Maintenance method and system for corner module configuration chassis based on product data management

Abstract

The invention relates to the technical field of product data management, and discloses a maintenance method and system for a corner module configuration chassis based on product data management. According to the method, digital twin bodies are introduced, multiple pieces of real-time working data can be managed and controlled, and the method specifically comprises the steps that a product data combination of the functional part is determined based on the working state of the functional part and the use scene of an angle module configuration chassis; a product data management system of the corner module configuration chassis is constructed according to the product data combination of each functional part and the corresponding functional type, so that the accuracy of the product data management system of the corner module configuration chassis is improved; meanwhile, the corresponding abnormal area is constructed based on the abnormal data of each abnormal part and the corresponding spatial position, and the corresponding dynamic maintenance event is determined according to the area position of the abnormal area, the corresponding influence range and the task list of the angle module configuration chassis, so that the accuracy of the dynamic maintenance event is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of product data management, and in particular to a product data management-based maintenance method and system for an angle module configuration chassis. BACKGROUND

[0002] With the development of science and technology, an angle module configuration chassis is present in a chassis in a modular structure and is applied to various vehicles. At this time, the angle module configuration chassis is applied to a corresponding logistics vehicle and serves as a chassis structure of the logistics vehicle. The angle module configuration chassis passes through different road scenes with the driving of the logistics vehicle and is worn out with the driving of the logistics vehicle. In the prior art, a use video of the angle module configuration chassis is collected, and corresponding abnormal parts are determined based on the recognition of the use video. However, the angle module configuration chassis has multiple functional components, and the product data combination and corresponding function categories of each functional component are ignored, which affects the accuracy of the product data management system of the angle module configuration chassis and leads to low accuracy of dynamic maintenance events. SUMMARY

[0003] The present application aims to overcome the shortcomings of the prior art, and provides a product data management-based maintenance method and system for an angle module configuration chassis.

[0004] The present application provides a product data management-based maintenance method for an angle module configuration chassis, which comprises the following steps: Based on the use video and the corresponding overall morphology of the angle module configuration chassis, a corresponding digital twin is constructed. Based on the detection of the digital twin, multiple functional components are determined, which are distributed at different positions. Based on the current use data of the angle module configuration chassis and the matching of the multiple functional components, multiple real-time working data are determined. According to the real-time working data and the corresponding functional components, the working state of the functional components is determined. Among the multiple functional components, based on the working state of the functional component and the use scene of the angle module configuration chassis, the product data combination of the functional component is determined. According to the product data combination of each functional component and the corresponding function category, a product data management system for the angle module configuration chassis is constructed. The product data management system construction logically constructs a multi-dimensional data cube, including a physical structure dimension, a data depth dimension, and a function breadth dimension. In the product data management system, based on the recognition of the product data management system, the working risk nodes of each functional component are predicted. According to the working risk nodes of each functional component, the working state of each functional component, and the abnormal signal of the angle module configuration chassis, multiple abnormal components are determined. The working risk nodes include specific risk types, prediction times, and confidence levels. Mark the abnormal data of each abnormal component, construct a corresponding abnormal area based on the abnormal data of each abnormal component and the corresponding spatial position, and determine a corresponding dynamic maintenance event according to the area position of the abnormal area, the corresponding influence range and the task list of the corner module configuration chassis.

[0005] The embodiment of the application provides a corner module configuration chassis maintenance system based on product data management, which is applied to the corner module configuration chassis maintenance method based on product data management.

[0006] Compared with the prior art, the beneficial effects of the application are: (1) A digital twin is constructed based on the use video of the corner module configuration chassis and the corresponding overall morphology; a plurality of functional components are determined based on the detection of the digital twin, and the plurality of functional components are distributed at different positions; a plurality of real-time working data are determined based on the matching of the current use data of the corner module configuration chassis and the plurality of functional components, and the working state of the functional component is determined according to the real-time working data and the corresponding functional component; the digital twin is introduced, and the plurality of real-time working data are further controlled, thereby improving the dynamic accuracy of the working state of the functional component.

[0007] (2) In each functional component, the product data combination of the functional component is determined based on the working state of the functional component and the use scene of the corner module configuration chassis; and a product data management system of the corner module configuration chassis is constructed according to the product data combination of each functional component and the corresponding functional category, so that the overall consideration of the product data combination of each functional component and the corresponding functional category is realized, and the accuracy of the product data management system of the corner module configuration chassis is improved.

[0008] (3) In the product data management system, the working risk nodes of each functional component are predicted based on the identification of the product data management system; a plurality of abnormal components are determined according to the working risk nodes of each functional component, the working state of each functional component and the abnormal signal of the corner module configuration chassis; the abnormal data of each abnormal component is marked, a corresponding abnormal area is constructed based on the abnormal data of each abnormal component and the corresponding spatial position, and a corresponding dynamic maintenance event is determined according to the area position of the abnormal area, the corresponding influence range and the task list of the corner module configuration chassis; the plurality of abnormal components are introduced, the abnormal area is further controlled, the overall consideration of the area position of the abnormal area, the corresponding influence range and the task list of the corner module configuration chassis is realized, and the accuracy of the dynamic maintenance event is improved. BRIEF DESCRIPTION OF DRAWINGS

[0009] Figure 1 is a flowchart of the corner module configuration chassis maintenance method based on product data management in the embodiment of the application; Figure 2 is a flowchart of step S11 in the maintenance method of the corner module configuration chassis based on product data management in the embodiments of the present application; Figure 3 is a flowchart of step S12 in the maintenance method of the corner module configuration chassis based on product data management in the embodiments of the present application; Figure 4 is a flowchart of step S13 in the maintenance method of the corner module configuration chassis based on product data management in the embodiments of the present application; Figure 5 is a flowchart of step S14 in the maintenance method of the corner module configuration chassis based on product data management in the embodiments of the present application; Figure 6 is a flowchart of step S15 in the maintenance method of the corner module configuration chassis based on product data management in the embodiments of the present application; Figure 7 is a structural composition schematic diagram of the maintenance system of the corner module configuration chassis based on product data management in the embodiments of the present application. DETAILED DESCRIPTION

[0010] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0011] Please refer to Figures 1 to 7 A maintenance method of a corner module configuration chassis based on product data management is applied to a product data management scene; the maintenance method of the corner module configuration chassis based on product data management comprises the following steps. Step S11: constructing a corresponding digital twin based on a use video of the corner module configuration chassis and a corresponding overall shape; determining a plurality of functional components based on detection of the digital twin, the plurality of functional components being distributed at different positions; Step S12: determining a plurality of real-time working data based on matching of current use data of the corner module configuration chassis and the plurality of functional components, and determining a working state of the functional component according to the real-time working data and the corresponding functional component; Step S13: among the functional components, determining a product data combination of the functional component based on the working state of the functional component and a use scene of the corner module configuration chassis; and constructing a product data management system of the corner module configuration chassis according to the product data combination of each functional component and a corresponding functional category; Step S14: in the product data management system, predicting a working risk node of each functional component based on identification of the product data management system, and determining a plurality of abnormal components according to the working risk node of each functional component, the working state of each functional component, and an abnormal signal of the corner module configuration chassis; Step S15: labeling the abnormal data of each abnormal component, constructing a corresponding abnormal area based on the abnormal data of each abnormal component and the corresponding spatial position, and determining a corresponding dynamic maintenance event according to the area position of the abnormal area, the corresponding influence range, and the task list of the corner module configuration chassis.

