Digital twin method and system for analyzing micro stress field of polymer blend system

Abstract

This invention discloses a digital twin method and system for analyzing the microscopic stress field of polymer blend systems, belonging to the field of digital twin technology. The method includes: acquiring response characteristics such as echo amplitude, propagation delay, spectral attenuation, phase shift, and scattering intensity within the blend system using an ultrasonic array; constructing a non-uniform mixing state field in three-dimensional space; and extracting spatial inhomogeneity features to generate structural cores. Each structural core is further mapped to a dedicated intelligent agent with a heterogeneous mechanical strategy, establishing a spatial competitive dominance relationship, and driving multi-agent adversarial evolution by combining polymer type, macroscopic proportion, and mixing history information. Each agent outputs adversarial actions in space and collectively acts on the local stress state, gradually reducing the local strain energy through iterative updates. When the system evolution reaches Nash equilibrium, a microscopic stress field of the blend system is generated, thereby enabling dynamic analysis of internal stress formation during cooling and solidification.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and in particular to a digital twin method and system for analyzing the microscopic stress field of polymer blend systems. Background Technology

[0002] Polymer blends, due to their ability to combine the performance advantages of multiple polymers, have been widely used in automotive parts, electronic packaging, medical devices, and functional films. The molding process of these systems typically involves complex processes such as multi-component flow, phase interface evolution, temperature gradient transfer, and cooling curing shrinkage, which easily leads to the formation of micro-stress distributions with significant spatial differences within the material. These micro-stress fields not only affect the dimensional stability, interfacial bonding strength, and mechanical durability of the parts, but are also closely related to subsequent problems such as warpage, cracking, and fatigue failure. Therefore, accurate characterization and analysis of the micro-stress field within polymer blend systems are of great significance.

[0003] In existing technologies, the study of internal stress in polymers largely relies on finite element simulation, polarized light detection, or local experimental measurements. However, these methods often struggle to simultaneously address the identification of internal three-dimensional structures, the characterization of complex phase distributions, and the dynamic evolution analysis of micro-stress. Especially for blends with non-uniform mixing characteristics and complex interface topologies, traditional methods still fall short in terms of resolution, real-time performance, and the ability to characterize the response of hidden internal structures. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0006] In a first aspect, the present invention provides a digital twin method for analyzing the microscopic stress field of a polymer blend system, which includes obtaining the types and macroscopic proportions of each polymer in the polymer blend system, and obtaining the response state inside the system by combining ultrasonic detection, and mapping and generating a non-uniform mixing state field characterizing the internal structure in three-dimensional space.

[0007] Based on the non-uniformity characteristics in the non-uniform mixed state field, multiple cores corresponding to local regions are adaptively generated;

[0008] Each core is treated as an intelligent agent, and a field is constructed in which there is a competitive dominance relationship over spatial location;

[0009] Based on the historical characteristics of the liquid stage of the polymer blend system, the various intelligent agents are driven to undergo adversarial evolution in the field;

[0010] When evolving to the Nash equilibrium state, a micro-stress field is generated in the polymer blend system.

[0011] As a preferred embodiment of the digital twin method for analyzing the microscopic stress field of the polymer blend system described in this invention, the response state is obtained by means of the echo amplitude, propagation delay, spectral attenuation, phase shift, and scattering intensity characteristics when the ultrasonic array penetrates the polymer blend system.

[0012] The non-uniform mixing state field includes mapping the acquired ultrasonic testing data according to the corresponding spatial coordinates in the space of the blended system to form the distribution result of the ultrasonic testing data in three-dimensional space.

[0013] As a preferred embodiment of the digital twin method for analyzing the microscopic stress field of the polymer blend system described in this invention, the non-uniformity features include multidimensional features representing the non-uniformity by measuring the changes in echo amplitude, propagation delay difference, spectral attenuation, phase shift, and scattering intensity of the ultrasonic detection data between adjacent spatial locations, thereby determining the local variation intensity of different spatial regions within the blend system.

[0014] As a preferred embodiment of the digital twin method for analyzing the micro-stress field of the polymer blend system described in this invention, the core of the adaptive generation of multiple corresponding local regions includes: adaptively clustering the spatial positions in three-dimensional space according to the multi-dimensional features at different coordinate positions to form multiple feature clustering regions.

[0015] In the three-dimensional space of the blend system, the results of the mixing and curing processes of all polymers are simplified:

[0016] Suppose that, starting from each cluster core, feature diffusion is performed to the surrounding spatial locations, so that each cluster core has a diffusion effect on the surrounding spatial locations to varying degrees;

[0017] The adaptive clustering includes constructing a local structure tensor by calculating the partial derivatives of the multidimensional features in three-dimensional space; performing eigenvalue decomposition on the local structure tensor to extract feature values ​​and two directions of the features, and generating internal texture features that characterize the continuity and bidirectional orientation of positional features.

[0018] By detecting the different locations of the intelligent agent in space:

[0019] Step 1: During the process of the detection agent traversing in three-dimensional space based on the internal texture features, the traversal optimization parameters of the agent's current spatial coordinates in the three dimensions are calculated and monitored in real time:

[0020] Local centroid distance: The relative distance between the current position and the global geometric centroid of the continuous texture in three-dimensional space;

[0021] Local attachment distance: The shortest spatial distance from the current position to the nearest texture skeleton or high-density feature trajectory;

[0022] Pointing convergence: The topological position of the current location in the texture bidirectional pointing vector field, calculated as the endpoint or convergence degree of the texture vector flow direction;

[0023] Step 2: Using a single-dimensional extreme value triggering mechanism, when the agent identifies that any single dimension parameter among the above three dimensions has reached a local maximum extreme value in the agent's traversal neighborhood, the probing agent immediately becomes a clustering core.

[0024] The single-dimensional extreme value triggering mechanism allows multiple heterogeneous clustering cores to be adaptively generated within the same complex texture region due to extreme value triggering in different dimensions.

