Mechanical product full-life-cycle collaborative design system based on digital twinning

Abstract

The invention discloses a mechanical product full-life-cycle collaborative design system based on digital twinning, and relates to the technical field of mechanical product design. The system comprises a scheme design module, a preliminary design module, a three-dimensional structure digital design module, a product manufacturing module, a service monitoring and early warning module, a collaborative engine module, a three-dimensional collaborative design database module and a data analysis module. According to the system, a collaborative design framework throughout the whole process of design, manufacturing and service is constructed, full-process sharing of data and knowledge is realized, and information of a service stage is fed back to design and manufacturing links through a closed-loop feedback mechanism; the problems that in the prior art, all links are split, degradation modes are difficult to recognize, and the service life prediction precision is low are effectively solved, and the scientificity and reliability of product design are improved.

Description

Technical Field

[0001] This invention relates to the field of mechanical product design technology, and in particular to a collaborative design system for the entire lifecycle of mechanical products based on digital twins. Background Technology

[0002] With the deepening development of industry and intelligent manufacturing, the design process of mechanical products is becoming increasingly complex, and the demand for multidisciplinary collaboration and full life cycle management is constantly increasing. Traditional mechanical product design methods often isolate the design, manufacturing, and service stages, lacking effective data and information sharing mechanisms. This makes it difficult for design optimization to fully incorporate actual feedback from the manufacturing and usage stages. In terms of product condition monitoring, existing methods rely heavily on experience-based judgment and simple statistical analysis, making it difficult to accurately capture the inherent laws of complex nonlinear degradation processes. In terms of remaining life prediction, traditional models have limited accuracy, and maintenance decisions often lack a scientific basis, affecting product reliability and service life.

[0003] In recent years, digital twin technology, as a key enabling technology connecting physical entities and virtual models, has been widely used in industry. However, existing digital twin applications mostly focus on local optimization of single stages, such as simulation verification in the design phase or status monitoring in the usage phase, and have not yet formed a collaborative design framework that runs through the entire product lifecycle. In addition, in terms of product degradation pattern recognition and lifespan prediction, traditional methods mostly use analysis methods based on statistical or shallow machine learning models, which have limited ability to extract features from high-dimensional, multi-source, and dynamically evolving operational data, making it difficult to effectively identify potential failure modes and evolution paths.

[0004] Topological data analysis (TDA), as an emerging data science method, possesses the unique advantage of extracting robust topological features from complex, high-dimensional data and has shown potential in multiple fields. However, in the full lifecycle management of mechanical products, especially in the collaborative analysis of integrated design, manufacturing, and service data, this technology is still in the initial exploratory stage and has not yet formed a systematic theoretical framework and technical system.

[0005] Therefore, there is an urgent need to build a collaborative design system that can deeply integrate digital twins and topological data analysis, so as to connect the data flow and knowledge flow of mechanical products throughout their entire life cycle, realize closed-loop optimization from design, manufacturing to service, and improve the scientific nature of product design, the precision of manufacturing, and the level of intelligent service. Summary of the Invention

[0006] To address these issues, this invention provides a collaborative design system for the entire lifecycle of mechanical products based on digital twins. This system solves the problems of fragmented design, manufacturing, and service processes, difficulty in identifying complex product degradation patterns, and low accuracy in predicting remaining lifespan in existing technologies.

[0007] To address the aforementioned technical problems, embodiments of the present invention provide a collaborative design system for the entire lifecycle of mechanical products based on digital twins. This system includes: The solution design module is used to design functional solutions for the product. The preliminary design module is used to design the product's structural, electrical, and process parameters. The 3D structure digital design module is used to optimize the 3D structure of the product based on collaborative feedback information; The product manufacturing module is used to complete the manufacturing process of the designed scheme; The service monitoring and early warning module is used to acquire operational data of the product during use based on topological data analysis theory, construct a topological representation of the product's operating status, identify degradation patterns, and predict the product's remaining service life. The module includes a multi-dimensional data topology structure representation unit, a degradation pattern identification unit, and a remaining service life prediction unit. The multi-dimensional data topology structure representation unit maps product operational data to a high-dimensional feature space and extracts topological features. The degradation pattern identification unit constructs a topological network of the product's degradation state based on the topological features and identifies degradation patterns. The remaining service life prediction unit constructs a topological state space evolution model based on the degradation patterns and the topological network, predicts the product's remaining service life, and generates an early warning signal when the remaining service life is lower than expected. The collaboration engine module is used to obtain collaborative feedback information from different designers and product users and pass it to the 3D structure digital design module. The 3D collaborative design database module is used to store design schemes, structural parameters, product operation data, and operational degradation status. The data analysis module is used to analyze and process the data in the 3D collaborative design database module.

[0008] Preferably, the multidimensional data topology representation unit includes: The data acquisition layer is used to collect multi-dimensional operational data of the product through a distributed sensor network. The preprocessing layer is used to clean the collected data, handle outliers, and synchronize the time. The point cloud mapping layer is used to map preprocessed multidimensional running data into a set of points in a high-dimensional space; The distance calculation layer is used to construct distance metric functions that adapt to the characteristics of mechanical products; Complex construction layer, used to construct Vietoris-Rips complex sequences based on distance thresholds; The homology computation layer is used to compute the topological features of complex sequences in different dimensions. The feature representation layer is used to generate persistent graphs and barcodes, quantifying the topological features of product operation data.

[0009] Preferably, the degradation pattern recognition unit includes: The feature space construction layer is used to transform persistent graph features into a mathematical representation of the persistent landscape and establish the feature space. The similarity calculation layer is used to calculate the topological similarity between different running states; Dimensionality reduction mapping layer is used to map a high-dimensional feature space to a low-dimensional space; The overlay construction layer is used to build overlapping overlays on the dimensionality-reduced parameter space. The segmented clustering layer is used to perform clustering analysis on data points within each coverage area; The network construction layer is used to represent clustering results as nodes, establish connections between nodes that share data points, and form a degenerate pattern topology network. The structural analysis layer is used to extract key network topology and identify product degradation patterns.

[0010] Preferably, the network construction layer represents the clustering results as nodes, and connections are established between nodes sharing data points to form a degradation mode topology network, specifically including: For coverage Each region in The data points are clustered to obtain a set of clusters. Each cluster In a network, a node is represented as a single node. If two nodes share data points, an edge is established between them, with the weight of the edge set as the proportion of the number of shared points to the total number of points. For two clusters and ,if Then, an edge is created between them, with the weight of the edge being: ; in, This indicates the number of points in the intersection of two clusters; This indicates the number of union points of two clusters; This represents the weight of the edge, with a value range of 1. ; The final constructed network is represented as ,in, It is a set of nodes, with each node corresponding to a cluster; It is a set of edges, representing the connections between nodes; It is the edge weight function, which represents the connection strength.

