Special polymer multi-environment coupling aging track prediction method and system

Abstract

The application relates to the cross field of artificial intelligence and material science, and discloses a special polymer multi-environment coupling aging track prediction method and system, which solves the problems that existing aging life prediction is difficult to handle multi-factor coupling and lacks micro-mechanism characterization. The method acquires an initial molecular topological graph and a service environment stress sequence, extracts polymer chain topological features by using a graph neural network, and extracts dynamic stress features by using a time series deep learning network; time and space features are aligned and fused through a cross-modal attention mechanism; and the macroscopic performance aging track curve and the microscopic bond rupture probability distribution of a polymer material are synchronously predicted. The application can realize accurate prediction of material performance under multi-coupling environment and provide micro-explanation of failure mechanism, and is suitable for service life evaluation of polymer materials in the fields of aviation, aerospace and the like.

Description

Technical Field

[0001] This invention relates to the field of interdisciplinary technology of artificial intelligence and materials science, and more specifically, to a method and system for predicting the multi-environment coupling aging trajectory of special polymers by integrating time series and graph neural networks. Background Technology

[0002] Polymer materials used in specialized fields such as aerospace and nuclear industries face the combined effects of multiple environmental factors during service, including thermal oxidation, ultraviolet radiation, and mechanical stress, leading to complex aging and degradation. Traditional material lifetime prediction relies primarily on simple empirical formulas such as the Arrhenius equation. However, these formulas typically only handle the linear effects of single environmental factors and struggle to characterize the nonlinear effects under multiple stress couplings. More importantly, existing lifetime prediction methods only provide macroscopic performance degradation trends, failing to reflect the microscopic physicochemical nature of molecular chain breakage and cross-linking network destruction, thus lacking scientific explanation for failure analysis. While the development of deep learning technology has led to data-driven prediction models, most are "black box" models, struggling to deeply align the dynamic evolution of the service environment with the response of the polymer topology. Therefore, an intelligent solution is urgently needed that can achieve both high-precision trajectory prediction and provide explanations for microscopic failures. Summary of the Invention

[0003] This invention aims to solve the technical problems of existing polymer aging prediction methods, which struggle to handle multi-factor coupling and lack microscopic mechanism characterization, and provides a method and system for predicting the aging trajectory of special polymers under multi-environment coupling.

[0004] To achieve the above objectives, the first technical solution adopted by the present invention is: a method for predicting the aging trajectory of special polymers in multiple environments, comprising the following steps: Acquire molecular network topology data of polymer materials in their initial state, as well as time-series environmental stress data during service. The topological features of the polymer chain are obtained by extracting features from the molecular network topology graph data using a graph neural network. A time-series deep learning network is used to extract features from the time-series environmental stress data to obtain multi-environment coupled dynamic features. By fusing the topological features of the polymer chain with the dynamic features of multi-environment coupling through a cross-modal attention mechanism, spatiotemporal coupled aging features are obtained. Based on the spatiotemporal coupled aging characteristics, the macroscopic performance aging trajectory curve of the polymer material and the microscopic breakage probability distribution of chemical bonds in the molecular network topology data are predicted simultaneously.

[0005] To achieve the above objectives, the second technical solution adopted by the present invention is: a special polymer multi-environment coupled aging trajectory prediction system, including a data acquisition module, a polymer topology coding module, a stress coding module, a cross-modal fusion module, and a joint prediction and interpretation module.