[0025] Among them, for any continuous texture feature in space, it is allowed to simultaneously penetrate and belong to multiple cluster families governed by different clustering cores in the topological extension.

[0026] As a preferred embodiment of the digital twin method for analyzing the micro-stress field of the polymer blend system described in this invention, wherein: the field having a competitive dominance relationship over spatial location includes treating each of the cluster cores as an independent intelligent agent;

[0027] Based on the influence diffusion of each agent to the surrounding spatial location, which decreases with spatial distance and multidimensional feature similarity;

[0028] In the process of the spread of influence, it is affected by the size and scope of the race:

[0029] In the adaptive clustering process, the cumulative topological length of all continuous texture features affected by a single cluster core and classified into that cluster family is extracted as the race size corresponding to the agent; and the race size is positively mapped to the initial reference radiation intensity for the agent to perform influence diffusion through a preset mapping function 1.

[0030] Extract the global topological envelope boundary occupied by all continuous texture features affected by a single cluster core in three-dimensional space, which is taken as the range of the agent; and map the spatial span of the range to the spatial attenuation coefficient of the agent's influence through a preset mapping function 2.

[0031] At any spatial coordinate position, after adjusting the spatial distance and the spatial attenuation coefficient, each intelligent agent jointly governs the stress calculation results at the coordinate position.

[0032] As a preferred embodiment of the digital twin method for analyzing the micro-stress field of the polymer blend system described in this invention, the adversarial evolution includes: Step 1: For the three dimensions of local centroid distance, local attachment distance and directional convergence degree in the single-dimensional extreme value triggering mechanism, three exclusive agent models with heterogeneous mechanical response strategies are independently pre-trained in combination with the physical force characteristics of the corresponding local topological morphology.

[0033] When determining the clustering core through the single-dimensional extreme value triggering mechanism, the single-dimensional parameter type that triggered the generation of the clustering core is traced back, and the clustering core is mapped to the corresponding dimension's dedicated intelligent agent;

[0034] Step 2: Inject the types and macroscopic proportions of each polymer in the polymer blend system, as well as the vector sequence of the mixing action, as features into each dedicated intelligent agent to establish the initial physical properties of each dedicated intelligent agent.

[0035] Each dedicated intelligent agent, based on its own pre-training and initial physical properties, aims to minimize the local strain energy within its own dominance space and outputs counteracting actions of its internal tension tensor.

[0036] Step 3: At any coordinate position in the three-dimensional space, using the common dominance ratio of each dedicated agent after adjustment by spatial distance and spatial attenuation coefficient, tensor weighted fusion is performed on the adversarial actions output by multiple dedicated agents to obtain stress calculation results.

[0037] The stress calculation results are used as state penalty information and transmitted in reverse along the dominance relationship of the agents to each dedicated agent that has an impact, so that each dedicated agent can continuously update its own adversarial strategy and state tensor based on the state penalty information.

[0038] As a preferred embodiment of the digital twin method for analyzing the micro-stress field of the polymer blend system described in this invention, the Nash equilibrium state includes generating the local strain energy corresponding to each coordinate position based on the stress calculation results at each coordinate position and combined with the iterative process of adversarial evolution.

[0039] During the adversarial evolution process, each dedicated intelligent agent continuously updates its own state tensor;

[0040] When the continuous updating of the state tensor fails to reduce the local strain energy at any coordinate position in the three-dimensional space or reach the maximum number of iterations, the adversarial evolution is determined to have reached a Nash equilibrium state.

[0041] Secondly, the present invention provides a digital twin system for analyzing the micro-stress field of a polymer blend system, comprising: a data acquisition unit, which acquires the types and macroscopic proportions of each polymer in the polymer blend system, and acquires the response state inside the system by combining ultrasonic detection, and maps and generates a non-uniform mixing state field characterizing the internal structure in three-dimensional space.

[0042] The simulation unit adaptively generates multiple cores corresponding to local regions based on the non-uniformity characteristics of the non-uniform mixed state field; each core is used as an intelligent agent, and a field with competitive dominance over spatial position is constructed.

[0043] The evolution unit, combining the historical characteristics of the liquid stage of the polymer blend system, drives each agent to evolve in the field in an adversarial manner; when evolving to the Nash equilibrium state, it generates the micro-stress field of the polymer blend system.

[0044] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the digital twin method for analyzing the microscopic stress field of polymer blend systems as described in the first aspect of the present invention.

[0045] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the digital twin method for analyzing the microscopic stress field of polymer blend systems as described in the first aspect of the present invention.

[0046] The beneficial effects of this invention are as follows: This invention enables refined characterization and stable solution of the micro-stress field under complex phase distribution and non-uniform mixing conditions within polymer blend systems, improving the ability to identify internal stress concentrations, local instability, and potential defect locations. Compared to traditional analysis methods relying on regular meshes or single detection results, this invention more fully characterizes the continuous texture, local topological differences, and multi-core coupling relationships within the material, more realistically reflecting the internal stress formation law of the blend system during cooling and solidification. Simultaneously, this invention achieves dynamic updating and convergence determination of stress distribution through a multi-agent adversarial evolution mechanism, which is beneficial to improving the stability, adaptability, and physical consistency of the analysis results. This method can provide effective support for the formulation design, molding process optimization, structural reliability assessment, and early failure warning of polymer blend materials. Attached Figure Description

[0047] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Fig. 1 A flowchart of a digital twin method for analyzing the microscopic stress field of polymer blends.

[0049] Fig. 2 A flowchart illustrating the clustering core generation process of a digital twin method for analyzing the microscopic stress field of polymer blends.