[0011] Preferably, the remaining lifetime prediction unit includes: The historical trajectory construction layer is used to construct a set of degenerate trajectories in the topological state space based on historical data. The state-space modeling layer is used to construct a topological state-space evolution model and describe the state transition process. The health index mapping layer is used to define the health index function and map the topological state to the health index, quantifying the degree of degradation. The evolution simulation layer is used to simulate possible future degradation trajectories based on the current state and the topological state space evolution model; The lifetime distribution calculation layer is used to calculate the time distribution of reaching the failure threshold based on the simulated degradation trajectory. The prediction update layer is used to continuously update the parameters and prediction results of the topological state space evolution model and the health index function by combining real-time data. The early warning generation layer is used to generate tiered early warning signals based on the prediction results.

[0012] Preferably, the state space modeling layer uses a Markov chain model to describe the state transition process, and the constructed topological state space evolution model is as follows: ; in, The elements of the state transition matrix represent the transition from state to state. Transition to state The probability of; It is observed from the state Transition to state The number of times; From state The total number of transitions to any state; To account for the impact of operating conditions on the transition probability, a conditional transition rule is introduced: ; in, Indicates operating conditions Below, from the state Transition to state The conditional probability; Indicates the state Under these conditions, the state is observed simultaneously. and conditions The joint conditional probability; Indicates the state Observation conditions The conditional probability.

[0013] Preferably, the health index function of the health index mapping layer is expressed as: ; in, For state Health index, range of values ; This represents the current topology state. For reference regarding health status; The current state Compared with reference health status Topological distance between them; It is the normalization factor.

[0014] Preferably, the evolution simulation layer simulates possible future degradation trajectories based on the current state and the topological state space evolution model, specifically including: Given the current state and state transition matrix A simulated trajectory is generated through the following steps: initialization: ; For time Based on transition probability Randomly select the next state ; Return to simulation trajectory ; in, Indicates time The state; Indicates the current state; Indicates from state Transition to state The probability of; Indicates the number of simulation steps; Repeat the above process to generate multiple possible evolutionary trajectories, reflecting the uncertainty of future states.

[0015] Preferably, the lifetime distribution calculation layer calculates the time distribution for reaching the failure threshold based on the simulated degradation trajectory, specifically including: Set failure threshold When the health index The simulation is considered invalid if the health index first falls below the failure threshold. For each simulated trajectory, the first arrival time is calculated, which is the time when the health index first falls below the failure threshold. ; in, This indicates the time of first arrival of the trajectory, i.e., the time of failure. Indicates time status Corresponding health index; Indicates the failure threshold; The first arrival times of all trajectories form a distribution, the mean of which can be used as a point estimate of the remaining lifetime, and the quantiles form a confidence interval. The point estimate of the remaining lifetime is as follows: ; in, Indicates the predicted remaining useful life; Indicates the number of simulated trajectories; Indicates the first The first arrival time of each trajectory.

[0016] Preferably, the service monitoring and early warning module, together with the 3D structure digital design module, the product manufacturing module, and the collaborative engine module, forms a closed-loop feedback mechanism for: The identified degradation-sensitive points are transmitted to the three-dimensional structure digital design module for structural optimization; The identified quality-sensitive points are transmitted to the product manufacturing module for manufacturing process optimization. The degradation mode and failure mechanism analysis results are passed to the collaborative engine module for multidisciplinary collaborative optimization.

[0017] As can be seen from the above technical solutions, this invention application has the following beneficial effects: (1) This invention constructs a unified digital twin framework that runs through the entire process of mechanical product design, manufacturing and service by coordinating modules such as scheme design, preliminary design, three-dimensional digital design, product manufacturing and service monitoring and early warning. The system utilizes the three-dimensional collaborative design database module and the collaborative engine module to realize cross-link data and knowledge sharing, effectively solving the problem of the separation between design, manufacturing and service in traditional methods, and forming a complete closed loop from virtual design to physical entity, and then optimizing the design through feedback from actual operation data.

[0018] (2) The core innovation of this invention lies in applying topological data analysis theory to monitoring and early warning during the product service phase. Through the multidimensional data topology structure representation, degradation pattern recognition, and remaining lifetime prediction units in the service monitoring and early warning module, the system can extract robust topological features (such as persistent graphs and barcodes) from high-dimensional, multi-source operational data, construct a topological network in the degradation state, and perform accurate remaining lifetime prediction based on the state space evolution model. This method can capture complex nonlinear degradation patterns that are difficult to detect by traditional statistical or simple machine learning models, thereby identifying potential faults in advance and improving prediction accuracy and reliability.

[0019] (3) This invention utilizes a closed-loop feedback mechanism comprised of a service monitoring and early warning module, a 3D structure digital design module, a product manufacturing module, and a collaborative engine module. This mechanism provides real-time feedback to the design and manufacturing stages on the degradation-sensitive points, quality-sensitive points, degradation modes, and failure mechanism analysis results identified during actual use. This enables product structure design to be optimized based on real degradation data, and manufacturing processes to be improved based on identified quality issues. This achieves data-driven continuous product iteration and full lifecycle optimization, ultimately enhancing product reliability, service life, and market competitiveness. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly described below. Referring to the drawings will make the features and advantages of the present invention clearer. The drawings are illustrative and should not be construed as limiting the present invention in any way. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a schematic diagram of the overall architecture of the system of the present invention; Figure 2 This is a schematic diagram of the service monitoring and early warning module of the present invention; Figure 3 This is a schematic diagram of the processing flow of the multidimensional data topology structure representation unit of the present invention; Figure 4 This is a schematic diagram of the topology network construction of the degradation pattern recognition unit of the present invention; Figure 5 This is a schematic diagram of the state-space evolution model of the remaining lifetime prediction unit of the present invention; Figure 6 This is a schematic diagram of the data flow and closed-loop feedback of the system of the present invention.

[0021] Explanation of reference numerals in the accompanying drawings: 10. Scheme Design Module; 20. Preliminary Design Module; 30. 3D Structure Digital Design Module; 40. Product Manufacturing Module; 50. Service Monitoring and Early Warning Module; 60. Collaborative Engine Module; 70. 3D Collaborative Design Database Module; 80. Data Analysis Module; 90. Cloud Platform Common Module; 51. Multidimensional Data Topology Representation Unit; 52. Degradation Pattern Recognition Unit; 53. Remaining Lifetime Prediction Unit; 511. Data Acquisition Layer; 512. Preprocessing Layer; 513. Point Cloud Mapping Layer; 514. Distance calculation layer; 515, Complex construction layer; 516, Homology calculation layer; 517, Feature representation layer; 521, Feature space construction layer; 522, Similarity calculation layer; 523, Dimensionality reduction mapping layer; 524, Cover construction layer; 525, Segmentation and clustering layer; 526, Network construction layer; 527, Structure analysis layer; 531, Historical trajectory construction layer; 532, State space modeling layer; 533, Health index mapping layer; 534, Evolution simulation layer; 535, Lifespan distribution calculation layer; 536, Prediction update layer; 537, Early warning generation layer. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Example 1: To address the problems in existing technologies such as the fragmentation of design, manufacturing, and service processes, the difficulty in identifying complex product degradation modes, and the low accuracy of remaining life prediction, such as... Figure 1 As shown, this invention proposes a collaborative design system for the entire lifecycle of mechanical products based on digital twins. The system includes a scheme design module 10, a preliminary design module 20, a three-dimensional structure digital design module 30, a product manufacturing module 40, a service monitoring and early warning module 50, a collaborative engine module 60, a three-dimensional collaborative design database module 70, a data analysis module 80, and a cloud platform common module 90.

