constructed, precisely pointing from the current low-density data point to the high-density target neighbor node. Let the anisotropic local density be... The spatial distance is The target neighbor node is The current node index is The directed edge targeting locking logic is established, and the calculation formula is expressed as follows: This forcibly connects the originally scattered individual failure cases according to the logic of potential physical deterioration, outlining the precise path of the evolution from microscopic hidden dangers to macroscopic failures.

[0082] Next, all data points contained in the high-dimensional fault feature matrix are traversed, and the density gradient directed edges generated in the previous steps are seamlessly connected and merged. A massive number of unidirectional vector threads continuously converge and merge in the mathematical space, ultimately constructing a density gradient directed forest that can comprehensively represent the spontaneous clustering trend of the data. In this virtual topological forest, probes examine the node state of each tree branch one by one, and the absolute root node that gathers multiple lower-level threads (i.e., in-degree greater than zero) and cannot find a higher-density destination (i.e., out-degree zero) is taken as the density peak of that local region. Let the density gradient directed forest matrix be... The in-degree vector of the node is The node's out-degree vector is The set of density peak nodes is An extremum point extraction formula based on graph theory topological features is constructed, and the calculation formula is expressed as follows: After the core array is successfully anchored, the established absolute density peaks serve as the starting point for the search. A highly rigorous reverse tracing is then conducted along the directed edges of the density gradient previously laid out in the directed forest. The probe continuously probes downwards in the opposite direction of the directed edges, retrieving all edge data points that directly or indirectly converge to the same specific density peak, and forcibly assigning them to the same exclusive cluster set. Finally, the resulting independent cluster sets are packaged and output as multiple locally high-density clusters existing within the high-dimensional fault feature matrix.

[0083] The present invention also discloses a collaborative management system for the entire process of motor components based on digital intelligence fusion, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the collaborative management method for the entire process of motor components based on digital intelligence fusion as described above.

[0084] The processor can be a central processing unit (CPU). Of course, depending on the actual use, it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc., and this application does not limit it.

[0085] The memory can be an internal storage unit of a computer device, such as a hard disk or RAM, or an external storage device, such as a plug-in hard disk, smart memory card (SMC), secure digital card (SD), or flash memory card (FC) provided on the computer device. Furthermore, the memory can be a combination of internal storage units and external storage devices of a computer device. The memory is used to store computer programs and other programs and data required by the computer device. The memory can also be used to temporarily store data that has been output or will be output. This application does not limit this.

[0086] The present invention also discloses a computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to be configured to perform the digital-intelligence fusion-based full-process collaborative management method for motor components described in any of the above embodiments.

[0087] The computer program can be stored in a machine-readable medium. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or certain middleware. The machine-readable medium includes any entity or device capable of carrying computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the machine-readable medium includes, but is not limited to, the above-mentioned components.

[0088] The method for collaborative management of motor components based on digital intelligence fusion described in the above embodiments is stored in the computer-readable storage medium and loaded and executed on the processor to facilitate the storage and application of the above method.

[0089] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of protection of this application is limited to these examples; within the framework of this application, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of different aspects of one or more embodiments of this application as described above, which are not provided in detail for the sake of brevity.

[0090] One or more embodiments in this application are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of this application. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments in this application should be included within the protection scope of this application.

Claims

1. A method for collaborative management of motor components throughout the entire process based on data and intelligence fusion, characterized in that, Includes the following steps: Microscopic physicochemical parameters of key motor components are collected as innate characteristic data during the supply stage of the supply chain, and process stress data of key motor components are collected as acquired characteristic data during the motor manufacturing and assembly stage of the supply chain. The innate characteristic data and the acquired characteristic data are mapped and associated through an orthogonal coupling algorithm to generate an initial health baseline that characterizes the initial damage state of each individual motor in the motor group. A damage measurement model incorporating an accelerated aging model is established based on the initial health baseline, and operational status data is collected in real time during the service phase of the motor group. Based on the operating status data and combined with the damage measurement model, the damage acceleration of the motor under the current operating conditions relative to the initial health baseline is calculated. The physical operating time is converted into virtual damage equivalent and accumulated and deducted to update the remaining virtual life of each individual motor in real time. When allocating production tasks, differentiated collaborative scheduling is performed for motors based on their remaining virtual lifespan; When the remaining virtual life of the motor is detected to be lower than the preset safety threshold, a predictive maintenance work order is generated based on the initial damage status marked in the initial health baseline, and a spare parts procurement request is sent to the supply chain based on the predictive maintenance work order. When multiple individual motors in a motor group are found to exhibit the same degradation characteristics, the innate and acquired characteristic data of the individual motors exhibiting degradation characteristics are traced back, and cluster analysis algorithms are used to identify the common factor combinations that lead to the degradation characteristics. Quality rectification strategies for motor components generated based on combinations of common factors are fed back to the supply chain.