[0012] Reference Figure 2 In step S11, the specific steps are: S111: during the use of the corner module configuration chassis, a use video of the corner module configuration chassis is determined based on dynamic shooting of the corner module configuration chassis by the camera, and at the same time, the overall shape of the corner module configuration chassis is determined according to the matching of the model of the corner module configuration chassis and the chassis database. S112: a plurality of solid models are determined according to the use video of the corner module configuration chassis and the corresponding overall shape, a corresponding digital twin is constructed based on the plurality of solid models, a plurality of functional components are determined according to the digital twin and the component list of the module configuration chassis, and each functional component loads a corresponding function.

[0013] In the embodiment of the present application, during the use of the corner module configuration chassis, a use video of the corner module configuration chassis is determined based on dynamic shooting of the corner module configuration chassis by the camera, and at the same time, the overall shape of the corner module configuration chassis is determined according to the matching of the model of the corner module configuration chassis and the chassis database, which is compatible with the overall consideration of the matching of the model of the corner module configuration chassis and the chassis database, and ensures the accuracy of the overall shape of the corner module configuration chassis.

[0014] At this time, for the collection of dynamic video, the data source is a high-definition industrial camera deployed at the front, side, rear of the vehicle and the maintenance station, and the unstructured video stream output by the camera contains rich dynamic information of the chassis under acceleration, braking, turning and other working conditions, and is preprocessed by using Gaussian filtering, histogram equalization and other technologies; while the static shape matching is a structured data retrieval process, which uses the VIN code read by the on-board diagnostic interface as an index to retrieve a digital master model in STEP AP242 or JT format from the PDM system. The model not only contains accurate B-rep geometric representation, but also includes PMI assembly relationship and material metadata.

[0015] A three-dimensional structure is recovered from two-dimensional images by SfM and MVS algorithms to generate a dense point cloud, and then ICP algorithm is used to realize coordinate alignment and data fusion of the point cloud and the CAD model to form a parameterized mesh with real-world imprint. Functional component identification relies on deep learning, such as three-dimensional semantic segmentation network PointNet++ for face-by-face classification of the mesh model to identify specific component instances such as tires and suspensions, and then through geometric comparison and spatial positioning with the BOM list, the identified instances are bound with unique identifiers such as Part_ID, and automatically queried from the PDM to provide steering torque or support and damping functions.

[0016] Specifically, when a logistics vehicle with an A-Modular-Chassis-v2.1 chassis enters the intelligent maintenance station, the multi-camera above the entrance and the trench immediately starts working, capturing the real-time state of the chassis from different angles. The video stream clearly records the mud splatter marks of the left front corner module, the slight abnormal vibration of the right rear suspension at idle speed, and the slight tilting posture of the vehicle body to the left. At the same time, the vehicle sends its VIN code to the main control system through the wireless communication module, and the system successfully retrieves the complete digital master model of the A-Modular-Chassis-v2.1 from the PDM within 0.1 seconds, which accurately includes CAD data of all components such as in-wheel motors and steering actuators in the four corner modules.

[0017] The system reconstructs a three-dimensional point cloud of the A chassis using multiple videos, which accurately reproduces the tilting and mud state of the vehicle body. Then, through ICP algorithm, it is registered and fused with the v2.1 version of the CAD master model to generate a digital twin that is structurally perfect but covered with real mud texture and posture consistent with reality.

[0018] The system performs three-dimensional semantic segmentation on the twin to accurately identify components such as four tires and suspension assemblies. When analyzing the right rear corner module, the algorithm detects abnormal vibration patterns in the suspension components in this area by comparing the small displacement of multiple frames of point clouds. The system immediately identifies it as Part_ID:SUS-004-RR in the BOM and assigns it the function definition of providing support and damping for the right rear wheel, transforming it from a geometric body to a functionally abnormal component that requires special maintenance.

[0019] Further, a plurality of three-dimensional models are determined according to the use video of the A-Modular-Chassis-v2.1 chassis and the corresponding overall morphology, a corresponding digital twin is constructed based on the plurality of three-dimensional models, a plurality of functional components are determined according to the digital twin and the component list of the modular chassis, each functional component loads a corresponding function, and the overall consideration of the digital twin and the component list of the modular chassis is compatible, ensuring the accuracy of the plurality of functional components.

[0020] At this time, the system uses SIFT, ORB and other algorithms to establish tens of thousands of stable corresponding point pairs from multi-view images, laying the foundation for subsequent three-dimensional reconstruction; camera pose estimation and sparse reconstruction rely on the SfM algorithm, which calculates the three-dimensional coordinates of matching points through the principle of triangulation, forming a sparse point cloud that outlines the basic profile of the chassis; in the dense reconstruction stage, MVS algorithms such as PatchMatchNet perform depth inference on each pixel to generate a high-density point cloud, which is then converted into a continuous three-dimensional mesh model in PLY or OBJ format, i.e. a stereo model.

[0021] The system achieves coarse alignment by identifying macro features such as axles, and then iteratively optimizes using the ICP algorithm to minimize the distance error between the stereo model and the CAD model until convergence; after registration, the stereo model's color, texture, normal vector deviation, and other real-world information are input onto the precise mesh of the CAD model, achieving data fusion.

[0022] Three-dimensional instance segmentation networks such as PointGroup not only perform semantic classification on the mesh (such as tires), but also distinguish between different instances (such as the left front tire); the identified instances are matched with the BOM list based on their volume, spatial coordinates, and other geometric parameters, and after successful association, the system automatically pulls the unique identifier, design life, and maintenance manual for the component from the PDM, especially providing core function definitions such as driving torque or steering control, completing the final transformation from geometric entities to functional components.

[0023] Specifically, the surrounding 5 high-definition cameras capture images from different angles, and the system successfully extracts tens of thousands of feature points such as hub bolt corner points and car body logo edges from the video stream; through the SfM algorithm, the system not only determines the precise spatial position of each camera, but also generates a sparse point cloud that clearly outlines the overall profile of the A chassis; then, the MVS algorithm performs pixel-level depth calculation on the chassis surface, generating a dense point cloud that captures the scratches on the car body and the obvious mud on the left front fender, and finally processing it into a detailed three-dimensional mesh stereo model, becoming a snapshot-style three-dimensional replica of the A chassis at that moment.

[0024] The system loads this stereo model with mud and scratches along with the A chassis v2.1 version CAD master model retrieved from the PDM; through the ICP algorithm, the stereo model is accurately fitted onto the CAD model, and it is found that the suspension part of the right rear corner module has a persistent deviation of millimeters from the CAD standard position, which confirms the slight vibration observed in the video; after fusion, a high-fidelity digital twin is born: it has the perfect v2.1 structure, but the surface is covered with real mud texture, the car body posture is slightly tilted, and the geometric form of the right rear suspension has a quantifiable difference from the design standard.