[0024] In this embodiment of the invention, the solution design module 10 is used to design the functional solution of the mechanical product. Specifically, the solution design module 10 establishes the functions that the product needs to achieve based on the product's operating environment, determines functional requirements and functional constraints, defines product performance and product performance indicators, constructs the topology of the product functional model, obtains the product's functional implementation strategy, establishes the product functional model, and evaluates the product functional implementation strategy.

[0025] In this embodiment of the invention, the preliminary design module 20 includes a mechanical submodule, an electrical submodule, and a process submodule, which respectively implement structural, electrical, and process parameter design. Specifically, the mechanical submodule includes a design scheme selection unit and a preliminary three-dimensional structural modeling unit, used to provide designers with existing design schemes or to design new design schemes themselves, and to create multiple preliminary three-dimensional mechanical structure models based on the product structure design scheme. The electrical submodule and the process submodule are responsible for electrical design and process design, respectively. The three submodules work together to achieve multidisciplinary collaborative design.

[0026] In this embodiment of the invention, the 3D structure digital design module 30 is used to optimize the 3D structure of the product based on collaborative feedback information. Specifically, the 3D structure digital design module 30 includes a knowledge mining subunit, a model optimization subunit, a simulation analysis subunit, and a 3D model display subunit. The knowledge mining subunit mines design knowledge for structural parameter completion from the 3D collaborative design database module 70; the model optimization subunit iteratively optimizes the 3D structure model of the digital product; the simulation analysis subunit performs simulation analysis on the performance of the 3D structure model and optimizes the 3D model in real time based on feedback data; the 3D model display subunit renders the optimized 3D structure model to generate the final product digital model.

[0027] In this embodiment of the invention, the product manufacturing module 40 is used to complete the manufacturing process of the design scheme. Specifically, the product manufacturing module 40 includes a 3D model generating virtual assembly model unit, a 3D model production process planning unit, and a 3D model CNC machining process automatic scheduling unit. The 3D model generating virtual assembly model unit performs simulation analysis on the 3D structural model to generate a virtual assembly model for the virtual assembly process; the 3D model production process planning unit generates the manufacturing process flow based on the virtual assembly model; and the 3D model CNC machining process automatic scheduling unit is used to generate CNC machining processes and establish the product CNC machining process flow.

[0028] In this embodiment of the invention, the service monitoring and early warning module 50 is the core innovative module of the invention. It is used to acquire the product's operating data during use based on topological data analysis theory, construct a topological representation of the product's operating status, identify degradation patterns, and predict the product's remaining service life. The innovative implementation of this module will be described in detail below.

[0029] In this embodiment of the invention, the collaboration engine module 60 is used to obtain collaborative feedback information from different designers and product users, and to transmit the collaborative feedback information to the 3D structure digital design module 30. The collaboration engine module 60 is key to realizing closed-loop feedback throughout the entire lifecycle; it feeds back the data and knowledge acquired in the service phase to the design phase, promoting continuous product optimization.

[0030] In this embodiment of the invention, the three-dimensional collaborative design database module 70 is used to store design schemes, structural parameters, product operating data, and operating degradation status. It serves as the central hub for data interaction among the various modules of the system, supporting data storage, retrieval, and sharing.

[0031] In this embodiment of the invention, the data analysis module 80 is used to analyze and process the data in the 3D collaborative design database module 70. The data analysis module 80 includes a data acquisition unit, a data cleaning unit, a data storage unit, a data visualization unit, and a knowledge mining and analysis unit. The data acquisition unit is used to acquire product data, usage data, design knowledge, model parameters and performance data, and simulation data; the data cleaning unit processes the data input from the data acquisition unit; the data storage unit stores the input data from the data acquisition unit and the data cleaning unit; the data visualization unit collects data from the data storage unit and displays it visually; the knowledge mining and analysis unit mines and analyzes the data in the design knowledge database and the 3D collaborative design database module 70 to extract design knowledge.

[0032] In this embodiment of the invention, the cloud platform common module 90 is used to provide other modules with data transmission, data management, information interaction, business integration, artificial intelligence, visual interaction and security and privacy functions, so as to realize remote collaborative design by designers from different regions and departments.

[0033] like Figure 2 As shown, the service monitoring and early warning module 50 includes a multi-dimensional data topology representation unit 51, a degradation pattern recognition unit 52, and a remaining lifetime prediction unit 53. These three units constitute a complete link from data acquisition and feature extraction to pattern recognition and lifetime prediction, which is the core innovation of this invention.

[0034] The multidimensional data topology representation unit 51 is the foundation of the service monitoring and early warning module 50, used to map product operation data to a high-dimensional feature space and extract topological features. For example... Figure 3 As shown, the multidimensional data topology structure representation unit 51 includes a data acquisition layer 511, a preprocessing layer 512, a point cloud mapping layer 513, a distance calculation layer 514, a complex construction layer 515, a homology calculation layer 516, and a feature representation layer 517.

[0035] In this embodiment of the invention, the data acquisition layer 511 collects multi-dimensional operational data of the product through a distributed sensor network. Specifically, for the CNC machine tool spindle system, the collected data includes multi-dimensional data such as temperature (temperature of the front and rear bearings of the spindle), vibration (acceleration and velocity in the X, Y, and Z directions), sound (spectral characteristics), and current (spindle motor current). The acquisition frequency varies depending on the data type; for example, temperature data is collected every 10 seconds, and vibration data is collected every 0.1 seconds to ensure the timeliness and completeness of the data.

[0036] In this embodiment of the invention, the preprocessing layer 512 cleans the collected data, handles outliers, and synchronizes the time. For outliers, a method based on interquartile range is used. The detection and processing are carried out using the following methods: ; in, The first quartile of the data represents the value at the 25th percentile after sorting. The third quartile represents the value at the 75th percentile after sorting the data. The interquartile range represents the degree of dispersion in the data distribution.

[0037] Data points If satisfied or Values ​​that do not meet the criteria are considered outliers and can be replaced using the mean, median, or linear interpolation methods. For data with different sampling frequencies, time synchronization is achieved through interpolation or downsampling methods to ensure data consistency. For example, for temperature and vibration data, the vibration data can be downsampled to the same frequency as the temperature data, or the temperature data can be linearly interpolated to the sampling points of the vibration data.