[0050] Fig. 3 A diagram of a computer setup for a digital twin method of analyzing the microscopic stress field in polymer blends. Detailed Implementation

[0051] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0052] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0053] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0054] Reference Figs. 1-3 As one embodiment of the present invention, this embodiment provides a digital twin method for analyzing the microscopic stress field of a polymer blend system, comprising the following steps:

[0055] S1: Obtain the types and macroscopic proportions of each polymer in the polymer blend system, and combine ultrasonic detection to obtain the response state inside the system, and map and generate a non-uniform mixed state field characterizing the internal structure in three-dimensional space.

[0056] By utilizing the differences in acoustic impedance, interface scattering effects, and structural inhomogeneity of different polymer components during acoustic propagation, the spatial structure within the blend system is non-destructively sensed, thereby obtaining raw detection data that reflects the mixing state and local structural differences within the material. Furthermore, the response state is obtained through an ultrasonic array penetrating the polymer blend system, capturing the echo amplitude, propagation delay, spectral attenuation, phase shift, and scattering intensity characteristics. The inhomogeneous mixing state field involves mapping the acquired ultrasonic detection data according to corresponding spatial coordinates within the blend system, forming a three-dimensional spatial distribution of the ultrasonic detection data. This transforms the originally discrete detection signal into a three-dimensional feature field with spatial continuity and structural expressive capabilities, thus forming an inhomogeneous mixing state field that reflects the degree of mixing, interface distribution, and local structural differences within the blend system.

[0057] The degree of change in ultrasonic signals during propagation at different spatial locations reflects the structural inhomogeneity within a polymer blend system caused by differences in component distribution, interfacial structure variations, and varying degrees of local mixing. By analyzing the changes in echo amplitude, propagation delay, spectral attenuation, phase shift, and scattering intensity between adjacent spatial locations using the ultrasonic detection data, the multidimensional characteristics of inhomogeneity are represented, and the intensity of local variations in different spatial regions within the blend system is determined.

[0058] It is important to understand that by jointly analyzing the changes in multiple acoustic characteristics between adjacent spatial locations, the spatial differences in the internal structure of materials can be characterized from multiple physical dimensions. This allows the non-uniformity to no longer rely solely on a single detection index, but rather to reflect the local structural changes within the blend system more comprehensively through a multi-feature collaborative characterization approach.

[0059] Based on this, the intensity of local changes in different spatial regions is determined through the multidimensional features. The purpose is to identify regions where the internal structure of the blend system changes significantly, thereby providing a quantitative basis for subsequent spatial feature clustering, core region generation, and local structural evolution analysis. This enables subsequent analysis to more accurately locate possible structural difference regions and interface distribution regions within the material.

[0060] S2: Based on the non-uniformity characteristics in the non-uniform mixed state field, multiple cores corresponding to local regions are adaptively generated.

[0061] Specifically, based on the multidimensional features at different coordinate positions, adaptive clustering is performed on the spatial positions in three-dimensional space to form multiple feature clustering regions.

[0062] In the three-dimensional space of the blend system, the results of the mixing and curing processes of all polymers are simplified:

[0063] Suppose that, starting from each cluster core, feature diffusion is performed to the surrounding spatial locations, so that each cluster core has a diffusion effect on the surrounding spatial locations to varying degrees.

[0064] This approach transforms the highly complex and continuously changing microstructure of real materials into several representative structural cores and their spatial influence relationships, thereby describing the spatial characteristic distribution within the blend system in a structured manner. Because polymer blend systems form numerous microstructures of varying scales, complex morphologies, and continuous changes during mixing and curing, directly modeling all spatial structures in detail would not only be computationally extremely complex but also difficult to extract key regions representing the overall structural characteristics. Therefore, by abstracting and simplifying the complex structural results and representing them as a finite number of cluster cores and their spatial influence ranges, the internal structure of the system can be effectively expressed while preserving the main structural features.

[0065] This simplification significantly reduces the computational complexity of 3D spatial structure modeling, making subsequent spatial relationship analysis and evolution calculations more controllable, while highlighting key structural regions within the material that dominate the overall mechanical behavior. Furthermore, this approach transforms continuously distributed complex structures into core nodes with clearly defined spatial relationships and their diffused influences, making it easier to establish connections between spatial locations. This provides a clear structural foundation for subsequently constructing competitive dominance relationships and performing adversarial evolution analysis.

[0066] By analyzing the spatial distribution of multidimensional structural features within a material, representative structural feature convergence locations or dominant structural locations are identified in three-dimensional space, thereby determining cluster core regions that can characterize local structural states. Unlike traditional clustering methods based on feature similarity, this method analyzes the local structural morphology in the spatial feature distribution, enabling the cluster cores to reflect key nodes or structural convergence regions in the changes of the material's internal structure.

[0067] Furthermore, the adaptive clustering includes constructing a local structure tensor by calculating the partial derivatives of the multidimensional features in three-dimensional space. During the calculation, the rate of change of each feature along the three coordinate directions at the current spatial position can be obtained using a three-dimensional gradient calculation algorithm. The three-dimensional gradient calculation algorithm can be any one of the following: central difference algorithm, Sobel operator, Scharr operator, three-dimensional Gaussian derivative filtering algorithm, or a combination thereof. Through this step, the trend of change of each feature along different directions at the current position can be obtained, thereby reflecting the fluctuations, inflections, and extensions of the feature in the local space.

[0068] After obtaining the spatial gradients of each feature dimension, a corresponding directional coupling descriptor can be constructed based on the gradient results of each feature dimension. Specifically, this involves jointly statistically analyzing the changes of the same feature dimension in different coordinate directions. This allows us to describe not only the strength of changes in a particular direction but also whether there are cooperative changes, coupled changes, or dominant extension directions between different directions. In this way, each feature dimension can form a matrix-based description of the local directional structure.