2. The method for collaborative management of motor components throughout the entire process based on data and intelligence fusion as described in claim 1, characterized in that, The process of collecting microscopic physicochemical parameters of key motor components as innate characteristic data during the supply chain stage and collecting process stress data of key motor components as acquired characteristic data during the motor manufacturing and assembly stage, and mapping and associating the innate and acquired characteristic data through an orthogonal coupling algorithm to generate an initial health baseline characterizing the initial damage state of each individual motor in the motor group includes the following steps: During the supply stage of the supply chain, the microscopic physicochemical parameters of key motor components are collected, and multidimensional feature extraction is performed on the microscopic physicochemical parameters to construct innate feature data that characterizes the inherent physical properties. In the motor manufacturing and assembly stage of the supply chain, process stress data of key motor components are collected, and time series analysis is performed on the process stress data to construct acquired characteristic data that characterizes the distribution of acquired stress. Principal component analysis is used to reduce the dimensionality and noise of both innate and acquired feature data to obtain a standardized input vector set. A two-dimensional orthogonal coupling matrix containing material sensitivity coefficients is constructed. The input vector set is input into the two-dimensional orthogonal coupling matrix for cross tensor calculation, and the cross penalty weights of each key motor component under different stress dimensions are output. The ideal health benchmark preset by the motor is weighted and corrected based on the cross-penalty weights of each key motor component to generate an initial health baseline that characterizes the initial damage state of the individual motor.

3. The method for collaborative management of motor components throughout the entire process based on data and intelligence fusion according to claim 2, characterized in that, The step of inputting the input vector set into a two-dimensional orthogonal coupling matrix to perform cross tensor calculation and outputting the cross penalty weights of each key motor component under different stress dimensions includes the following steps: The input vector set is divided into a material dimension subset and a process dimension subset; Extract the orthogonal basis vectors from the two-dimensional orthogonal coupling matrix, use the orthogonal basis vectors to perform spatial projection on the material dimension subset and the process dimension subset, and calculate the nonlinear coupling coefficient between the material dimension subset and the process dimension subset in the projection space; The nonlinear coupling coefficient is compared with a preset damage sensitivity threshold to filter out strongly coupled nodes that exceed the damage sensitivity threshold. The selected strongly coupled nodes are reverse-mapped to calculate the corresponding stress penalty value, and the stress penalty value is aggregated and output as the cross-penalty weight of each key motor component under different stress dimensions.

4. The method for collaborative management of motor components throughout the entire process based on data and intelligence fusion according to claim 3, characterized in that, The step of spatially projecting the material dimension subset and the process dimension subset using orthogonal basis vectors and calculating the nonlinear coupling coefficient between the material dimension subset and the process dimension subset in the projected space includes the following steps: Calculate the projected coordinates of the material dimension subset and the process dimension subset on the orthogonal basis vectors; A high-order covariance matrix is ​​constructed based on projected coordinates, and the principal eigenvalues ​​of the high-order covariance matrix are extracted by eigenvalue decomposition. Substitute the principal eigenvalues ​​into the preset Lorentz nonlinear activation function to calculate the initial coupling values ​​of the material dimension subset and the process dimension subset in the projection space. By using an ambient temperature compensation factor to correct the initial coupling value, the nonlinear coupling coefficients of the material dimension subset and the process dimension subset in the projection space are obtained.