[0038] In this embodiment of the invention, the point cloud mapping layer 513 maps the preprocessed multidimensional running data into a set of points in a high-dimensional space. Each point This represents a multidimensional state vector within a time window. In one embodiment of the invention, for a CNC machine tool spindle system with four types of sensors—temperature, vibration, sound, and current—each point... It can be represented as: ; in, For the first The front bearing temperature within a time window, in degrees Celsius; For the first The rear bearing temperature within a time window, in degrees Celsius; , , They represent the first The velocity in the X, Y, and Z directions within a time window, in millimeters per second; , , They represent the first The acceleration in the X, Y, and Z directions within a time window, in meters per second²; Indicates the first a time window Each spectral feature is typically a dimensionless normalized value; Indicates the first Motor current for each time window, in amperes; The vector dimension is equal to the total number of all features (in this example). ); This represents the total number of time windows.

[0039] To ensure the comparability of different physical quantities, data needs to be normalized. Commonly used normalization methods are min-max normalization or Z-fraction normalization. or ; in: This is the original data; The data is after normalization; and These are the minimum and maximum values ​​of the data, respectively. The mean of the data; Let be the standard deviation of the data. In this invention, the Z-score normalization method is preferably used to reduce the impact of outliers.

[0040] In this embodiment of the invention, the distance calculation layer 514 constructs a distance metric function adapted to the characteristics of the mechanical product. Traditional Euclidean distance often fails to accurately reflect the true differences between the multidimensional states of mechanical products. This invention employs weighted Mahalanobis distance, defined as: ; in, For any two points and The weighted Mahalanobis distance between them; The covariance matrix of the data; This is the weight matrix; Indicates matrix transpose; This represents the inverse of the covariance matrix.

[0041] Specifically, for CNC machine tool spindle systems, weights can be set based on expert experience or data analysis results. For example, the weight of vibration characteristics can be set to 0.4, temperature characteristics to 0.3, current characteristics to 0.2, and sound characteristics to 0.1. This means that the weight matrix... The diagonal elements are set to these values ​​accordingly. For example, for data with dimensions of temperature, vibration, current, and sound in that order, The diagonal elements can be set to .

[0042] In this embodiment of the invention, the complex construction layer 515 is based on a distance threshold. Construct the Vietoris-Rips complex sequence. Specifically, for any two points... and ,when At that time, connections are established between them to form a simple complex. Different thresholds The selection forms a series of nested complexes, constituting a filtering sequence of topological features. .

[0043] threshold The choice of threshold has a significant impact on the complex construction result. In this embodiment of the invention, the initial threshold can be determined in the following way. : ; in, This is an adjustment coefficient, ranging from 1.5 to 2.5, used to control the sparsity of the initial complex; The total number of points in the point cloud; Point The distance to its nearest neighbor; This represents the average distance from all points to their nearest neighbors.

[0044] In this embodiment of the invention, the threshold growth sequence can be set as follows: ,in: This is the growth rate, typically ranging from 1.2 to 1.5, used to control the rate at which the threshold grows. The number of increments is typically 10-20, depending on the data distribution characteristics and computing resources. In embodiments of the present invention, a preferred value is... , , These parameters perform well in a variety of mechanical systems.

[0045] In this embodiment of the invention, the homology computation layer 516 computes the topological features of the complex sequence in different dimensions. In algebraic topology, the Betti number... Used to describe Number of dimensional topological features: Indicates the number of connected components. Indicates the number of rings (one-dimensional holes), This represents the number of cavities (two-dimensional holes). As the threshold... As the number of elements increases, topological features will be generated and disappear, forming persistent homology.

[0046] Furthermore, this invention captures the topological features of the data by calculating the evolution of the Betti number in different dimensions as a function of a threshold. Specifically, for each threshold... Calculate its corresponding complex Betti number , and The curve of the Betti number as a function of the threshold was obtained. .

[0047] In this embodiment of the invention, the feature representation layer 517 generates a persistent graph and a barcode to quantify the topological features of the product's operational data. The persistent graph is a two-dimensional scatter plot describing the lifecycle of topological features; the horizontal axis represents the threshold for feature appearance, and the vertical axis represents the threshold for feature disappearance. Each point... Represent a topological feature, where The threshold for the occurrence (birth) of the feature. This is the threshold for feature disappearance (death).

[0048] Barcodes use horizontal line segments to represent the lifecycle of features, with the length of the line segment corresponding to the persistence of the feature. The starting point of each line segment corresponds to the threshold at which the feature appears, and the ending point corresponds to the threshold at which the feature disappears.

[0049] Specifically, the persistence graph and barcode show significant differences for different health states of CNC machine tool spindle systems, providing an important basis for identifying degraded states. For example, a healthy system typically exhibits a simpler topology, with fewer points in the persistence graph, concentrated near the diagonal; while a degraded system typically exhibits a more complex topology, with more points in the persistence graph, and some points far from the diagonal, indicating the presence of more persistent topological features.

[0050] Furthermore, the multidimensional data topology representation unit 51 also employs a multi-scale feature extraction strategy, including: Microscale: Set the time window width to 2-10 seconds to capture transient behavioral features; Mesoscale: Set the time window width to 1-5 minutes to capture work cycle characteristics; Macro scale: Set the time window width to 1 to 24 hours to capture long-term evolutionary trends.

[0051] Through multi-scale analysis, the system can comprehensively capture the characteristics of product operation at different time scales, improving the completeness and accuracy of state characterization. For example, for CNC machine tool spindle systems, the microscale can capture the transient vibration characteristics of bearings, the mesoscale can capture performance changes within a machining cycle, and the macroscale can capture long-term degradation trends such as bearing wear.

[0052] The degradation pattern recognition unit 52 constructs a topological network of product degradation status based on topological features to identify degradation patterns. For example... Figure 4As shown, the degradation pattern recognition unit 52 includes a feature space construction layer 521, a similarity calculation layer 522, a dimensionality reduction mapping layer 523, a cover construction layer 524, a segmentation and clustering layer 525, a network construction layer 526, and a structure analysis layer 527.

[0053] In this embodiment of the invention, the feature space construction layer 521 transforms persistent graph features into comparable mathematical representations, establishing a feature space. There are various mathematical representation methods for persistent graphs; this invention employs a persistent landscape representation. For persistent graphs... ,in Indicates the first Birth and death thresholds for each topological feature, persistent landscape function Defined as: ; in: For the first A landscape function at point The value at; The x-axis variable typically takes values ​​within a certain range. ; k=1,2,… indicates the first, second, and third digits of the set of digits. A landscape function, typically taken as... From 1 to 5 or 10; For the first Birth threshold for each topological feature; For the first The death threshold of each topological feature; This represents the total number of points in the persistent graph. Calculation points The corresponding trigonometric functions are The value at that location.