[0069] Subsequently, the matrix-based descriptions of the features from each dimension are fused to construct a local structure tensor for the current location. During fusion, weighted summation algorithms, adaptive weight fusion algorithms, or feature-based reliability-based fusion models can be employed. For example, different features can be assigned different contribution ratios based on their signal-to-noise ratio, variance stability, information entropy, local contrast, or preset empirical weights, thereby preventing excessive noise from any single feature from dominating the results. Through this process, the originally dispersed multidimensional ultrasonic features are uniformly mapped into a tensor object capable of representing the local structural features of the current spatial location.

[0070] Furthermore, to improve the stability of the local structure tensor, a neighborhood window can be introduced around the current spatial location to smoothly aggregate the local structure tensors at multiple locations within the neighborhood. The aggregation method can employ Gaussian weighted smoothing, sliding window averaging, local anisotropic smoothing, or robust neighborhood statistical methods. By smoothing the results within the neighborhood, the instability caused by ultrasonic detection noise, discrete point fluctuations, and sudden local anomalies can be effectively suppressed, making the final constructed local structure tensor more representative of the true internal structural state near the current location.

[0071] Therefore, the construction of the local structure tensor is essentially a process of "multi-dimensional feature extraction—spatial gradient calculation—directional relationship coupling—multi-feature fusion—neighborhood stabilization processing." Through this process, the original ultrasound detection data can be transformed into a local structure representation with clear spatial directional significance.

[0072] Next, eigenvalue decomposition is performed on the local structure tensor to extract eigenvalues ​​and their two directions, generating internal texture features that characterize the continuity and bidirectional orientation of the positional features. Specifically, since the local structure tensor is essentially a symmetric matrix describing the distribution of local spatial directions, it can be processed using a symmetric matrix eigenvalue decomposition algorithm. This eigenvalue decomposition algorithm can employ QR decomposition, the Jacobi iterative algorithm, a Householder transformation combined with QR eigenvalue solving algorithm, or directly call the symmetric matrix eigenvalue decomposition module in a mature linear algebra solver library. By performing eigenvalue decomposition, the local structure tensor can be decomposed into several eigenvalues ​​and their corresponding eigenvectors.

[0073] Each eigenvalue represents the intensity of change or saliency of the current local structure in a certain principal direction; the corresponding eigenvector represents the specific direction of that principal direction in three-dimensional space. In other words, through eigenvalue decomposition, several of the most representative dominant directions can be separated from the multi-directional information originally mixed in the tensor.

[0074] In this invention, after intrinsic decomposition, the resulting eigenvalues ​​are sorted according to their magnitude, and the two eigenvectors corresponding to the two largest eigenvalues ​​are extracted as the two dominant directions of the internal texture at the current location. The reason for extracting two directions is that this invention focuses not on a single linear extension, but on the bidirectional orientation of local textures in three-dimensional space; that is, the internal structure around a certain location may extend, connect, converge, or bifurcate along two significant directions. By extracting the two dominant directions, the continuous distribution trend of the material's internal texture in space can be described more accurately. Simultaneously, the relative relationships between eigenvalues ​​can also be used to determine the organizational state of the local structure near the current location. For example, when both the first and second dominant eigenvalues ​​are large and the third direction is weak, it indicates a significant bidirectional extension structure around the location; when only one dominant eigenvalue is prominent, it indicates that the location is closer to a unidirectional dominant texture structure; when the differences between multiple eigenvalues ​​are small, it indicates that the local directionality at that location is not obvious, and it is closer to a disordered or scattered structure. Therefore, intrinsic decomposition can not only output directions but also provide quantitative evidence regarding the continuity, directionality, and organizational degree of local structures.

[0075] To improve the robustness of intrinsic decomposition results in complex spaces, practical implementations can combine eigenvalue sorting algorithms, orientation consistency constraint algorithms, and neighborhood principal orientation continuous correction algorithms. For example, between adjacent spatial locations, the obtained principal orientation can be continuously corrected to avoid abrupt changes in the principal orientation in space due to local noise or sign flipping. Furthermore, neighborhood voting mechanisms or orientation smoothing mechanisms can be used to make the principal orientation field more continuous and stable, facilitating subsequent traversal of the texture field by the detection agent.

[0076] It's important to understand that in polymer physics, polymer chains are highly entangled and interpenetrating. A long chain (a texture) meanders through space; its head may be controlled by the A-phase region (A core), and its tail may be controlled by the B-phase region (B core). If it is forcibly assigned to a local core, the stress transmission will be artificially severed at the boundary.

[0077] By detecting the different locations of the intelligent agent in space:

[0078] Step 1: During the process of the detection agent traversing in three-dimensional space based on the internal texture features, the traversal optimization parameters of the agent's current spatial coordinates in the three dimensions are calculated and monitored in real time:

[0079] Local centroid distance: The relative distance between the current position and the global geometric centroid of the continuous texture in three-dimensional space.

[0080] Local attachment distance: The shortest spatial distance from the current position to the nearest texture skeleton or high-density feature trajectory.

[0081] Pointing convergence: The topological position of the current location within the bidirectional directional vector field of the texture, calculated as either the endpoint or convergence degree of the texture vector flow direction. Specifically, pointing convergence characterizes the spatial role of the current location within the texture vector flow structure; it can represent either the endpoint of the texture vector flow direction or the convergence position of the texture vectors. When the current location is at the end of the texture vector flow path, the texture vectors in the neighborhood no longer maintain stable extension after reaching this location, exhibiting significant directional decay or reduced continuity. In this case, the location can be identified as the endpoint of the texture flow direction, and its pointing convergence is calculated by analyzing the termination degree of the neighborhood vectors or changes in vector density. On the other hand, when multiple texture vectors from different spatial directions exist around the current location, and these vectors gradually point to the same location during spatial extension, it indicates that this location is a convergence region of the texture vector flow direction. In this case, the convergence strength can be determined by statistically analyzing the number of vectors pointing to this location in the neighborhood, directional consistency, or flow density, and the corresponding pointing convergence is calculated accordingly. By simultaneously identifying the endpoints and convergence points of the texture flow, the organization of the internal texture structure in three-dimensional space can be more comprehensively characterized, enabling the detection agent to accurately identify key nodes or termination regions in the texture structure, thereby providing a reliable basis for the subsequent generation of clustering cores based on the extreme value triggering mechanism.