[0025] The system runs a three-dimensional instance segmentation algorithm on the fused twin, successfully segmenting out four independent corner module assemblies, dozens of component instances such as shock absorbers; the system matches the corner module instance located at the right rear with the BOM, successfully associates it to the component ID ASM-004-RR, and identifies the shock absorber instance inside it, associated to ID SHK-004-RR; by querying the PDM, the system loads the right rear corner module with integrated drive, steering, and suspension functions for ASM-004-RR, and loads the function definition of attenuating the vibration of the right rear wheel for SHK-004-RR; at this point, the component with geometric deviation is no longer an unnamed model, but is clearly defined as a right rear corner module shock absorber, and has a complete engineering identity, laying a solid foundation for subsequent state judgment.

[0026] Reference Figure 3 In step S12, the specific steps are: S121: Real-time monitoring of the use process of the corner module configuration chassis, and collecting current use data of the corner module configuration chassis, determining a plurality of function data according to the recognition of the current use data of the corner module configuration chassis; S122: Determining a plurality of real-time working data according to the matching of the plurality of function data and the corresponding function components; each real-time working data dynamically changes with the change of time; collecting the current state of the function component, and determining the state change amount of the function component according to the current state of the function component and the corresponding real-time working data, to determine the working state of the function component.

[0027] In the embodiment of the present application, the use process of the corner module configuration chassis is monitored in real time, and the current use data of the corner module configuration chassis is collected, and a plurality of function data is determined according to the recognition of the current use data of the corner module configuration chassis, which is compatible with the overall consideration of the recognition of the current use data of the corner module configuration chassis, and ensures the accuracy of the plurality of function data.

[0028] At this time, the heterogeneity of the data source is reflected in the difference of physical interface and data format. CAN / CAN-FD bus transmits structured messages in broadcast form, while IMU, temperature sensor, etc. output continuous data stream through SPI or analog interface, and ADAS sensor produces image or point cloud and other unstructured big data.

[0029] The edge computing unit integrates multiple CAN / CAN-FD, Ethernet and analog / digital input interfaces as a data aggregation center; time synchronization is the cornerstone of this step, through PTP protocol or GPS PPS signal, to ensure that data from different sources such as motor speed and vehicle body acceleration are accurately aligned on the time axis within milliseconds or even microseconds.

[0030] DBC file plays the role of a decoding dictionary for CAN messages, which clearly defines the start bit, length, byte sequence, scaling factor, and offset of each signal in the message, so that the system can accurately convert a specific bit segment in the original hexadecimal code stream, such as 0x18FF50E5, into a physical quantity such as the right rear hub motor speed = 450.5 rpm.

[0031] For sensor signals, their physical characteristics need to be applied, for example, the resistance value of a PT1000 temperature sensor needs to be calculated through a calibration function such as the Callendar-VanDusen equation to obtain an accurate temperature reading; all the extracted engineering quantities, such as motor torque, steering angle, suspension acceleration, etc., are systematically classified into driving, steering, suspension, etc. functional data according to their described physical objects and functions, completing the transition from raw signals to structured information.

[0032] Specifically, the autonomous logistics vehicle with an angular module configuration chassis leaves the intelligent maintenance station and drives on urban roads, and the edge computing unit starts comprehensive data capture. It listens to the right rear corner module controller (RMCU) through the CAN-FD bus and periodically sends status messages with ID 0x18FF50E5; at the same time, it reads the real-time resistance value of the PT1000 temperature sensor integrated with the right rear hub motor through the SPI interface, and receives the three-axis acceleration and angular velocity raw data output by the IMU on the RMCU at a frequency of 100 Hz; in addition, the rear camera and the bottom laser radar also synchronously collect the image of the rear vehicle and the three-dimensional profile information of the road surface; all these data streams are collected by the edge computing unit and are uniformly stamped with GPS timestamps, forming a large and time-aligned raw data set.

[0033] The edge computing unit begins to analyze these raw data; it decodes the CAN message 0x18FF50E5 according to the A chassis DBC file, and successfully extracts three functional data: the right rear hub motor actual torque is 85.2 Nm, the rotor temperature is 78.5°C, and the steering angle is -2.1°; at the same time, it processes the raw data from the IMU, after coordinate transformation and filtering, it extracts the right rear corner module Z-axis acceleration of -9.85 m / s².

[0034] Furthermore, multiple real-time working data are determined based on the matching of multiple functional data and corresponding functional components; each real-time working data changes dynamically over time; the current state of the functional component is collected, and the state change of the functional component is determined based on the current state of the functional component and the corresponding real-time working data to determine the working state of the functional component. This approach takes into account both the current state of the functional component and the corresponding real-time working data, ensuring the accuracy of the state change of the functional component. At the same time, a digital twin is introduced, and multiple real-time working data are further controlled, improving the dynamic accuracy of the working state of the functional component.

[0035] At this point, a precise mapping table is established, which defines the logical association between functional data signal IDs and functional component IDs. This association includes not only the direct attribution of the right rear wheel hub motor temperature signal to the right rear wheel hub motor, but also the aggregate calculation of the vehicle's energy consumption obtained by calculating the power of multiple controllers, and even the mapping of the distance to the left front obstacle to the spatial position of the left front corner module through coordinate system transformation. After matching is completed, the system creates one or more real-time working data streams in the form of D_i(t) for each component and stores them in the time series database to support efficient time series analysis.

[0036] Time-domain analysis not only calculates statistical characteristics such as mean, variance, and RMS, but also captures dynamic indicators such as temperature change rate through differential analysis; frequency-domain analysis uses FFT to convert vibration or current signals to the frequency domain, and identifies early faults by identifying bearing fault characteristic frequencies or motor current harmonics; model baseline comparison compares real-time data with physical simulation or historical statistical models, and the calculated residuals become a key quantitative indicator for measuring performance deviation.

[0037] Multi-dimensional feature fusion is employed to avoid false alarms from a single indicator. The fusion method can be an expert system based on the IF-THEN rule, a machine learning classifier such as SVM or XGBoost that can learn complex nonlinear relationships, or a fuzzy logic system that can handle uncertainty. The final output is not only a discrete label such as normal or abnormal, but also includes a confidence score that quantifies the reliability of the judgment, providing richer information dimensions for subsequent decision-making.

[0038] Specifically, during the driving process, the corner module chassis passed through a long uphill section and continued driving on that section. The system collected the functional data that the temperature of the right rear wheel hub motor was 85.2°C. Based on the mapping relationship in the PDM, it accurately matched this data to the right rear wheel hub motor component (ID:MTR-004-RR) in the digital twin. At the same time, the torque and speed data of this component were also matched, which together formed three parallel real-time working data streams for this component: temperature (t), torque (t), and speed (t). These data fluctuated in real time on the monitoring screen, forming a dynamic working profile of the component.

[0039] The system performed an in-depth analysis of the MTR-004-RR data stream: Time domain analysis showed that the average temperature over the past 60 seconds was 86.5°C, with a temperature change rate of +0.8°C / min; Frequency domain analysis, after performing an FFT on the torque data stream, revealed a new spectral peak with increasing amplitude at the third harmonic of the fundamental frequency, suggesting early tooth surface wear in the reducer; Model baseline comparison, by inputting the current data into the motor thermal model, found that the actual temperature was 4.5°C higher than the model prediction, resulting in a significant residual.