5. The method for collaborative management of motor components throughout the entire process based on data and intelligence fusion according to claim 1, characterized in that, The process of calculating the damage acceleration of the motor relative to the initial healthy baseline under the current operating conditions based on operating status data and combined with a damage measurement model, converting physical operating time into virtual damage equivalent for cumulative deduction, and updating the remaining virtual life of each individual motor in real time includes the following steps: Ambient temperature, stator current amplitude, and rotor vibration frequency are extracted from the operating status data as dynamic input parameters for the damage measurement model; The initial damage state in the initial healthy baseline is transformed into the initial boundary conditions of the accelerated aging model. Run the damage measurement model and calculate the transient damage rate of the motor under the current operating conditions based on dynamic input parameters and initial boundary conditions; The damage acceleration under the current working condition is obtained by time integral calculation of the transient damage rate, and the virtual damage equivalent is obtained by multiplying the damage acceleration by the physical running time. The virtual damage equivalent is cumulatively deducted from the preset total design life of the motor unit to update the remaining virtual life of the motor unit in real time.

6. The method for collaborative management of motor components throughout the entire process based on digital intelligence fusion according to claim 5, characterized in that, The step of calculating the damage acceleration under the current operating condition by integrating the transient damage rate over time, and then multiplying the damage acceleration by the physical running time to obtain the virtual damage equivalent includes the following steps: The Runge-Kutta algorithm was used to calculate the transient damage rate by time integration, and preliminary integration results were obtained. The initial integration result is smoothed by using the Kalman filter algorithm, and the smoothed initial integration result is defined as the damage acceleration under the current working condition. The basic damage equivalent is calculated by multiplying the damage acceleration by the corresponding physical running time. The basic damage equivalent is then corrected using a nonlinear correction factor based on fluctuations in the running state data to calculate the virtual damage equivalent.

7. The method for collaborative management of motor components throughout the entire process based on digital intelligence fusion according to claim 1, characterized in that, The process of identifying common factor combinations that lead to the decline characteristics using cluster analysis algorithms includes the following steps: The inherent and acquired characteristic data of the motor group exhibiting degradation characteristics are extracted, and the inherent and acquired characteristic data are fused to construct a high-dimensional fault characteristic matrix; The density peak clustering algorithm is used to iteratively cluster the high-dimensional fault feature matrix to identify multiple local high-density clusters in the high-dimensional fault feature matrix; Calculate the intra-class similarity and inter-class dissimilarity of each local high-density cluster; Based on intra-class similarity and inter-class dissimilarity, discrete clusters with intra-class similarity below a preset similarity threshold are removed, while core clusters are retained; Extract the combination of feature elements that appear most frequently in the core cluster as the common factor combination that leads to the decay feature.

8. The method for collaborative management of motor components throughout the entire process based on digital intelligence fusion according to claim 7, characterized in that, The step of iteratively clustering the high-dimensional fault feature matrix using the density peak clustering algorithm to identify multiple local high-density clusters in the high-dimensional fault feature matrix includes the following steps: Extract the local neighborhood set of each data point in the high-dimensional fault feature matrix, and dynamically calculate the local feature weight matrix of each data point based on the variance distribution of each feature dimension in the local neighborhood set; The local feature weight matrix is ​​used to perform anisotropic space mapping on the data points in the high-dimensional fault feature matrix, and the higher-order cutoff distance of each data point is calculated in the anisotropic space to obtain the anisotropic local density of each data point. For each data point, search for the target neighbor node in the local neighborhood set whose anisotropic local density is greater than that of each data point and whose anisotropic spatial distance is the smallest, and establish a density gradient directed edge from each data point to the target neighbor node. Traverse all data points in the high-dimensional fault feature matrix and connect all density gradient directed edges to construct a density gradient directed forest that represents the clustering trend of the data. The root node in the density gradient directed forest with an in-degree greater than zero and an out-degree of zero is identified as the density peak. By performing reverse tracing along the density gradient directed edges in the density gradient directed forest, all data points converging to the same density peak are divided into the same cluster set, and the aggregated cluster sets are output as multiple local high-density clusters in the high-dimensional fault feature matrix.

9. A collaborative management system for the entire process of motor components based on digital intelligence fusion, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the full-process collaborative management method for motor components based on digital intelligence fusion as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When executed by the processor, the instruction causes the processor to be configured to perform the collaborative management method for the entire process of motor components based on digital intelligence fusion, as described in any one of claims 1 to 8.

Images