[0082] Step 2: Using a single-dimensional extreme value triggering mechanism, when the agent identifies any single-dimensional parameter among the above three dimensions as reaching a local maximum extreme value (i.e., the topological boundary farthest / closest to the centroid, the skeleton anchor point absolutely closest to the texture, or the physical end of the texture pointing to a highly convergent point), the probing agent immediately generates a clustering core.

[0083] The single-dimensional extremum triggering mechanism allows multiple heterogeneous clustering cores to be adaptively generated within the same complex texture region due to extremum triggering in different dimensions.

[0084] Specifically, for any continuous texture feature in space, it is allowed to simultaneously penetrate and belong to multiple cluster families governed by different cluster cores in its topological extension. By recording the spatial distribution anchor points of the texture feature in each cluster family, the influence of each cluster core can cross the local space along the texture network under its jurisdiction, thereby realizing the mechanical response and stress transmission influence on the global space of the blend system.

[0085] The structural textures within polymer blends typically exhibit significant complexity and multi-scale characteristics, with different types of key structural locations often dominated by distinct spatial features. For example, some regions may represent the spatial boundaries of the overall structure, others may function as stable anchor points for the texture skeleton or structural channels, while still others may appear as nodes formed by the convergence of multiple structural directions. Therefore, by employing a single-dimensional extremum triggering method, different types of structural features can be independently identified as potential structural cores, thus avoiding structural identification biases caused by relying on a single criterion and making the generation of cores more consistent with the true spatial organization of the material.

[0086] Building upon this, allowing multiple cores with different properties within the same structural region can enrich the expression of spatial structures. Since the structural network in a hybrid system is often composed of nodes with various properties, generating only a single center can easily overlook key locations with different physical meanings within the structure. The existence of multiple types of cores allows for the simultaneous characterization of various spatial roles, such as structural boundaries, skeleton nodes, and directional convergence nodes, thus making the expression of spatial structures more complete.

[0087] Simultaneously, allowing continuous textures to connect multiple cores topologically helps maintain the overall connectivity of the material's internal structural network. In real material structures, structural channels or texture paths often span multiple local regions and connect multiple key nodes. Forcing them into a single core would disrupt the true interrelationships between structures. Therefore, by recording the distribution of textures within the effective range of different cores, a spatial relationship network between cores can be established, enabling each core to not only represent a local region but also influence a larger area through the texture network.

[0088] S3: Treat each core as an intelligent agent and construct a field in which there is a competitive dominance relationship over spatial location.

[0089] Each agent performs influence diffusion to its surrounding spatial location, which decays with spatial distance and multidimensional feature similarity. This influence diffusion process is affected by race size and range.

[0090] In the adaptive clustering process, the cumulative topological length of all continuous texture features affected by a single cluster core and classified into that cluster family is extracted as the race size corresponding to the agent; and the race size is forward mapped to the initial reference radiation intensity for the agent to perform influence diffusion through a preset mapping function 1.

[0091] Extract the global topological envelope boundary that all continuous texture features affected by a single cluster core occupy in three-dimensional space, which is taken as the range of the agent; and map the spatial span of the range to the spatial decay coefficient of the agent's influence through a preset mapping function 2.

[0092] At any spatial coordinate position, after adjusting the spatial distance and the spatial attenuation coefficient, each intelligent agent jointly governs the stress calculation results at the coordinate position.

[0093] Specifically, in this embodiment, mapping function 1 is used to convert the race size corresponding to the clustering core into the initial baseline radiation intensity when the core performs influence diffusion in space. Specifically, the race size reflects the scale of the continuous texture network controlled by a core, that is, the overall topological extension of the structural channels associated with that core. Therefore, in this embodiment, a monotonically increasing mapping function can be used, so that the larger the race size, the higher the initial radiation intensity of the corresponding agent.

[0094] For example, one implementation can employ a normalized linear mapping approach: first, the race size of all cores is normalized, and then it is mapped proportionally to the initial radiation intensity of the corresponding agent. This allows cores with larger texture network sizes to have a stronger initial influence in the competitive dominance field. This mapping method enables cores with larger structural sizes to play a more significant role in the spatial influence diffusion process, thus better reflecting the dominant role of the material's internal structural network in overall mechanical behavior.

[0095] In other alternative embodiments, mapping function 1 can also take other forms, such as logarithmic function mapping, exponential function mapping, or sigmoid function mapping. For example, when there are large differences in structural scale, a logarithmic function can be used to compress the size of the population to avoid a few large-scale cores from exerting excessive dominance over the overall space; or an exponential function can be used to enhance the influence weight of large-scale structural cores to highlight the dominant role of the main structural network. Therefore, mapping function 1 can be flexibly selected according to different material structural characteristics or analytical needs.

[0096] Mapping function 2 is used to transform the spatial extent corresponding to the cluster core into a spatial attenuation coefficient during the diffusion of the core's influence. Here, the spatial extent reflects the overall scale covered by the texture network associated with a core in three-dimensional space, that is, the distribution span of the core's influence structure in space.

[0097] In this embodiment, an inverse proportional mapping method can be adopted, meaning that the larger the spatial range, the smaller the corresponding spatial attenuation coefficient, making the influence of the core decay more slowly in space, thus covering a wider spatial area. In this way, a core controlling a larger spatial structural range can maintain a certain influence over a wider area, while a small core controlling a localized structure is more likely to form a localized influence, thereby creating a multi-scale structural influence relationship.