[0040] The system's XGBoost decision engine integrates multiple state change quantities, such as temperature mean, temperature change rate, temperature residual, and 3 times the peak amplitude of the spectrum. After calculation, the final output working state is an abnormal coupling of overheating and mechanical wear, with a high confidence level of 88%. This clear state is then marked on the MTR-004-RR component of the digital twin, making it highlighted in red in the 3D model, and immediately passed to step S13, providing the core basis for triggering the corresponding product data combination and generating accurate maintenance warnings.

[0041] refer to Figure 4 In step S13, the specific steps are as follows: S131: Based on the recognition of the usage video of the corner module chassis, multiple scene features are determined, and the usage scenario of the corner module chassis is determined according to the feature shape, corresponding feature position and current working task of each scene feature; S132: Determine the product data combination of the functional component based on the usage scenario of the corner module chassis, the working status of the functional component, and the corresponding working data, and mark the functional type of the functional component; S133: Determine the first level of product data management content based on the product data combination of each functional component and the digital twin of the corner module configuration chassis; determine the second level of product data management content based on the functional types of each functional component and the digital twin of the corner module configuration chassis; and construct the product data management system of the corner module configuration chassis based on the first level of product data management content and the second level of product data management content.

[0042] In the embodiments of this application, multiple scene features are determined based on the recognition of the usage video of the corner module chassis. The usage scenario of the corner module chassis is determined according to the feature shape, corresponding feature position and current working task of each scene feature. This approach takes into account the feature shape, corresponding feature position and current working task of each scene feature, ensuring the accuracy of the usage scenario of the corner module chassis.

[0043] At this point, the system performs advanced computer vision analysis on the usage video acquired in S111, extracting multi-dimensional scene features. This is not just object detection, but also environmental understanding: Road features: using semantic segmentation networks (such as DeepLab, U-Net) to identify road surface materials (asphalt, cement, unpaved roads), lane lines, traffic signs, speed bumps, etc.; Environmental features: using object detection networks (such as YOLO, Faster R-CNN) to identify key surrounding objects, such as pedestrians, non-motorized vehicles, large vehicles, traffic light status, construction areas, etc.; Weather and lighting features: analyzing the overall tone, contrast, and dynamic range of the image to determine weather conditions (sunny, rainy, foggy) and lighting conditions (daytime, nighttime, inside a tunnel).

[0044] The extracted features have morphological features (such as the geometry of speed bumps) and location features (such as their position in the image coordinate system, which is then mapped to distance and orientation in the vehicle coordinate system). The system integrates these discrete features with the vehicle's current task (such as high-speed cruising, logistics delivery within the park, and emergency braking). It employs a rule-based inference engine or a lightweight classification model. For example, the rule could be: IF Road features = 'unpaved road' AND Task = 'logistics delivery' AND Environment features = 'open space' THEN Usage scenario = 'off-road delivery'. Finally, it outputs a structured, semantic usage scenario label, such as following other vehicles in congested urban areas, overtaking on highways, driving in the rain at night, and working on bumpy roads.

[0045] Furthermore, based on the usage scenarios of the corner module chassis, the working status of the functional component, and the corresponding working data, the product data combination of the functional component is determined, and the functional type of the functional component is marked. This approach takes into account the overall usage scenarios of the corner module chassis, the working status of the functional component, and the corresponding working data, ensuring the accuracy of the product data combination of the functional component.

[0046] At this point, the engine's rule base is not a simple IF-THEN statement, but encapsulates profound engineering logic. For example, bumpy road operation scenarios will be associated with the random vibration load model of the suspension system, overheat warning status will trigger attention to material temperature resistance and lubrication data, and abnormal torque fluctuation data will point to the geometric accuracy and wear model of the transmission system.

[0047] When a rule is triggered, the engine will accurately extract a subset of data from the PDM, including CAD models, FEA stress cloud diagrams, bench test reports, and historical failure case libraries, forming a dynamic product data combination focused on the current core issue.

[0048] The system predefines enumerated values ​​for components in dimensions such as load (standby / no-load to overload), dynamic (static to impact), and mode (normal to fail-safe). The marking process uses logic such as threshold judgment (e.g., marking as overload if torque exceeds 110% of the rated value), pattern recognition (e.g., analyzing the suspension acceleration spectrum to determine if it is a high-frequency dynamic), and state mapping (e.g., thermal performance degradation warning is directly mapped to performance limitation mode). Finally, the system combines the labels of multiple dimensions into a comprehensive functional category label that accurately describes the current situation of the component.

[0049] Specifically, when a logistics vehicle with a corner modular chassis enters the park and operates in a low-speed mixed-traffic scenario, its right rear wheel hub motor is monitored to be in an overload state. When the working data shows that the continuous torque exceeds 120Nm and the temperature reaches 95°C, the decision engine receives the input {Scenario: 'Low-speed mixed-traffic operation in the park', Status: 'Overload operation', Key data: 'High torque, High temperature'}, triggering an advanced rule. The system then dynamically constructs a product data combination for the motor from the PDM, including the motor thermal model, permanent magnet demagnetization curve, cooling system parameters, overload protection logic settings, and a library of overload fault cases for the same model of motor, precisely focusing on the core issues of overheating and overload.

[0050] The system categorizes the motor by function: after querying its rated torque of 100 Nm, the current torque of 120 Nm exceeds the 110% threshold, so its load dimension is marked as overload; analyzing its gradual changes in low-speed, high-torque data, the dynamic dimension is marked as gradual change; given its overload operating state, the mode dimension is marked as performance limitation mode; the motor is assigned the precise function category label of overload-gradual change-performance limitation mode. This highly condensed semantic information provides crucial context for the subsequent S14 step to predict the early arrival of its thermal fatigue risk node and for the S15 step to formulate targeted maintenance decisions.

[0051] Therefore, the first level of product data management content is determined based on the product data combination of each functional component and the digital twin of the corner module chassis. The second level of product data management content is determined based on the functional types of each functional component and the digital twin of the corner module chassis. A product data management system for the corner module chassis is constructed based on the first and second levels of product data management content. This system is compatible with the overall consideration of the functional types of each functional component and the digital twin of the corner module chassis, ensuring the accuracy of the second level of product data management content. At the same time, it realizes the overall consideration of the product data combination of each functional component and the corresponding functional types, improving the accuracy of the product data management system for the corner module chassis.

[0052] At this point, the system does not simply copy the data, but attaches the index of the product data combination as a live attribute package to the digital twin instance through the API interface; when engineers interact, the system retrieves the most authoritative data from source systems such as PDM and PLM in real time, and automatically overlays real-time working data curves and simulation models for comparison, ensuring the timeliness and context relevance of the data.

[0053] By using function category tags as a new index dimension independent of the physical BOM, the system can execute SQL-like aggregate queries, such as filtering out all components with overload in all function categories, regardless of whether they are motors or controllers; based on this, the system generates a system-level function view, such as highlighting all high-frequency dynamic components.

