CN-122024953-A - Causal reasoning-based epoxy resin material performance prediction method and causal reasoning-based epoxy resin material performance prediction system

Abstract

Introducing chemical bond constraint, energy constraint and a crosslinking reaction nonlinear processing mechanism to constrain search space and direction judgment of a causal discovery algorithm; and carrying out causal reasoning based on the causal graph to realize quantitative prediction of target macroscopic performance and output interpretable attribution results. According to the application, causal reasoning is introduced into the performance prediction of the epoxy resin material, and the universal causal discovery algorithm is subjected to territorial improvement by fusing chemical bond constraint, energy constraint and crosslinking nonlinear processing mechanism, so that the model not only has high prediction precision, but also can reveal an intrinsic causal conduction path between formula-structure-performance, and the interpretative prediction and the anti-facts intervention of the material design are realized.

Inventors

• DU CHEN
• WEI ZHIYUN

Assignees

• 上海灵纭科技有限公司

Dates

Publication Date

20260512

Application Date

20260122

Claims (10)

1. 1. The causal reasoning-based epoxy resin material performance prediction method is characterized by comprising the following steps of: Introducing a chemical bond constraint, an energy constraint and a crosslinking reaction nonlinear processing mechanism to constrain the search space and direction judgment of a causal discovery algorithm; constructing a causal graph for predicting the performance of the epoxy resin material based on a constrained causal discovery algorithm; And carrying out causal reasoning based on the causal graph, realizing quantitative prediction of the target macroscopic performance, and outputting an interpretable attribution result.
2. 2. The causal reasoning-based epoxy resin material performance prediction method of claim 1, wherein the introducing chemical bond constraints, energy constraints and crosslinking reaction nonlinear processing mechanisms constrains search space and direction decisions of a causal discovery algorithm, comprising: Introducing chemical bond constraint to limit edges which do not accord with chemical bonding rules in the causal graph; introducing energy constraint to ensure that the causal relationship in the causal graph accords with a molecular energy change rule; Introducing a crosslinking reaction nonlinear treatment mechanism to treat the crosslinking reaction nonlinear characteristic of the epoxy resin.
3. 3. The causal reasoning-based epoxy resin material performance prediction method of claim 1, wherein the constraint-based causal discovery algorithm constructs a causal graph for predicting the epoxy resin material performance, comprising: Obtaining molecular design parameters of epoxy resin as input variables; generating microstructures of a plurality of epoxy resin systems through molecular dynamics simulation, extracting intermediate structural features reflecting structural states and macroscopic properties of the intermediate structural features, and constructing a data set; Based on the data set, a constrained causal discovery algorithm is combined with a multi-algorithm integration strategy to identify causal paths from input variables, intermediate structural features to macroscopic performance and generate a causal graph.
4. 4. The method for predicting the performance of an epoxy resin material based on causal reasoning of claim 3, wherein the molecular design parameters of the epoxy resin comprise epoxy resin monomer type and molecular weight, curing agent type and molecular weight, molar ratio of epoxy group to amine hydrogen, functional group density and initial conformational parameters.
5. 5. A causal reasoning based epoxy resin material property prediction method according to claim 3, wherein the generating microstructures of a plurality of epoxy resin systems by molecular dynamics simulation and extracting intermediate structural features reflecting structural states and macroscopic properties thereof, constructing a dataset comprises: automatically generating an initial model of an epoxy resin system containing different molecular types, proportions and initial conformations through scripts; sequentially carrying out energy minimization and molecular dynamics pre-balancing under an NPT system on each initial model to obtain a pre-balanced microstructure; Based on the pre-balanced microstructure, adopting an iterative reaction algorithm, randomly selecting and executing a bonding operation in an atomic pair meeting the distance and reactivity criteria, updating molecular topology and force field parameters, then performing relaxation simulation, repeating until reaching a preset crosslinking degree, forming a crosslinked network microstructure, and extracting crosslinking density and free volume fraction from the crosslinked network microstructure as intermediate structural characteristics; Performing balance simulation on the final crosslinked network microstructure, and obtaining macroscopic performance by applying external conditions; And (3) correlating the intermediate structural features with macroscopic properties corresponding to different molecular types, proportions and initial conformations of each system to construct a structured data set.
6. 6. A causal reasoning based epoxy material performance prediction method according to claim 3, wherein the causal finding algorithm after applying constraints based on the dataset, in combination with a multi-algorithm integration strategy, identifies causal paths from input variables, intermediate structural features to macroscopic performance, generating a causal graph comprising: Selecting a set of causal discovery algorithms based on different principles and assumptions from a causal discovery algorithm library, wherein the causal discovery algorithm library comprises a PC algorithm, a GES algorithm and a LiNGAM algorithm; Inputting the data set into each selected algorithm respectively, and independently running each algorithm to generate a corresponding initial causal graph, wherein the nodes comprise input variables, intermediate structural features and macroscopic performance; integrating all the initial causal graphs to obtain a unified integrated causal graph; Pruning and complementing the integrated causal graph based on priori knowledge in the chemical field, removing edges which do not accord with chemical rules or complementing omitted key causal relationships, and obtaining a final causal graph.
7. 7. The causal reasoning based epoxy material performance prediction method of any of claims 1-6, wherein the causal graph comprises a plurality of nodes and directed edges between the connecting nodes, forming a hierarchical causal network; The nodes are divided into input variable nodes, intermediate structure characteristic nodes and macroscopic performance nodes.
8. 8. The causal reasoning-based epoxy resin material performance prediction method of claim 7, wherein the reasoning based on the causal graph realizes quantitative prediction of target macroscopic performance, and outputs interpretable attribution results, comprising: Based on a causal graph formed by input variable nodes, intermediate structural feature nodes and macroscopic performance nodes, constructing a corresponding structural equation for directed causal edges connecting the nodes to form a structural equation model; executing causal intervention on the structural equation model, calculating response values of intermediate structural feature nodes layer by layer according to the topological sequence of the causal graph by setting values of input variable nodes, and transmitting the results to macroscopic performance nodes to realize quantitative prediction of target macroscopic performance; And responding to the prediction result interpretation request, identifying a causal path from the input variable node to the target macroscopic performance node through the intermediate structural feature node, extracting a node sequence on the path and parameters of a corresponding structural equation, and analyzing which intermediate structural features the input variables influence the macroscopic performance.
9. 9. The causal reasoning-based epoxy resin material performance prediction method of claim 1, comprising deploying a causal discovery algorithm based on chemical bond constraints, energy constraints and crosslinking reaction nonlinear processing mechanism constraints on a cloud computing platform, supporting interfacing with a material database, and displaying a causal graph, a causal path and an intervention suggestion through an interactive visual interface.
10. 10. An epoxy resin material performance prediction system based on causal reasoning, comprising: the constraint module introduces chemical bond constraint, energy constraint and a crosslinking reaction nonlinear processing mechanism to constrain the search space and direction judgment of the causal discovery algorithm; the generation module is used for constructing a causal graph for predicting the performance of the epoxy resin material based on a constrained causal discovery algorithm; And the prediction module is used for carrying out causal reasoning based on the causal graph, realizing quantitative prediction on the target macroscopic performance and outputting an interpretable attribution result.