[0098] In other alternative embodiments, the mapping function 2 can also employ different attenuation models. For example, a linear mapping method can be used to maintain a linear relationship between the spatial span and the attenuation coefficient; an exponential attenuation model or a Gaussian attenuation model can also be used to make the attenuation of the core influence in space smoother and more continuous. Furthermore, in some embodiments, the spatial attenuation coefficient can be adaptively determined through a learning model or a statistical model to better conform to the spatial action law of a specific material structure.

[0099] By observing the varying strengths and ranges of influence of different structural cores in space, the synergistic and competitive relationships among the internal structures of a material can be more realistically reflected. At any spatial location, the combined influence of multiple cores means that the stress calculation at that location is simultaneously affected by multiple structural nodes, resulting in a mechanical response expression that more closely matches the actual structural characteristics of the material.

[0100] S4: Combining the historical characteristics of the liquid stage of the polymer blend system, drive each intelligent agent to undergo adversarial evolution in the field.

[0101] Step 1: For the three dimensions of local centroid distance, local attachment distance and pointing convergence degree in the single-dimensional extreme value triggering mechanism, three dedicated agent models with heterogeneous mechanical response strategies are independently pre-trained by combining the physical force characteristics of the corresponding local topology.

[0102] When determining the clustering core through the single-dimensional extreme value triggering mechanism, the single-dimensional parameter type that triggered the generation of the clustering core is traced back, and the clustering core is mapped to the corresponding dimension's dedicated intelligent agent.

[0103] Step 2: Inject the types and macroscopic proportions of each polymer in the polymer blend system, as well as the vector sequence of the mixing action, as features into each dedicated intelligent agent to establish the initial physical properties of each dedicated intelligent agent.

[0104] Each dedicated intelligent agent, based on its own pre-training and initial physical properties, aims to minimize the local strain energy within its own dominance space and outputs counteracting actions of its internal tension tensor.

[0105] Step 3: At any coordinate position in the three-dimensional space, the adversarial actions output by multiple dedicated agents are fused using tensor weighting based on the common dominance ratio of each dedicated agent after adjustment by spatial distance and spatial attenuation coefficient, to obtain the stress calculation result.

[0106] The stress calculation results are used as state penalty information and transmitted in reverse along the dominance relationship of the agents to each dedicated agent that has an impact. Each dedicated agent continuously updates its own adversarial strategy and state tensor based on the state penalty information, thereby completing the adversarial evolution of the micro-stress field in the pure phase change cooling process without external mechanical load.

[0107] In fact, the process of adversarial evolution is a process in which the local stress state formed by the joint action of multiple dedicated intelligent agents at each location in three-dimensional space is continuously updated, and the local strain energy corresponding to that location tends to decrease and stabilize as a whole during the iterative process.

[0108] It should be noted that the vector sequence of mixing actions is a set of time-series features used to characterize the dynamic historical information of the mixing and stirring process of the polymer blend system before entering the cooling and solidification stage. This sequence is used to record the operational behavior applied to the material system by the mixing equipment or stirring device in the liquid stage and its state changing over time, thereby reflecting the flow state, shearing action, and changes in mixing intensity experienced during the formation of the internal structure of the blend system.

[0109] Specifically, during the mixing process of a blend system, the mixing equipment typically generates a series of continuously changing operational actions, such as changes in stirring speed, adjustment of stirring direction, changes in shear strength, mixing duration, and material feeding sequence. These operational behaviors directly affect the spatial dispersion state of different polymer components, the interface formation process, and the generation mode of local structural networks. Therefore, this invention records these mixing operations in chronological order and encodes the corresponding operational state at each moment into an action vector, thereby forming a time-varying sequence of mixing action vectors.

[0110] In one implementation, each action vector can contain several parameters describing the mixing state, such as stirring speed, stirring direction, instantaneous shear strength, mixing energy input, material flow rate, or mixing time step. By arranging these action vectors in chronological order, an action sequence describing the entire mixing process can be formed. This sequence reflects the flow evolution path experienced by the material before cooling, enabling the system to capture the influence of different mixing strategies on the formation of the material's internal structure.

[0111] By introducing a vector sequence of mixed actions, the processing history information of the blending system's formation stage is incorporated into subsequent stress evolution calculations. This allows each dedicated agent to not only rely on the current spatial structure during initialization but also to consider the material's formation process in the liquid stage, thus more accurately simulating the stress formation mechanism of different structural regions during cooling and solidification. In this way, the calculation results of the micro-stress field can better reflect the impact of the actual processing on the material's internal structure and stress distribution.

[0112] It's important to understand that different structural locations within a material often play different stress roles—for example, structural boundary regions, texture skeleton regions, and texture convergence nodes have different stress transmission and strain release mechanisms. Therefore, by configuring dedicated agents with differentiated strategies for clustering cores from different sources, the system can exhibit mechanical response characteristics that match the local topology during evolution. Simultaneously, by injecting material composition ratios and mixing history information into each agent, it acquires physical properties reflecting the material formation process in its initial state. This allows stress evolution to not only depend on the current spatial structure but also reflect the influence of processing history on the structural state. In three-dimensional space, each agent acts collectively at the same location according to its own influence range. Through continuous iterative strategy updates, the local stress state is gradually adjusted under the competition and synergy of multiple agents, ultimately leading to a reduction and stabilization of the overall local strain energy at each location in space. This enables dynamic simulation of the micro-stress formation process of the blend system under phase change cooling without external loads.

[0113] S5: When evolving to the Nash equilibrium state, a micro-stress field is generated in the polymer blend system.

[0114] In the aforementioned steps, each agent continuously outputs adversarial actions and participates in updating the spatial stress state according to its own strategy. Without a stable criterion, the evolutionary process may continue indefinitely, making it difficult to determine the final outcome. Therefore, by monitoring the changes in local strain energy at various locations in three-dimensional space, when the system can no longer further reduce the overall strain energy after continuous iteration, or when it reaches a preset iteration limit, the competitive relationship between the agents can be considered to have stabilized. At this point, any single agent adjusting its strategy cannot significantly change the overall energy state, thus forming a stable strategy combination state. This state corresponds to Nash equilibrium in a multi-agent system, indicating that the mechanical interactions between the internal structures of the material have reached a relative balance. Therefore, the current spatial stress distribution can be taken as the final microscopic stress field result.