[0054] The product data management system is logically constructed as a multi-dimensional data cube, including physical structure dimension, data depth dimension, and functional breadth dimension. Its physical structure dimension (X-axis) is defined by the hierarchy of the digital twin, the data depth dimension (Y-axis) is composed of the data combinations associated with each component, and the functional breadth dimension (Z-axis) is formed by functional category labels and aggregated views. This cube supports standard OLAP operations: drill-down is to view in-depth data from the whole vehicle to the components, roll-up is to aggregate component data to functional categories, slicing is to fix a certain scenario or state for specialized analysis, and rotation is to switch the analysis perspective. This multi-dimensional analysis capability is unattainable by traditional PDM systems.

[0055] Specifically, the system does not simply copy data, but attaches the index of the product data combination as a live attribute package to the digital twin instance through the API interface; when engineers interact, the system retrieves the most authoritative data from source systems such as PDM and PLM in real time, and automatically overlays real-time working data curves and simulation models for comparison, ensuring the timeliness and context relevance of the data.

[0056] The core technology of breadth aggregation lies in using functional category tags as a new index dimension independent of the physical BOM. This allows the system to execute SQL-like aggregation queries, such as filtering out all components with overload in all functional categories, without caring whether they are motors or controllers. Based on this, the system generates a system-level functional view, such as highlighting all high-frequency dynamic components, realizing a shift in management perspective from focusing on where the objects are to focusing on how functions are distributed.

[0057] The product data management system is built by logically constructing a multi-dimensional data cube. Its physical structure dimension (X-axis) is defined by the hierarchy of the digital twin, the data depth dimension (Y-axis) is composed of the data combinations associated with each component, and the functional breadth dimension (Z-axis) is formed by functional category labels and aggregated views. This cube supports standard OLAP operations: drill-down is to view in-depth data from the whole vehicle to the components, roll-up is to aggregate component data to functional categories, slicing is to fix a certain scenario or state for special analysis, and rotation is to switch the analysis perspective. This multi-dimensional analysis capability is unattainable by traditional PDM systems.

[0058] refer to Figure 5 In step S14, the specific steps are as follows: S141: Real-time monitoring of the product data management system, dynamic identification of the product data management system, output of the data risk range of each functional component, prediction of the working risk node of each functional component based on the data risk range of each functional component and the product data combination of that functional component, so as to mark the working risk node of each functional component. S142: Determine the abnormal signal of the corner module configuration chassis based on the abnormal detection of the corner module configuration chassis, and determine the first abnormal content according to the abnormal signal of the corner module configuration chassis and the working risk nodes of each functional component; S143: Determine the second abnormal content based on the abnormal signals of the corner module chassis and the working status of each functional component, and determine multiple abnormal components based on the first abnormal content, the second abnormal content and the mapping relationship of abnormal components.

[0059] In the embodiments of this application, the product data management system is monitored in real time, the product data management system is dynamically identified, and the data risk range of each functional component is output. The working risk node of each functional component is predicted based on the data risk range of each functional component and the product data combination of the functional component, so as to mark the working risk node of each functional component. The method of marking the working risk node of each functional component is introduced.

[0060] At this point, monitoring not only covers the instantaneous values ​​and trends of real-time operating data such as temperature and pressure, but also regards the changes in function type labels (such as from normal mode to performance limit mode) and the switching of product data combinations as important risk indicators; risk quantification goes beyond simple threshold comparison. It defines a health function for each key parameter (such as the S-curve of temperature) and dynamically adjusts the threshold according to the usage scenario and function type. Finally, it calculates a comprehensive risk index through weighted average, realizing a refined and contextualized assessment of the health status of components.

[0061] The system draws upon deep knowledge from the product data combination: physical model deduction utilizes FEA or thermal-structural coupling models, inputting the current operating conditions for transient simulation to predict future temperature fields and stress distributions; data-driven models input real-time feature sequences into pre-trained models such as LSTM, outputting remaining service life (RUL) and failure modes; and for progressive failures such as fatigue, damage accumulation models such as Miner's Rule are used to calculate the number of cycles to reach the critical point. The results of these models are fused to form a working risk node, which includes specific risk types such as thermal demagnetization of permanent magnets, prediction time, and confidence level, achieving a precise mapping from abstract risks to specific failures.

[0062] Specifically, the system observed that the motor's function type had changed from medium load-gradient-normal mode to overload-gradient-performance limiting mode, and the product data combination had also been switched to a dedicated combination; the system detected that its real-time temperature was 95°C, corresponding to a temperature risk index of 0.8, and the torque was 125Nm (120% over the rated value), corresponding to a torque risk index of 0.9. Based on the preset weights, the comprehensive risk index was calculated to be 0.84, and its data risk range was ultimately rated as High.

[0063] Because the risk range reached High, the system initiated deep prediction. It retrieved the motor thermal-magnetic coupling simulation model and the permanent magnet annealing RUL model based on historical data from the product data set. The physical model deduction showed that under this condition, the local hot spot of the permanent magnet would reach the demagnetization critical temperature of 150°C after 12 hours. The data-driven RUL model predicted that its remaining service life was 13.5 hours, and the failure mode was permanent magnet thermal demagnetization. The system fused the results of the two models and finally marked a working risk node on the MTR-004-RR component of the digital twin: the risk type was permanent magnet thermal demagnetization, the prediction time was about 12.5 hours, and the confidence level was 92%. This clear risk node provided a direct basis for generating accurate maintenance events in the future.

[0064] Furthermore, based on the anomaly detection of the corner module configuration chassis, the abnormal signals of the corner module configuration chassis are determined. The first abnormal content is determined according to the abnormal signals of the corner module configuration chassis and the working risk nodes of each functional component. This takes into account the overall consideration of the abnormal signals of the corner module configuration chassis and the working risk nodes of each functional component, ensuring the accuracy of the first abnormal content.

[0065] At this point, the DTC code provided by OBD is the most direct hard threshold over-limit signal, while model-based residual analysis detects more subtle anomalies by comparing the difference between the actual output and the model prediction. For complex systems, data-driven models such as LSTM autoencoders are used to learn normal patterns and use reconstruction error as an anomaly criterion, which can effectively identify pattern deviations. Frequency domain analysis focuses on vibration and acoustic signals, and uses FFT to identify energy spectral peaks at specific fault frequencies. All these raw signals from various sources are ultimately standardized into a structured anomalous signal object, ensuring the uniformity of subsequent processing.

[0066] The knowledge base stores knowledge in the form of (risk node) -- [precursor features] > (abnormal signal features). When an abnormal signal is captured, the system extracts its key features and searches the knowledge base for all risk nodes with that feature as a precursor. The association matching algorithm quantifies the association strength based on the degree of feature matching. For example, if a 15% increase in current is detected, and a rule in the knowledge base indicates that the precursor is an increase in current of more than 10%, then the association strength is very high. The first abnormal content generated is a structured diagnostic hypothesis that not only describes the phenomenon but also points out the future risks it predicts.

[0067] Specifically, the system applied an LSTM autoencoder model to the three-phase current signal of the motor. In the past 10 minutes, the model reconstruction error was below 0.02, but at the current moment, the error suddenly jumped to 0.15, far exceeding the threshold of 0.05. The system immediately determined that this was a mode deviation anomaly and generated a standardized abnormal signal object with a key feature value of reconstruction error = 0.15.