[0054] In this embodiment of the invention, the persistent landscape is transformed into a set of functions, facilitating subsequent analysis and comparison. In practical calculations, continuous landscape functions are typically discretized into vectors, i.e., within an interval... Take evenly from the top At points, calculate the value of each landscape function at these points, forming... A 500-dimensional feature vector. In embodiments of the present invention, m=100 and k=5 are typically chosen to obtain a 500-dimensional feature vector.

[0055] The similarity calculation layer 522 calculates the topological similarity between different running states. This invention uses Wasserstein distance to calculate the similarity between persistent graphs, which is defined as: ; in, and These are two persistent graphs, both of which are sets of points; yes and Wasserstein distance between them; From arrive A bijection represents a matching relationship between points; It is the order of the distance, controlling the sensitivity of distance calculation; Point and its mapping point The Euclidean distance between them; Denotes the infimum, that is, all possible bijections. Among them, the one that minimizes the total distance.

[0056] In embodiments of the present invention, typically selected This refers to the second-order Wassstein distance. For CNC machine tool spindle systems, state classification accuracy based on persistent graph similarity can reach over 95%, far exceeding traditional methods. Calculating the Wassstein distance is an optimal transport problem, which can be solved using algorithms such as the Hungarian algorithm.

[0057] In this embodiment of the invention, the dimensionality reduction mapping layer 523 maps the high-dimensional feature space to a low-dimensional space. This invention employs a multidimensional scaling (MDS) method for dimensionality reduction while preserving the topological structure. The goal of multidimensional scaling is to maintain the distance relationships between data points in the low-dimensional space; its optimization objective is: ; in, It is the midpoint of the original high-dimensional space. and points The distance between them, i.e., the Wasserstein distance ; and These are their mapping points in low-dimensional space, usually two-dimensional or three-dimensional points; It is the Euclidean distance in low-dimensional space; It is the set of coordinates of all points in low-dimensional space. It represents the total number of state points.

[0058] Optimization of MDS typically employs methods such as gradient descent. In this embodiment of the invention, the feature space is usually reduced to two dimensions for easier visualization and analysis. For a CNC machine tool spindle system, points in different health states form distinct clusters in the reduced-dimensional plane, reflecting the evolution of the system state.

[0059] In this embodiment of the invention, the coverage construction layer 524 constructs overlapping coverages for the dimensionality-reduced parameter space. The invention employs a uniform coverage strategy, dividing the parameter space into a series of overlapping intervals to form coverages. In practical implementation, for a two-dimensional parameter space... It can be divided into Each rectangular area has an overlap rate of 30% to 50% between adjacent areas.

[0060] Furthermore, for two-dimensional space The method for constructing a cover includes the following steps: 1. Determine the number of divisions along the x-axis. Number of divisions along the y-axis ; 2. Calculate the width of each region: ,high: ,in , , The overlap rate; 3. Determine the center point of the region: , , ; 4. Using each center point Centered on, construct a width of Height is rectangular area .

[0061] Excessive overlap can lead to an overly complex network structure, while insufficient overlap may result in the loss of connections between data. In this embodiment of the invention, a preferred overlap ratio is... Coverage This means that adjacent areas overlap by 40%. For CNC machine tool spindle systems, this setting can better capture the structural features of the state space.

[0062] In this embodiment of the invention, the segmented clustering layer 525 performs clustering analysis on the data points within each coverage area. This invention primarily employs the density-based clustering algorithm DBSCAN, whose core idea is to define clusters based on density, meaning a cluster is the largest density-connected set of density-connected points. The two key parameters of the DBSCAN algorithm are: , represents the neighborhood radius; , which represents the minimum number of neighboring points required to become a core point.

[0063] The main steps of the DBSCAN algorithm are: Step 1: For each point Find its neighborhood radius All points within, if the number is greater than or equal to ,but As the core point, among which This represents the minimum number of neighboring points required to become a core point. Step 2: For each unvisited core point Create a new cluster and set its neighborhood radius. All points within the cluster are added to the cluster; Step 3: For each new point in the cluster, if it is a core point, set its neighborhood radius. All unvisited points within the cluster are added to the cluster; Step 4: Repeat step 4 until the cluster stops growing; Step 5: Mark any point that is not assigned to any cluster as a noise point.

[0064] In an embodiment of the present invention, It can be set to 15% to 25% of the diameter of the local area. It can be set to 3% to 5% of the total number of points. For CNC machine tool spindle systems, 1 to 3 clusters are usually formed in each coverage area, corresponding to different operating states or degradation stages.

[0065] In this embodiment of the invention, the network construction layer 526 represents the clustering results as nodes, and connections are established between nodes sharing data points to form a degradation mode topology network. Specifically, for coverage Each region in The data points are clustered to obtain a set of clusters. Each cluster In a network, this is represented as a node. If two nodes (from clusters in different coverage areas) share data points, an edge is established between them, and the weight of the edge can be set as the proportion of the number of shared points to the total number of points.

[0066] Specifically, for two clusters and ,if Then, an edge is created between them, with the weight of the edge being: ; in, This indicates the number of points in the intersection of two clusters; This indicates the number of union points of two clusters; This represents the weight of the edge, with a value range of 1. ; The final constructed network is represented as ,in, It is a set of nodes, with each node corresponding to a cluster; It is a set of edges, representing the connections between nodes; It is the edge weight function, which represents the connection strength.

[0067] In this embodiment of the invention, the structural analysis layer 527 extracts the key topology of the network and identifies product degradation patterns. This invention primarily focuses on the following structural features in the network: Loop structure: corresponds to cyclic degradation mode, such as periodic failure in mechanical system; Branching structure: Corresponds to bifurcated degradation paths, reflecting different directions of fault development; Centrality metrics, such as degree centrality and betweenness centrality, are used to identify key degenerate states. For the network ,node Degree centrality is defined as: ; in, For nodes Degree centrality; For the edge The weights; node Betweenness centrality is defined as: ; in, Indicates from node To the node The number of shortest paths; Indicates passing through nodes from arrive The number of shortest paths; For nodes Betweenness centrality.

[0068] Nodes with high degree centrality typically represent important states, while nodes with high betweenness centrality typically represent key points in state transitions. Loop structures can be identified by computing the basic loops of the network, while branch structures can be identified by analyzing the network's connectivity and tangent points.

[0069] In this embodiment of the invention, for the spindle system of CNC machine tools, various degradation modes can be identified by analyzing the topology network structure, such as early bearing wear, poor lubrication, and shaft imbalance. The identification accuracy can reach more than 90%, and potential faults can be detected 30% to 50% earlier than traditional methods.

[0070] The remaining useful life prediction unit 53 constructs a topological state-space evolution model to predict the remaining useful life of the product. For example... Figure 5 As shown, the remaining lifespan prediction unit 53 includes a historical trajectory construction layer 531, a state space modeling layer 532, a health index mapping layer 533, an evolution simulation layer 534, a lifespan distribution calculation layer 535, a prediction update layer 536, and an early warning generation layer 537.