Description

Causal reasoning-based epoxy resin material performance prediction method and causal reasoning-based epoxy resin material performance prediction system Technical Field The application relates to the crossing field of material science and artificial intelligence, in particular to a causal reasoning-based epoxy resin material performance prediction method and system. Background Epoxy resin is used as an important thermosetting polymer material and is widely applied to the fields of aerospace, electronic packaging, composite materials and the like. The traditional epoxy resin material design mainly depends on experience and error testing methods, and has the problems of long development period and high cost. The existing material performance prediction method mainly comprises a machine learning-based method, wherein although a mapping relation between input characteristics and output performance can be established, the method lacks of interpretability and cannot reveal a causal mechanism between a material structure and performance. Although the method based on the physical model has clear physical meaning, the method is difficult to establish an accurate physical model for a complex epoxy resin system. The method based on molecular dynamics simulation can describe the material behavior from an atomic level, but has high calculation cost and is difficult to solve the problem of large-scale material screening. The causal reasoning is used as an emerging artificial intelligence method, can discover causal relation among variables from observed data, solves the problems, and provides a new thought for material design. However, the existing causal discovery algorithm is mainly designed aiming at general data, and does not consider the specificity of the material science field, in particular the characteristics of chemical bond constraint, energy constraint, crosslinking reaction nonlinearity and the like of the epoxy resin material. Disclosure of Invention In view of the above-mentioned drawbacks of the prior art, the present application is directed to a causal reasoning-based method and system for predicting the performance of an epoxy resin material. The application provides a causal reasoning-based epoxy resin material performance prediction method, which is characterized by comprising the following steps of: Introducing a chemical bond constraint, an energy constraint and a crosslinking reaction nonlinear processing mechanism to constrain the search space and direction judgment of a causal discovery algorithm; constructing a causal graph for predicting the performance of the epoxy resin material based on a constrained causal discovery algorithm; And carrying out causal reasoning based on the causal graph, realizing quantitative prediction of the target macroscopic performance, and outputting an interpretable attribution result. Optionally, the introducing chemical bond constraint, energy constraint and crosslinking reaction nonlinear processing mechanism constrains search space and direction determination of the causal discovery algorithm, including: Introducing chemical bond constraint to limit edges which do not accord with chemical bonding rules in the causal graph; introducing energy constraint to ensure that the causal relationship in the causal graph accords with a molecular energy change rule; Introducing a crosslinking reaction nonlinear treatment mechanism to treat the crosslinking reaction nonlinear characteristic of the epoxy resin. Optionally, the constructing a causal graph for predicting the performance of the epoxy resin material based on the constrained causal discovery algorithm includes: Obtaining molecular design parameters of epoxy resin as input variables; generating microstructures of a plurality of epoxy resin systems through molecular dynamics simulation, extracting intermediate structural features reflecting structural states and macroscopic properties of the intermediate structural features, and constructing a data set; Based on the data set, a constrained causal discovery algorithm is combined with a multi-algorithm integration strategy to identify causal paths from input variables, intermediate structural features to macroscopic performance and generate a causal graph. Optionally, the molecular design parameters of the epoxy resin include epoxy resin monomer type and its molecular weight, curing agent type and its molecular weight, molar ratio of epoxy groups to amine hydrogens, functional group density, and initial conformational parameters. Optionally, the generating microstructures of the plurality of epoxy resin systems through molecular dynamics simulation, extracting intermediate structural features reflecting structural states and macroscopic properties thereof, and constructing a data set includes: automatically generating an initial model of an epoxy resin system containing different molecular types, proportions and initial conformations through scripts; sequentially carryi