[0068] The system is aware that S141 has marked the permanent magnet thermal demagnetization risk as a working risk node on this component. The system then searches the causal chain knowledge base and finds a highly relevant piece of knowledge: (Permanent magnet thermal demagnetization risk) -- [Precursor characteristics: current waveform distortion, leading to mode deviation] > (increased LSTM reconstruction error). Since the captured abnormal signal perfectly matches the precursor characteristics in the knowledge base, the system successfully associates the abnormal signal with the working risk node and generates the first abnormal content: the current signal of the right rear wheel hub motor (MTR-004-RR) has deviated from its mode. This abnormality is highly consistent with the typical precursor of the predicted 'permanent magnet thermal demagnetization' risk node, and comes with a high confidence level of 95%. This conclusion is no longer a simple abnormality of the motor, but rather that the abnormality of the motor is a sign that the permanent magnet has begun to demagnetize. This provides an extremely critical and forward-looking diagnostic basis for subsequent accurate positioning and maintenance decisions.

[0069] Therefore, the second abnormal content is determined based on the abnormal signals of the corner module chassis and the working status of each functional component. Multiple abnormal components are determined based on the mapping relationship between the first abnormal content, the second abnormal content and the abnormal components. This approach takes into account the overall consideration of the mapping relationship between the first abnormal content, the second abnormal content and the abnormal components, and ensures the accuracy of multiple abnormal components.

[0070] At this point, the system does not view the abnormal signal in isolation, but interprets it within the context of its operating state. This correlation may be a consistent confirmation, such as an abnormal current exceeding the limit under overload operation, where the two corroborate each other; or it may be a contradictory revelation, such as an abnormal temperature rise under normal standby state, which points to a more hidden fault. In addition, the operating state provides the necessary context for the abnormal signal, allowing the severity of the same signal under different operating conditions to be distinguished. The generated second abnormal content is an objective, precise, and structured statement of the current facts.

[0071] Bayesian networks are a typical example. They treat operational risk nodes, operational states, anomalous signals, and anomalous components as network nodes. By inputting first and second anomalous contents as evidence, they infer from the past to calculate the posterior probability of each component being a true anomalous component. The fusion decision logic includes: when two anomalous contents point to the same component, an evidence reinforcement effect is generated, significantly increasing its confidence; when they point to different components, an evidence conflict handling mechanism is triggered, generating a system-level anomalous hypothesis or requesting more data; the system can also construct a complete evidence chain from the anomalous signal to the risk node and then to the final component; only components with a posterior probability exceeding the decision threshold are locked as anomalous components.

[0072] Specifically, when the system detects an abnormal current signal pattern deviation in the right rear hub motor (MTR-004-RR) of the corner module chassis, the system finds that the MTR-004-RR was in overload operation when the abnormality occurred. Since the current abnormality is a typical accompanying phenomenon under overload operation, and the two are highly consistent, the system immediately generates a second abnormality: the right rear hub motor (MTR-004-RR) was detected to have a current signal pattern deviation under the 'overload operation' state, objectively describing the current situation.

[0073] The system's Bayesian network received two pieces of input evidence: Evidence 1 is the first anomaly, indicating that the current anomaly is highly consistent with the risk of thermal demagnetization of the permanent magnet; Evidence 2 is the second anomaly, confirming that the anomaly occurred under overload operation. Since both pieces of evidence clearly point to MTR-004-RR, the evidence in the network is strengthened, and the system calculates that the posterior probability of this component being a true anomaly is as high as 98.7%, which far exceeds the 95% decision threshold. Therefore, the system ultimately determines that MTR-004-RR is the only anomalous component, with an anomaly confidence level of 98.7%. This high-confidence diagnostic result will be directly transmitted to S15 to generate an accurate and timely dynamic maintenance event, such as immediately arranging for the replacement or in-depth inspection of the right rear wheel hub motor.

[0074] refer to Figure 6 In step S15, the specific steps are as follows: S151: Based on the detection of each abnormal component, mark the abnormal data of each abnormal component, and construct the corresponding abnormal area according to the abnormal data of each abnormal component, the corresponding spatial position, and the digital twin of the corner module configuration chassis. S152: In each abnormal area, mark the location of the abnormal area, and determine the scope of influence of the abnormal area based on the shape of the area and the corresponding multiple abnormal components; at the same time, collect the task list of the corner module chassis, and determine the work impact of the corner module chassis during the working process based on the task list of the corner module chassis and the location of the abnormal area. S153; Based on the content of the work's impact, the scope of impact of each abnormal area, and the current work tasks of the corner module chassis, determine the dynamic maintenance data, and based on the dynamic maintenance data and the product data management system of the corner module chassis, determine the corresponding dynamic maintenance events.

[0075] In the embodiments of this application, abnormal data of each abnormal component is marked based on the detection of each abnormal component. The corresponding abnormal region is constructed according to the abnormal data of each abnormal component, the corresponding spatial position and the digital twin of the corner module configuration chassis. This takes into account the overall consideration of the abnormal data of each abnormal component, the corresponding spatial position and the digital twin of the corner module configuration chassis, and ensures the accuracy of the corresponding abnormal region.

[0076] At this point, the system will automatically gather all the information fragments related to the specific abnormal component from the entire process from S121 to S143. This includes the original data snapshot that triggered the abnormality, the status determined by S122, the risk predicted by S141, the signal captured by S142, and the diagnostic conclusion and confidence level generated by S143. This information is integrated into a unified data object and bound to the component ID, forming a comprehensive and multi-dimensional abnormality profile, which provides a solid information foundation for subsequent in-depth analysis.

[0077] The abnormal regions are dynamically selected based on the different abnormal situations: For a single abnormal component, a single-point region strategy is usually adopted, that is, based on the component's geometric model, a slightly larger enclosing region is generated through a buffer algorithm; for multiple spatially adjacent abnormal components, spatial clustering algorithms such as DBSCAN are used to aggregate them into clusters, and then their convex hulls or alpha shapes are calculated to form aggregated regions; for specific anomalies such as thermal risks or liquid leaks, the system may even call physical simulation models to generate propagation regions based on physical predictions by solving heat conduction equations, etc. These generated regions are rendered as eye-catching geometric objects such as semi-transparent red and flashing outlines, and are presented intuitively in the digital twin.

[0078] Specifically, once the system identifies the right rear hub motor (MTR-004-RR) of the corner module chassis as an abnormal component, the system automatically backtracks and aggregates all information related to MTR-004-RR: real-time operating data (temperature 95°C, torque 125Nm), operating status (overload operation), operating risk nodes (permanent magnet thermal demagnetization risk, expected to occur in 12.5 hours), abnormal signals (current signal mode deviation, reconstruction error 0.15), and an abnormality confidence level of 98.7%. The system packages this information into a structured data object and binds it to the model instance of MTR-004-RR in the digital twin, forming a complete abnormality file.

[0079] The system obtained the precise three-dimensional coordinates and geometric model of MTR-004-RR from the digital twin. Since there is only one abnormal component at present, the system adopts a single-point region construction strategy. Based on the geometric model of the motor, a three-dimensional buffer algorithm with a radius of 5 cm is applied to generate a smooth ellipsoidal abnormal region. On the visualization interface of the digital twin, at the right rear corner module of chassis A, a semi-transparent red ellipsoid is rhythmically and slowly flashing. Engineers can intuitively see that the core of the problem is precisely locked within this spatial range.