[0071] In this embodiment of the invention, the historical trajectory construction layer 531 constructs a set of degenerate trajectories in the topological state space based on historical data. The persistent graph feature sequence of the product's historical operating data is considered as a trajectory in the topological space. Each Represents the topological state at a certain moment.

[0072] In embodiments of the present invention, operational data from multiple products of the same model can be collected to construct a trajectory library containing different degradation modes, providing a data foundation for subsequent modeling. For example, for a CNC machine tool spindle system, operational data from more than 10 machine tools can be collected, with each machine tool operating for at least 6 months, forming a complete degradation trajectory dataset.

[0073] In this embodiment of the invention, the state-space modeling layer 532 establishes a topological state-space evolution model to describe the state transition process. This invention uses a Markov chain model to describe the state transition process and constructs a state transition matrix. ,in Indicates from state Transition to state The probability of this. The estimation of the state transition matrix is ​​based on historical trajectory data: ; in, The elements of the state transition matrix represent the transition from state to state. Transition to state The probability of; It is observed from the state Transition to state The number of times; From state The total number of transitions to any state; To account for the impact of operating conditions on the transition probability, a conditional transition rule is introduced: ; in, Indicates operating conditions Below, from the state Transition to state The conditional probability; Indicates the state Under these conditions, the state is observed simultaneously. and conditions The joint conditional probability; Indicates the state Observation conditions The conditional probability.

[0074] Specifically, for a CNC machine tool spindle system, the operating state can be discretized into 10 to 20 representative states, and a state transition matrix can be constructed to describe the evolution law of the system. Operating conditions Factors such as load, rotational speed, and ambient temperature can significantly affect the degradation rate of equipment.

[0075] In this embodiment of the invention, the health index mapping layer 533 maps the topological state to a health index, quantifying the degree of degradation. This invention defines a health index function. topology state Mapped to The health index represents a range of values, where 1 indicates perfect health and 0 indicates complete failure. The health index can be calculated as follows: ; in, For state Health index, range of values ; This represents the current topology state. For reference to health status, data is usually taken from the operation of new products or products that have just been repaired; The current state Compared with reference health status The topological distance between them can be obtained using the Wasserstein distance; It is a normalization factor, which can be set to the maximum distance value observed to ensure that the health index falls within the [0,1] interval.

[0076] Specifically, the health status can be referenced by selecting the topology status corresponding to the operating data of new products or products that have just been repaired. For CNC machine tool spindle systems, the health index has a good correlation with traditional performance indicators (such as vibration amplitude, temperature, etc.), but can reflect the trend of performance degradation earlier.

[0077] In this embodiment of the invention, the evolution simulation layer 534 simulates possible future degradation trajectories based on the current state and the topological state space evolution model. This invention employs the Monte Carlo simulation method to generate multiple possible evolution trajectories, specifically including: Given the current state and state transition matrix A simulated trajectory is generated through the following steps: initialization: ; For time Based on transition probability Randomly select the next state ; Return to simulation trajectory ; in, Indicates time The state; Indicates the current state; Indicates from state Transition to state The probability of; Indicates the number of simulation steps; Repeating the above process (e.g., 1000 times) generates multiple possible evolutionary trajectories, reflecting the uncertainty of future states. These trajectories constitute the probability distribution of the product's future state evolution, providing a basis for predicting remaining lifespan.

[0078] In this embodiment of the invention, the lifetime distribution calculation layer 535 calculates the time distribution for reaching the failure threshold based on the simulated degradation trajectory, specifically including: Set failure threshold (like When the health index The simulation is considered invalid at a certain time; for each simulated trajectory, the First Passage Time is calculated, which is the time when the health index first falls below the failure threshold. ; in, This indicates the time of first arrival of the trajectory, i.e., the time of failure. Indicates time status Corresponding health index; Indicates the failure threshold; The first arrival times of all trajectories form a distribution, the mean of which can be used as a point estimate of the remaining lifetime, and the quantiles form a confidence interval. The point estimate of the remaining lifetime is as follows: ; in, Indicates the projected remaining useful life. Indicates the number of simulated trajectories; Indicates the first The first arrival time of each trajectory.

[0079] Furthermore, the 95% confidence interval can be expressed as: ,in and These represent the 2.5% and 97.5% percentiles of the first arrival time, respectively.

[0080] Specifically, for CNC machine tool spindle systems, the relative error of remaining life prediction can be controlled within 15%, providing a reliable basis for maintenance decisions. Failure threshold. The setting can be adjusted according to actual application needs, and is usually between 0.2 and 0.3. A lower threshold corresponds to a stricter failure standard.

[0081] In this embodiment of the invention, the prediction update layer 536 continuously updates the prediction results based on real-time data. This invention employs a Bayesian update framework, continuously updating the parameters of the state transition matrix and the health index function as new data is acquired. Specifically, the prior distribution of the parameters is combined with the likelihood function to obtain the posterior distribution of the parameters: ; in, Represents model parameters, including the state transition matrix. and health index function Parameters; Represents observation data; It is the prior distribution of the parameters, reflecting the initial beliefs about the parameters; It is the likelihood function, representing the likelihood of the parameter Observed data The probability of; It is the posterior distribution of the parameters, representing the distribution after observing the data. The subsequent belief in updating the parameters; It indicates a direct proportional relationship.

[0082] By continuously updating the posterior distribution of parameters, the model can adaptively adjust and improve prediction accuracy. In practical applications, methods such as particle filtering can be used to implement Bayesian updates to handle nonlinear and non-Gaussian state-space models.

[0083] In this embodiment of the invention, the early warning generation layer 537 generates tiered early warning signals based on the prediction results. The invention designs a four-level early warning mechanism: Normal (green): Predicted remaining lifespan > 2 times maintenance cycle; Note (yellow): The predicted remaining lifespan is between 1 and 2 times the maintenance cycle; Warning (orange): Predicted remaining lifespan is between 0.5 and 1 times the maintenance cycle; Emergency (Red): Predicted remaining lifespan < 0.5 times maintenance cycle.

[0084] Among them: the maintenance cycle is the time interval for regular equipment maintenance, which varies depending on the equipment type and usage scenario; the classification of warning levels takes into account the importance and security of the product and can be adjusted according to the specific application scenario.

[0085] Specifically, for CNC machine tool spindle systems, the maintenance cycle is typically 3-6 months. Early warning signals can be sent to relevant personnel via system interface, email, or SMS to remind them to take timely maintenance measures. Furthermore, different warning levels correspond to different maintenance recommendations; for example, a yellow warning level suggests increasing monitoring frequency, an orange warning level suggests planning spare parts procurement, and a red warning level suggests scheduling machine shutdown for maintenance.