[0115] The Nash equilibrium state includes generating the local strain energy corresponding to each coordinate position based on the stress calculation results at each coordinate position and in conjunction with the iterative process of adversarial evolution. During the adversarial evolution process, each dedicated agent continuously updates its own state tensor.

[0116] When the continuous updating of the state tensor fails to reduce the local strain energy at any coordinate position in the three-dimensional space or reach the maximum number of iterations, the adversarial evolution is determined to have reached a Nash equilibrium state.

[0117] It's important to note that in each adversarial evolution iteration, the stress calculation results for each spatial location are obtained by fusing the adversarial actions output by multiple agents. Subsequently, by combining the local elastic or viscoelastic response relationship of the material during the cooling stage, the stress state is correlated with the corresponding local strain state to obtain the local strain energy at that location. Generally speaking, local strain energy can be understood as the elastic deformation energy stored in the material at that location due to internal stress, and its magnitude is related to the stress intensity and the corresponding strain degree. In implementation, this can be achieved by calculating the energy relationship between the current stress tensor and the local strain tensor, or by converting the stress state into the corresponding strain energy density through a material constitutive model, and then multiplying it by the volume of the spatial unit to obtain the local strain energy. By updating and monitoring the changes in the local strain energy at all locations in three-dimensional space in each iteration, when the overall strain energy of the system no longer decreases significantly, it can be determined that the evolution process has stabilized, and the final microscopic stress field distribution can be determined accordingly.

[0118] This embodiment also provides a digital twin system for analyzing the microscopic stress field of a polymer blend system, including: a data acquisition unit, which acquires the types and macroscopic proportions of each polymer in the polymer blend system, and combines ultrasonic detection to acquire the response state inside the system, and maps and generates a non-uniform mixed state field characterizing the internal structure in three-dimensional space.

[0119] The simulation unit adaptively generates multiple cores corresponding to local regions based on the non-uniformity characteristics of the non-uniform mixed state field; each core is used as an intelligent agent, and a field with competitive dominance over spatial position is constructed.

[0120] The evolution unit, combining the historical characteristics of the liquid stage of the polymer blend system, drives each agent to evolve in the field in an adversarial manner; when evolving to the Nash equilibrium state, it generates the micro-stress field of the polymer blend system.

[0121] This embodiment also provides a computer device applicable to the digital twin method for analyzing the micro-stress field of polymer blend systems, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the digital twin method for analyzing the micro-stress field of polymer blend systems as proposed in the above embodiment.

[0122] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0123] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the digital twin method for analyzing the microscopic stress field of polymer blend systems as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0124] In summary, this invention achieves dynamic simulation and precise analysis of the internal stress formation mechanism of the polymer blend system by: acquiring multidimensional acoustic response characteristics of the polymer blend system using an ultrasonic array and constructing a non-uniform mixing state field in three-dimensional space to characterize the internal structural distribution; extracting spatial non-uniformity characteristics and adaptively generating multiple structural cores; further abstracting each structural core into an intelligent agent with independent behavioral capabilities and constructing a spatial competition and dominance relationship based on structural topological characteristics, enabling each intelligent agent to form influence diffusion and interaction in space; subsequently, combining the material composition information and mixing history characteristics of the liquid stage of the blend system, driving dedicated intelligent agents with different mechanical response strategies to perform adversarial evolution in space, and continuously updating the stress state at each spatial location through the collaborative action of multiple intelligent agents; finally, by monitoring the changes in the overall local strain energy of the system, generating the micro-stress field of the polymer blend system when the evolution process tends to stabilize and reaches Nash equilibrium, thereby realizing dynamic simulation and precise analysis of the internal stress formation mechanism of the blend system during cooling and solidification.

[0125] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A digital twin method for analyzing the microscopic stress field of polymer blend systems, characterized in that: This includes obtaining the types and macroscopic proportions of each polymer in a polymer blend system, and combining ultrasonic detection to obtain the response state inside the system, and mapping and generating a non-uniform mixing state field that characterizes the internal structure in three-dimensional space. Based on the non-uniformity characteristics in the non-uniform mixed state field, multiple cores corresponding to local regions are adaptively generated; Each core is treated as an intelligent agent, and a field is constructed in which there is a competitive dominance relationship over spatial location; Based on the historical characteristics of the liquid stage of the polymer blend system, the various intelligent agents are driven to undergo adversarial evolution in the field; When evolving to the Nash equilibrium state, a micro-stress field is generated in the polymer blend system.

2. The digital twin method for analyzing the microscopic stress field of polymer blend systems as described in claim 1, characterized in that: The response state is the echo amplitude, propagation delay, spectral attenuation, phase shift, and scattering intensity characteristics obtained by the ultrasonic array when it penetrates the polymer blend system. The non-uniform mixing state field includes mapping the acquired ultrasonic testing data according to the corresponding spatial coordinates in the space of the blended system to form the distribution result of the ultrasonic testing data in three-dimensional space.

3. The digital twin method for analyzing the microscopic stress field of polymer blend systems as described in claim 2, characterized in that: The non-uniformity features include multi-dimensional features representing non-uniformity by measuring the changes in echo amplitude, propagation delay difference, spectral attenuation, phase shift, and scattering intensity of the ultrasonic detection data between adjacent spatial locations, thereby determining the local variation intensity of different spatial regions within the blend system.