[0080] Furthermore, in each abnormal area, the location of the abnormal area is marked, and the influence range of the abnormal area is determined based on the area shape and the corresponding multiple abnormal components. At the same time, the task list of the corner module chassis is collected, and the work impact content of the corner module chassis during the operation is determined based on the task list of the corner module chassis and the location of the abnormal area. This comprehensive consideration of the task list of the corner module chassis and the location of the abnormal area ensures the accuracy of the work impact content of the corner module chassis during the operation.

[0081] At this point, the system accurately marks the location and shape of the abnormal area, performs spatial adjacency queries, and identifies all non-abnormal components that are directly in contact with or spatially adjacent to it. More importantly, the system starts from the abnormal component and performs a breadth-first or depth-first search along the topology graph to trace strongly correlated paths such as energy, structure, and information, for example, tracing from the abnormal motor to its controller and high-voltage wiring harness. The final determined scope of influence is a set consisting of the abnormal area, adjacent components, and path-traced components. The system can even assign influence weights to each component based on topological distance and relationship strength, forming a fault influence cluster that is physically and functionally closely related.

[0082] The system collects a dynamic task list including the current task, priority, performance requirements, and subsequent task queue. Through a predefined matrix, it analyzes the core functions required to complete each task. By cross-referencing the degree of functional impairment involved in abnormal areas with task requirements, the system can quantify the impact, such as a 50% decrease in driving function performance. Finally, it generates structured work impact content, detailing the delay of the current task, the risk to subsequent tasks, and specific safety prompts.

[0083] Specifically, after the system constructs an abnormal region for the right rear hub motor (MTR-004-RR) of the corner module chassis, the system uses the topology graph of the digital twin to perform an adjacency query and finds that the components directly adjacent to the abnormal region are the right rear hub reducer and the right rear steering actuator. Through associated path tracing, the system also finds the energy path (to the right rear motor controller) and the structural path (to the right lower control arm) starting from MTR-004-RR. The system determines the scope of influence as {right rear hub motor, right rear hub reducer, right rear steering actuator, right rear motor controller, right lower control arm}, forming a fault influence cluster.

[0084] The system identified the current task as Delivery Task A (high priority, 25 minutes remaining), with the subsequent task being Return to the Park Charging Station. Through function-task dependency matrix analysis, the system found that the right rear-wheel drive and right rear-wheel steering functions, which are affected, have limited impact on the current task but pose a high risk to the subsequent task requiring long-distance travel. The system generated the following impact information: 'Delivery Task A' can continue, but power response will be weakened, with an estimated delay of 10 minutes; without maintenance, the 'Return to Charging Station' task carries a high risk, with continuous load potentially causing the vehicle to break down en route; and it recommends immediate maintenance after completing the current task.

[0085] Therefore, dynamic maintenance data is determined based on the content of the work's impact, the impact range of each abnormal area, and the current work tasks of the corner module chassis. Corresponding dynamic maintenance events are then determined based on this dynamic maintenance data and the corner module chassis's product data management system. This approach integrates the overall considerations of dynamic maintenance data and the corner module chassis's product data management system, ensuring the accuracy of the corresponding dynamic maintenance events. Simultaneously, multiple abnormal components are introduced to further control the abnormal areas, achieving a holistic consideration of the abnormal area's location, its corresponding impact range, and the corner module chassis's task list, thus improving the accuracy of dynamic maintenance events.

[0086] At this point, the decision engine receives not only qualitative descriptions of the impact, but also quantitative information on key components such as delay time, task priority, and scope of impact. At its core is a rule engine or decision tree based on Multi-Attribute Decision Analysis (MADA), which incorporates complex rules formulated by experts. For example, if a task can be completed but is delayed and has a high priority, the strategy is to maintain it immediately after the task is completed, or if the security risk is high, to shut it down immediately. Based on the matching rules, the engine outputs a structured decision object that clarifies the macro strategy, timing, level, and preliminary list of required resources for maintenance.

[0087] The system uses dynamic maintenance data as a query request. The PDM system utilizes its complete product knowledge base to accurately match general descriptions such as hub motor assemblies to specific spare part numbers; it automatically retrieves and associates the latest version of the standard operating procedure (SOP), and even interactive 3D animations; at the same time, it supplements relevant safety operating procedures and quality control points. These strategies and details are integrated into a complete dynamic maintenance event package, which is accurately pushed to relevant parties through the TCU, realizing a closed loop from decision-making to execution.

[0088] Specifically, after the system completes the impact assessment for the corner module chassis, the decision engine integrates the impact content of the current task that can be completed but is delayed by 10 minutes, the high-risk work content of subsequent tasks, the impact range including the right rear wheel hub motor, and the high priority of the current task, and matches the strategy of immediate maintenance after the task is matched; the system then generates dynamic maintenance data: the maintenance strategy is to maintain immediately after the task, the recommended execution time is to execute immediately after the current task is completed, the maintenance level is component-level replacement, and the required resource list includes the wheel hub motor assembly, torque wrench kit, and chassis maintenance engineer.

[0089] The system queries the PDM system for replacement information on the hub motor assembly. The PDM system returns the precise spare part number MTR-004-RR-ASSY-V2.1, a 10-minute 3D interactive disassembly and assembly animation, and a safety checklist for high-voltage power-off operation. The system integrates this information and finally generates and pushes a dynamic maintenance event: the driver terminal receives a concise instruction to go to the designated repair point after the task is completed; the dispatch center sees the abnormal vehicle highlighted on the digital map and automatically dispatches the work order to the repair team; and the repair team terminal receives a complete set of technical data work orders containing the precise spare part number, 3D repair guide, and safety instructions, and the entire maintenance process is seamlessly activated.

[0090] Please see Figure 7 , Figure 7 This is a schematic diagram of the structural composition of the maintenance system for the corner modular chassis based on product data management according to an embodiment of the present invention; the maintenance system for the corner modular chassis based on product data management includes: Detection module 21 is used to construct a corresponding digital twin based on the usage video and overall shape of the corner module chassis; multiple functional components are determined based on the detection of the digital twin, and the multiple functional components are distributed in different positions; The working status module 22 is used to determine multiple real-time working data based on the current usage data of the corner module configuration chassis and the matching of multiple functional components, and to determine the working status of the functional component based on the real-time working data and the corresponding functional component. Product data management system module 23 is used to determine the product data combination of each functional component based on its working status and the usage scenario of the corner module chassis; and to construct the product data management system of the corner module chassis based on the product data combination of each functional component and the corresponding functional type. The abnormal component module 24 is used to predict the working risk nodes of each functional component based on the identification of the product data management system in the product data management system, and to identify multiple abnormal components based on the working risk nodes of each functional component, the working status of each functional component and the abnormal signals of the corner module configuration chassis. The dynamic maintenance event module 25 is used to mark the abnormal data of each abnormal component, construct the corresponding abnormal area based on the abnormal data of each abnormal component and the corresponding spatial location, and determine the corresponding dynamic maintenance event according to the area location of the abnormal area, the corresponding impact range and the task list of the corner module configuration chassis.