[0086] like Figure 6 As shown, the modules of the system of the present invention achieve collaborative work through data flow and a closed-loop feedback mechanism. Specifically, the service monitoring and early warning module 50, the 3D structure digital design module 30, the product manufacturing module 40, and the collaborative engine module 60 constitute a closed-loop feedback mechanism for: Identify degradation-sensitive points (i.e., product design structures or components that are most likely to experience performance degradation or induce failure during long-term operation) and pass them to the 3D structure digital design module 30 for structural optimization. Identify quality-sensitive points (i.e., key processes or parameters that have the greatest impact on the performance and degradation process of the final product during manufacturing) and pass them to the product manufacturing module 40 for manufacturing process optimization; The degradation mode (i.e., the specific laws and forms of product performance evolution over time, such as uniform wear and local spalling of bearings) and the failure mechanism analysis results (i.e., the fundamental physical and chemical causes of product performance degradation, such as fatigue fracture and wear failure) are passed to the collaborative engine module 60 for multidisciplinary (such as mechanical, material, and electrical) collaborative optimization.

[0087] In an embodiment of the present invention, for a CNC machine tool spindle system, topological data analysis reveals a correlation between early bearing wear and shaft design parameters. The system automatically feeds this information back to the three-dimensional structure digital design module 30, allowing designers to optimize the shaft's structural design and enhance bearing support rigidity, thereby increasing the bearing life of the new generation of products by 30%. Simultaneously, a correlation between bearing degradation and assembly accuracy is discovered. The system feeds this information back to the product manufacturing module 40, optimizing the assembly process and controlling the tolerances of key dimensions, thus improving assembly quality by 15%.

[0088] The following detailed description, using a specific embodiment, illustrates the application of the present invention's system in monitoring CNC machine tool spindle systems.

[0089] This invention relates to a system for monitoring the spindle system of a certain type of CNC machine tool. The system is equipped with a temperature sensor (measuring the temperature of the front and rear bearings), a vibration sensor (measuring acceleration and velocity in the X, Y, and Z directions), a sound sensor (collecting spectral characteristics), and a current sensor (measuring the spindle motor current). The sensor data is transmitted to the system server via an industrial Ethernet network and processed by the service monitoring and early warning module 50.

[0090] The multidimensional data topology representation unit 51 collects and processes sensor data to construct a point cloud dataset. Point cloud mapping employs three time scales: a 10-second window (microscopic scale), a 3-minute window (mesoscopic scale), and a 2-hour window (macroscopic scale). For vibration data, the sampling frequency is 1000Hz. For each window, time-domain features (mean, standard deviation, peak value, peak-to-peak value, skewness, kurtosis) and frequency-domain features (characteristic frequency amplitude, frequency band energy ratio) are extracted. Distance calculation uses weighted Mahalanobis distance, with vibration features weighted at 0.4, temperature features at 0.3, current features at 0.2, and sound features at 0.1.

[0091] Complex construction was performed using the Vietoris-Rips method, with an initial threshold. ,growth rate A 15-step complex sequence is constructed. By calculating the evolution of 0-dimensional, 1-dimensional, and 2-dimensional Betti numbers, persistent graphs and barcodes are generated to capture the topological features of the data.

[0092] The degradation pattern recognition unit 52 transforms the persistent graph into a comparable mathematical representation based on the persistent landscape representation, and uses second-order Wasserstein distance to calculate the similarity between states. Dimensionality reduction employs the MDS method to map the feature space to a two-dimensional plane, constructing... The coverage is such that adjacent regions overlap by 40%. The DBSCAN algorithm is used for segmentation and clustering, with parameters... Set to 20% of the local region diameter. The number of nodes was set to 4% of the total number of nodes. The final constructed topology network contained approximately 100 nodes and 150 edges. By analyzing the network structure, four main degradation modes were identified: early bearing wear, poor lubrication, shaft imbalance, and motor overload.

[0093] The remaining life prediction unit 53 constructed transition matrices for 10 representative states based on historical data. The health index mapping adopted a topological distance calculation method with reference health states, and the failure threshold... =0.25. 1000 possible evolutionary trajectories are generated using Monte Carlo simulation to calculate the remaining lifetime distribution. The system updates the prediction results daily and generates corresponding warning signals based on the predicted remaining lifetime.

[0094] In this embodiment, the system successfully predicted the failure of a CNC machine tool spindle bearing, issuing a yellow warning 45 days in advance, an orange warning 20 days in advance, and a red warning 7 days in advance. This gave maintenance personnel ample time to prepare spare parts and arrange planned downtime for maintenance, avoiding production interruptions caused by sudden failures and saving maintenance costs by approximately 25%.

[0095] Simultaneously, the system feeds back the correlation between bearing failures and shaft design parameters to the design department, prompting designers to optimize the shaft's structural design and enhance bearing support rigidity. It also feeds back the correlation between failures and assembly accuracy to the manufacturing department, optimizing the assembly process. These optimization measures have been applied in the new generation of products, resulting in a 30% increase in bearing life and a 15% improvement in assembly quality, fully demonstrating the closed-loop feedback value of this invention's system.

[0096] In summary, this invention provides a collaborative design system for the entire lifecycle of mechanical products based on digital twins. It introduces topological data analysis theory into product condition monitoring and lifespan prediction, constructing a data-driven collaborative optimization framework covering the entire process from design and manufacturing to service. The core innovation of the system lies in its service monitoring and early warning module. Through technological innovations at three levels—multi-dimensional data topology representation, degradation pattern recognition, and remaining lifespan prediction—it achieves accurate description and prediction of the complex degradation process of products. The system's closed-loop feedback mechanism feeds back data and knowledge acquired during the service phase to the design and manufacturing phases, promoting continuous product optimization and improving product reliability and lifespan. Examples demonstrate that this system has achieved significant results in monitoring CNC machine tool spindle systems, not only improving the accuracy of fault prediction but also providing valuable knowledge support for design and manufacturing optimization. This system provides a new technical path for the intelligent design, manufacturing, and service of mechanical products, promoting the transformation of digital twin technology from single-stage applications to full lifecycle collaboration, and has broad application prospects.

[0097] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A collaborative design system for the entire lifecycle of mechanical products based on digital twins, characterized in that, include: The solution design module is used to design functional solutions for the product. The preliminary design module is used to design the product's structural, electrical, and process parameters. The 3D structure digital design module is used to optimize the 3D structure of the product based on collaborative feedback information; The product manufacturing module is used to complete the manufacturing process of the designed scheme; The service monitoring and early warning module is used to acquire operational data of the product during use based on topological data analysis theory, construct a topological representation of the product's operating status, identify degradation patterns, and predict the product's remaining service life. The module includes a multi-dimensional data topology structure representation unit, a degradation pattern identification unit, and a remaining service life prediction unit. The multi-dimensional data topology structure representation unit maps product operational data to a high-dimensional feature space and extracts topological features. The degradation pattern identification unit constructs a topological network of the product's degradation state based on the topological features and identifies degradation patterns. The remaining service life prediction unit constructs a topological state space evolution model based on the degradation patterns and the topological network, predicts the product's remaining service life, and generates an early warning signal when the remaining service life is lower than expected. The collaboration engine module is used to obtain collaborative feedback information from different designers and product users and pass it to the 3D structure digital design module. The 3D collaborative design database module is used to store design schemes, structural parameters, product operation data, and operational degradation status. The data analysis module is used to analyze and process the data in the 3D collaborative design database module.