4. The digital twin method for analyzing the microscopic stress field of polymer blend systems as described in claim 3, characterized in that: The core of the adaptive generation of multiple corresponding local regions includes adaptively clustering spatial locations in three-dimensional space based on multi-dimensional features at different coordinate positions to form multiple feature clustering regions. In the three-dimensional space of the blend system, the results of the mixing and curing processes of all polymers are simplified: Suppose that, starting from each cluster core, feature diffusion is performed to the surrounding spatial locations, so that each cluster core has a diffusion effect on the surrounding spatial locations to varying degrees; The adaptive clustering includes constructing a local structure tensor by calculating the partial derivatives of the multidimensional features in three-dimensional space; performing eigenvalue decomposition on the local structure tensor to extract feature values ​​and two directions of the features, and generating internal texture features that characterize the continuity and bidirectional orientation of positional features. By detecting the different locations of the intelligent agent in space: Step 1: During the process of the detection agent traversing in three-dimensional space based on the internal texture features, the traversal optimization parameters of the agent's current spatial coordinates in the three dimensions are calculated and monitored in real time: Local centroid distance: The relative distance between the current position and the global geometric centroid of the continuous texture in three-dimensional space; Local attachment distance: The shortest spatial distance from the current position to the nearest texture skeleton or high-density feature trajectory; Pointing convergence: The topological position of the current location in the texture bidirectional pointing vector field, calculated as the endpoint or convergence degree of the texture vector flow direction; Step 2: Using a single-dimensional extreme value triggering mechanism, when the agent identifies that any single dimension parameter among the above three dimensions has reached a local maximum extreme value in the agent's traversal neighborhood, the probing agent immediately becomes a clustering core. The single-dimensional extreme value triggering mechanism allows multiple heterogeneous clustering cores to be adaptively generated within the same complex texture region due to extreme value triggering in different dimensions. Among them, for any continuous texture feature in space, it is allowed to simultaneously penetrate and belong to multiple cluster families governed by different clustering cores in the topological extension.

5. The digital twin method for analyzing the microscopic stress field of polymer blend systems as described in claim 4, characterized in that: The field that has a competitive dominance relationship over spatial location includes treating each cluster core as an independent intelligent agent; Based on the influence diffusion of each agent to the surrounding spatial location, which decreases with spatial distance and multidimensional feature similarity; In the process of the spread of influence, it is affected by the size and scope of the race: In the adaptive clustering process, the cumulative topological length of all continuous texture features affected by a single cluster core and classified into that cluster family is extracted as the race size corresponding to the agent; and the race size is positively mapped to the initial reference radiation intensity for the agent to perform influence diffusion through a preset mapping function 1. Extract the global topological envelope boundary occupied by all continuous texture features affected by a single cluster core in three-dimensional space, which is taken as the range of the agent; and map the spatial span of the range to the spatial attenuation coefficient of the agent's influence through a preset mapping function 2. At any spatial coordinate position, after adjusting the spatial distance and the spatial attenuation coefficient, each intelligent agent jointly governs the stress calculation results at the coordinate position.

6. The digital twin method for analyzing the microscopic stress field of polymer blend systems as described in claim 5, characterized in that: The adversarial evolution includes the following step 1: For the three dimensions of local centroid distance, local attachment distance and direction convergence degree in the single-dimensional extreme value triggering mechanism, three exclusive agent models with heterogeneous mechanical response strategies are independently pre-trained in combination with the physical force characteristics of the corresponding local topological morphology. When determining the clustering core through the single-dimensional extreme value triggering mechanism, the single-dimensional parameter type that triggered the generation of the clustering core is traced back, and the clustering core is mapped to the corresponding dimension's dedicated intelligent agent; Step 2: The types and macroscopic proportions of each polymer in the polymer blend system, as well as the vector sequence of the mixing action, are used as features and injected into each dedicated intelligent agent to establish the initial physical properties of each dedicated intelligent agent. Each dedicated intelligent agent, based on its own pre-training and initial physical properties, aims to minimize the local strain energy within its own dominance space and outputs counteracting actions of its internal tension tensor. Step 3: At any coordinate position in the three-dimensional space, using the common dominance ratio of each dedicated agent after adjustment by spatial distance and spatial attenuation coefficient, tensor weighted fusion is performed on the adversarial actions output by multiple dedicated agents to obtain stress calculation results. The stress calculation results are used as state penalty information and transmitted in reverse along the dominance relationship of the agents to each dedicated agent that has an impact, so that each dedicated agent can continuously update its own adversarial strategy and state tensor based on the state penalty information.

7. The digital twin method for analyzing the microscopic stress field of polymer blend systems as described in claim 6, characterized in that: The Nash equilibrium state includes generating the local strain energy corresponding to each coordinate position based on the stress calculation results at each coordinate position and in combination with the iterative process of adversarial evolution. During the adversarial evolution process, each dedicated intelligent agent continuously updates its own state tensor; When the continuous updating of the state tensor fails to reduce the local strain energy at any coordinate position in the three-dimensional space or reach the maximum number of iterations, the adversarial evolution is determined to have reached a Nash equilibrium state.

8. A digital twin system for analyzing the microscopic stress field of polymer blend systems, based on the digital twin method for analyzing the microscopic stress field of polymer blend systems according to any one of claims 1 to 7, characterized in that: It includes a data acquisition unit that acquires the types and macroscopic proportions of each polymer in the polymer blend system, and combines ultrasonic detection to acquire the response state inside the system, and maps and generates a non-uniform mixing state field that characterizes the internal structure in three-dimensional space. The simulation unit adaptively generates multiple cores corresponding to local regions based on the non-uniformity characteristics of the non-uniform mixed state field; each core is used as an intelligent agent, and a field with competitive dominance over spatial position is constructed. The evolution unit, combining the historical characteristics of the liquid stage of the polymer blend system, drives each agent to evolve in the field in an adversarial manner; when evolving to the Nash equilibrium state, it generates the micro-stress field of the polymer blend system.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the digital twin method for analyzing the microscopic stress field of the polymer blend system according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the digital twin method for analyzing the microscopic stress field of the polymer blend system according to any one of claims 1 to 7.

Images