[0091] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A maintenance method for a corner module chassis based on product data management, characterized in that, include: A digital twin is constructed based on the usage video of the corner module chassis and its corresponding overall shape; Multiple functional components are identified based on the detection of digital twins, and these components are distributed in different locations. Based on the current usage data of the corner module chassis and the matching of multiple functional components, multiple real-time working data are determined, and the working status of the functional component is determined according to the real-time working data and the corresponding functional component. In each functional component, the product data combination of the functional component is determined based on its working status and the usage scenario of the corner module chassis; the product data management system of the corner module chassis is constructed according to the product data combination of each functional component and the corresponding functional type; the construction of the product data management system is a logically constructed multi-dimensional data cube, including physical structure dimension, data depth dimension, and functional breadth dimension; In the product data management system, the operational risk nodes of each functional component are predicted based on the identification of the product data management system. Multiple abnormal components are identified based on the operational risk nodes of each functional component, the operational status of each functional component, and the abnormal signals of the corner module configuration chassis. The operational risk nodes include the specific risk type, prediction time, and confidence level. The abnormal data of each abnormal component is marked. Based on the abnormal data of each abnormal component and its corresponding spatial location, a corresponding abnormal region is constructed. The corresponding dynamic maintenance event is determined according to the regional location of the abnormal region, its corresponding impact range, and the task list of the corner module configuration chassis.

2. The maintenance method for a corner module chassis based on product data management according to claim 1, characterized in that, The usage video and corresponding overall shape of the corner module configuration chassis are used to construct a digital twin; multiple functional components are determined based on the detection of the digital twin, and these functional components are distributed in different locations, including: During the use of the corner module chassis, the usage video of the corner module chassis is determined based on the dynamic shooting of the corner module chassis by the camera. At the same time, the overall shape of the corner module chassis is determined according to the matching of the model of the corner module chassis and the chassis database. Based on the usage video and corresponding overall shape of the corner modular chassis, multiple 3D models are determined. A corresponding digital twin is constructed based on the multiple 3D models. Based on the component list of the digital twin and the modular chassis, multiple functional components are determined, and each functional component carries a corresponding function.

3. The maintenance method for a corner module chassis based on product data management according to claim 1, characterized in that, The current usage data of the corner module chassis is used to determine multiple real-time operating data by matching multiple functional components. Based on this real-time operating data and the corresponding functional component, the operating status of that functional component is determined, including: The system monitors the usage of the corner module chassis in real time and collects its current usage data. Based on the identification of the current usage data of the corner module chassis, multiple functional data are determined. Multiple real-time working data are determined by matching multiple functional data with corresponding functional components; each real-time working data changes dynamically over time; the current state of the functional component is collected, and the state change of the functional component is determined based on the current state of the functional component and the corresponding real-time working data, so as to determine the working state of the functional component.

4. The maintenance method for a corner module chassis based on product data management according to claim 1, characterized in that, In each functional component, the product data combination of that functional component is determined based on its working status and the usage scenario of the corner modular chassis; a product data management system for the corner modular chassis is constructed based on the product data combinations of each functional component and their corresponding functional types, including: Multiple scene features are determined based on the recognition of usage videos of the corner modular chassis. The usage scenario of the corner modular chassis is determined according to the feature shape, corresponding feature position and current working task of each scene feature. Based on the usage scenario of the corner module chassis, the working status of the functional component, and the corresponding working data, determine the product data combination of the functional component and mark the functional type of the functional component.

5. The maintenance method for a corner module chassis based on product data management according to claim 4, characterized in that, The process of determining the product data combination of each functional component based on its working state and the usage scenario of the corner modular chassis, and constructing a product data management system for the corner modular chassis based on the product data combinations and corresponding functional types of each functional component, further includes: The first level of product data management content is determined based on the combination of product data of each functional component and the digital twin of the corner module chassis. The second level of product data management content is determined based on the functional types of each functional component and the digital twin of the corner module chassis. The product data management system of the corner module chassis is constructed based on the first level of product data management content and the second level of product data management content.

6. The maintenance method for a corner module chassis based on product data management according to claim 1, characterized in that, In the product data management system, operational risk nodes for each functional component are predicted based on the identification within the product data management system. Multiple abnormal components are identified based on these operational risk nodes, the operational status of each functional component, and abnormal signals from the corner module chassis. These include: The system monitors the product data management system in real time, dynamically identifies the system, and outputs the data risk range of each functional component. Based on the data risk range of each functional component and the combination of product data for that functional component, the system predicts the operational risk nodes of that functional component and marks the operational risk nodes of each functional component.

7. The maintenance method for a corner module chassis based on product data management according to claim 6, characterized in that, In the product data management system, based on the identification of the product data management system, the operational risk nodes of each functional component are predicted. Based on the operational risk nodes of each functional component, the operational status of each functional component, and the abnormal signals of the corner module chassis, multiple abnormal components are identified. This also includes: Anomaly signals of the corner module chassis are determined based on anomaly detection of the corner module chassis, and the first anomaly content is determined based on the anomaly signals of the corner module chassis and the working risk nodes of each functional component. The second abnormal content is determined based on the abnormal signals of the corner module chassis and the working status of each functional component. Multiple abnormal components are determined based on the mapping relationship between the first abnormal content, the second abnormal content and the abnormal components.

8. The maintenance method for a corner module chassis based on product data management according to claim 1, characterized in that, The abnormal data of each abnormal component is marked. Based on the abnormal data and corresponding spatial location of each abnormal component, a corresponding abnormal region is constructed. According to the regional location of the abnormal region, the corresponding impact range, and the task list of the corner module configuration chassis, the corresponding dynamic maintenance event is determined, including: Based on the detection of each abnormal component, the abnormal data of each abnormal component is marked, and the corresponding abnormal area is constructed according to the abnormal data of each abnormal component, the corresponding spatial position, and the digital twin of the corner module configuration chassis.

9. The maintenance method for a corner module chassis based on product data management according to claim 8, characterized in that, The method of marking abnormal data for each abnormal component, constructing corresponding abnormal regions based on the abnormal data and corresponding spatial locations of each abnormal component, determining corresponding dynamic maintenance events based on the regional location of the abnormal region, the corresponding impact range, and the task list of the corner module configuration chassis, also includes: In each abnormal area, the location of the abnormal area is marked, and the influence range of the abnormal area is determined based on the area shape and the corresponding multiple abnormal components; at the same time, the task list of the corner module configuration chassis is collected, and the work impact of the corner module configuration chassis during the working process is determined based on the task list of the corner module configuration chassis and the location of the abnormal area. Based on the content of the work's impact, the scope of impact of each abnormal area, and the current work tasks of the corner module chassis, dynamic maintenance data is determined. Based on this dynamic maintenance data and the product data management system of the corner module chassis, corresponding dynamic maintenance events are determined.

10. A maintenance system for a corner module chassis based on product data management, characterized in that, The maintenance system for the corner modular chassis based on product data management is applied to the maintenance method for the corner modular chassis based on product data management as described in any one of claims 1-9.

Images