2. The collaborative design system for the entire lifecycle of mechanical products based on digital twins as described in claim 1, characterized in that, The multidimensional data topology representation unit includes: The data acquisition layer is used to collect multi-dimensional operational data of the product through a distributed sensor network. The preprocessing layer is used to clean the collected data, handle outliers, and synchronize the time. The point cloud mapping layer is used to map preprocessed multidimensional running data into a set of points in a high-dimensional space; The distance calculation layer is used to construct distance metric functions that adapt to the characteristics of mechanical products; Complex construction layer, used to construct Vietoris-Rips complex sequences based on distance thresholds; The homology computation layer is used to compute the topological features of complex sequences in different dimensions. The feature representation layer is used to generate persistent graphs and barcodes, quantifying the topological features of product operation data.

3. The collaborative design system for the entire lifecycle of mechanical products based on digital twins as described in claim 1, characterized in that, The degradation pattern recognition unit includes: The feature space construction layer is used to transform persistent graph features into a mathematical representation of the persistent landscape and establish the feature space. The similarity calculation layer is used to calculate the topological similarity between different running states; Dimensionality reduction mapping layer is used to map a high-dimensional feature space to a low-dimensional space; The overlay construction layer is used to build overlapping overlays on the dimensionality-reduced parameter space. The segmented clustering layer is used to perform clustering analysis on data points within each coverage area; The network construction layer is used to represent clustering results as nodes, establish connections between nodes that share data points, and form a degenerate pattern topology network. The structural analysis layer is used to extract key network topology and identify product degradation patterns.

4. The collaborative design system for the entire lifecycle of mechanical products based on digital twins as described in claim 3, characterized in that, The network construction layer represents the clustering results as nodes, and establishes connections between nodes that share data points to form a degradation mode topology network, specifically including: For coverage Each region in The data points are clustered to obtain a set of clusters. Each cluster In a network, a node is represented as a single node. If two nodes share data points, an edge is established between them, with the weight of the edge set as the proportion of the number of shared points to the total number of points. For two clusters and ,if Then, an edge is created between them, with the weight of the edge being: ； in, This indicates the number of points in the intersection of two clusters; This indicates the number of union points of two clusters; This represents the weight of the edge, with a value range of 1. ; The final constructed network is represented as ,in, It is a set of nodes, with each node corresponding to a cluster; It is a set of edges, representing the connection relationships between nodes; It is the edge weight function, which represents the connection strength.

5. The collaborative design system for the entire lifecycle of mechanical products based on digital twins according to claim 1, characterized in that, The remaining lifetime prediction unit includes: The historical trajectory construction layer is used to construct a set of degenerate trajectories in the topological state space based on historical data. The state-space modeling layer is used to construct a topological state-space evolution model and describe the state transition process. The health index mapping layer is used to define the health index function and map the topological state to the health index, quantifying the degree of degradation. The evolution simulation layer is used to simulate possible future degradation trajectories based on the current state and the topological state space evolution model; The lifetime distribution calculation layer is used to calculate the time distribution of reaching the failure threshold based on the simulated degradation trajectory. The prediction update layer is used to continuously update the parameters and prediction results of the topological state space evolution model and the health index function by combining real-time data. The early warning generation layer is used to generate tiered early warning signals based on the prediction results.

6. The collaborative design system for the entire lifecycle of mechanical products based on digital twins as described in claim 5, characterized in that, The state-space modeling layer uses a Markov chain model to describe the state transition process, and the constructed topological state-space evolution model is as follows: ； in, The elements of the state transition matrix represent the transition from state to state. Transition to state The probability of; It is observed from the state Transition to state The number of times; From state The total number of transitions to any state; To account for the impact of operating conditions on the transition probability, a conditional transition rule is introduced: ； in, Indicates operating conditions Below, from the state Transition to state The conditional probability; Indicates the state Under these conditions, the state is observed simultaneously. and conditions The joint conditional probability; Indicates the state Observation conditions The conditional probability.

7. The collaborative design system for the entire lifecycle of mechanical products based on digital twins as described in claim 5, characterized in that, The health index function of the health index mapping layer is expressed as follows: ； in, For state Health index, range of values ; This represents the current topology state. For reference regarding health status; The current state Compared with reference health status Topological distance between them; It is the normalization factor.

8. The collaborative design system for the entire lifecycle of mechanical products based on digital twins according to claim 5, characterized in that, The evolution simulation layer simulates possible future degradation trajectories based on the current state and the topological state space evolution model, specifically including: Given the current state and state transition matrix A simulated trajectory is generated through the following steps: initialization: ; For time Based on transition probability Randomly select the next state ; Return to simulation trajectory ; in, Indicates time The state; Indicates the current state; Indicates from state Transition to state The probability of; Indicates the number of simulation steps; Repeat the above process to generate multiple possible evolutionary trajectories, reflecting the uncertainty of future states.

9. The collaborative design system for the entire lifecycle of mechanical products based on digital twins according to claim 5, characterized in that, The lifetime distribution calculation layer calculates the time distribution for reaching the failure threshold based on the simulated degradation trajectory, specifically including: Set failure threshold When the health index The simulation is considered invalid if the health index first falls below the failure threshold. For each simulated trajectory, the first arrival time is calculated, which is the time when the health index first falls below the failure threshold. ； in, This indicates the time of first arrival of the trajectory, i.e., the time of failure. Indicates time status Corresponding health index; Indicates the failure threshold; The first arrival times of all trajectories form a distribution, the mean of which can be used as a point estimate of the remaining lifetime, and the quantiles form a confidence interval. The point estimate of the remaining lifetime is as follows: ； in, Indicates the predicted remaining useful life; Indicates the number of simulated trajectories; Indicates the first The first arrival time of each trajectory.

10. The collaborative design system for the entire lifecycle of mechanical products based on digital twins according to claim 1, characterized in that, The service monitoring and early warning module, together with the 3D structure digital design module, the product manufacturing module, and the collaborative engine module, forms a closed-loop feedback mechanism for: The identified degradation-sensitive points are transmitted to the three-dimensional structure digital design module for structural optimization; The identified quality-sensitive points are transmitted to the product manufacturing module for manufacturing process optimization. The degradation mode and failure mechanism analysis results are passed to the collaborative engine module for multidisciplinary collaborative optimization.

Images