[0006] The beneficial effects of this invention are as follows: by aligning graph neural networks and time series networks across modes, dynamic modeling of environmental stress and molecular structure response is achieved; the cross-modal attention mechanism not only improves the accuracy of macroscopic lifetime prediction, but also explicitly points out the weak links in the molecular network that are prone to breakage under specific stress conditions, providing a scientific basis for the service safety and long-life design of special polymer materials. Attached Figure Description

[0007] Figure 1 A flowchart of a method for predicting the aging trajectory of a special polymer in a multi-environment coupling manner, provided in an embodiment of the present invention; Figure 2 A schematic diagram of the architecture of the spatiotemporally interwoven deep learning prediction model provided in an embodiment of the present invention; Figure 3 This is a logical diagram illustrating the cross-modal attention mechanism for processing microscopic topology and macroscopic stress, as provided in an embodiment of the present invention. Detailed Implementation

[0008] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0009] The present invention provides a method for predicting the aging trajectory of special polymers in a multi-environment coupling manner, the core of which lies in constructing an end-to-end mapping of "microstructure-dynamic environment-macro performance".

[0010] In step S1, the system first models the initial cross-linking network of the polymer material as a graph structure. Nodes represent atoms or specific functional groups, and edges represent covalent bonds or cross-linking points. Spatial features are extracted using a graph neural network (GNN). Simultaneously, the stress encoding module processes time-series data containing temperature, ultraviolet intensity, and mechanical load using a recurrent neural network or Transformer architecture.

[0011] The core of this embodiment lies in the cross-modal attention mechanism in step S4. The system maps the microscopic topological features extracted by the GNN to query vectors and the dynamic features of environmental stress to key and value vectors.

[0012] Through this mechanism, the model can automatically learn which type of chemical bond (such as main chain carbon-carbon bond or cross-linking point) has a higher contribution weight under specific service environments (such as high temperature pulses).

[0013] Finally, the system utilizes a multi-task learning head to output prediction results in parallel. The macroscopic branch outputs the tensile strength decay curve of the material; the microscopic interpretation branch calculates the breakage probability distribution of each edge (chemical bond) in the graph at the current moment based on attention weights. For example, when predicting the aging of a certain type of polyurethane material, the model can explicitly indicate the high breakage risk of urethane groups under high temperature and high humidity conditions, thus achieving physicochemical interpretability of the prediction results.

Claims

1. A method for predicting the aging trajectory of special polymers in multiple environments, characterized in that, Includes the following steps: The process involves acquiring molecular network topology data of polymer materials in their initial state, as well as time-series environmental stress data during service; using a graph neural network to extract features from the molecular network topology data to obtain polymer chain topological structure features; and using a time-series deep learning network to extract features from the time-series environmental stress data to obtain multi-environment coupling dynamic features. By fusing the topological features of the polymer chain with the dynamic features of multi-environment coupling through a cross-modal attention mechanism, spatiotemporal coupled aging features are obtained. Based on the spatiotemporal coupled aging characteristics, the macroscopic performance aging trajectory curve of the polymer material and the microscopic breakage probability distribution of chemical bonds in the molecular network topology data are predicted simultaneously.

2. The method for predicting the aging trajectory of special polymers in a multi-environment coupling according to claim 1, characterized in that, The nodes in the molecular network topology graph data represent atoms or functional groups, and the edges represent chemical bonds or cross-linking points. The extraction of polymer chain topology features using graph neural networks includes message passing on the graph to fuse the spatial location features and chemical bond energy features between nodes.

3. The method for predicting the aging trajectory of special polymers in a multi-environment coupling according to claim 1, characterized in that, The time-series environmental stress data includes at least two of the following: thermal stress curves, ultraviolet radiation dose sequences, oxygen concentration sequences, and external mechanical stress curves.

4. The method for predicting the aging trajectory of special polymers under multi-environment coupling according to claim 1, characterized in that, The cross-modal attention mechanism maps the topological features of the polymer chain to query vectors and the multi-environment coupled dynamic features to bond vectors and value vectors, calculates the contribution weights of stress features at different time steps to different chemical bond topological features, and achieves end-to-end fusion of the spatiotemporal coupled aging features.

5. A special polymer multi-environment coupled aging trajectory prediction system, characterized in that, The system includes a data acquisition module, a polymer topology coding module, a stress coding module, a cross-modal fusion module, and a joint prediction and interpretation module, and is configured to perform the method described in any one of claims 1 to 4.

Images