Aluminum alloy fatigue life prediction method based on machine learning and physical model fusion

Through the integrated system of XGBoost optimized by combining graph neural network, particle swarm algorithm and physical model parameter optimization, the problem of insufficient accuracy and adaptability of traditional aluminum alloy fatigue life prediction methods is solved, and efficient and accurate fatigue life prediction is achieved, which enhances the Model reliability and transparency.

CN119066980AInactive Publication Date: 2024-12-03LIAONING TECHNICAL UNIVERSITY
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411249781.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-06
Publication Date
2024-12-03
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Traditional aluminum alloy fatigue life prediction methods rely on empirical physical models, with limited prediction accuracy and insufficient adaptability to complex environments. Pure data-driven methods lack physical significance, making them difficult to gain trust in the engineering community, and require a large amount of labeled data to be difficult to obtain.

Method used

Combined with graph neural network (GNN), XGBoost (PSO-XGBoost) optimized by particle swarm algorithm and physical model parameter optimization, an integrated system is formed, and through data expansion, parameter optimization and physical model adjustment, the prediction accuracy and automation level are improved.

Benefits of technology

It significantly improves the accuracy and efficiency of fatigue life prediction of aluminum alloys, enhances model transparency and reliability, reduces experimental costs, is suitable for complex application needs, and improves the ability to understand and choose decisions in material performance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066980A_ABST
    Figure CN119066980A_ABST
Patent Text Reader

Abstract

The invention relates to an extreme gradient boosting (PSO-XGBoost) model and physical model parameter optimization system in combination with a graph neural network (GNN) and particle swarm optimization. The system is used for improving the prediction precision and reliability of the fatigue life of an aluminum alloy. The system flow comprises the following steps: 1, obtaining aluminum alloy fatigue data through an experiment, and carrying out feature enhancement by using a graph neural network; 2, optimizing XGBoost model parameters by using a particle swarm algorithm to ensure a minimum prediction error; 3, adopting a machine learning model to predict result feedback and adjust key parameters of a physical model; and 4, verifying the accuracy of the model by comparing experimental data. The system reduces human errors through an automatic process, is suitable for material fatigue performance analysis, effectively reduces research and engineering costs, and has a wide application prospect.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of material science and data analysis, and relates to a method for predicting fatigue life of aluminum alloy. Background Art

[0002] Aluminum alloy is a widely used structural material, and accurate prediction of its fatigue life is crucial to the safety and reliability of engineering structures. Traditional methods mainly rely on empirical physical models. Although these models provide theoretical support, their prediction accuracy is limited by insufficient experimental data or errors, and they are not adaptable to complex environments. With the development of machine learning technology, data-driven fatigue life prediction has become a new trend. However, although pure data-driven methods can process large amounts of data to discover complex associations, they often lack physical meaning and are difficult to gain trust from the engineering community. In addition, these models usually require a large amount of labeled data, which is difficult to obtain in actual material testing. Therefore, there is an urgent need for a new system that combines the advantages of machine learning and physical models to provide accurate and physically interpretable predictions to meet complex application requirements. This system can improve prediction efficiency and accuracy, reduce experimental costs, and enhance model transparency and reliability, which is of great value to aircraft, automobiles, high-speed rail, ships and other fields. Summary of the invention

[0003] In view of the limitations of the existing technology, this paper proposes an aluminum alloy fatigue life prediction system that combines graph neural network (GNN), particle swarm optimization XGBoost (PSO-XGBoost) and physical model parameter optimization. The system integrates modern machine learning technology with traditional physical models, significantly improves the accuracy and automation level of prediction, and overcomes the shortcomings of previous technologies.

[0004] Technical implementation details: S1. Data collection and processing: Experimental data collection: Through standardized fatigue tests, the fatigue life data of aluminum alloys at different stress levels are systematically collected, and the stress amplitude, maximum and minimum stresses, and corresponding fatigue life are recorded. Data expansion: A three-layer graph convolutional network graph neural network (GNN) is used for data expansion, with the feature dimension set to 64 and the ReLU activation function used. The network is trained using the Adam optimizer, with a learning rate of 0.001, a batch size of 32, and 200 iterations.

[0005] S2, XGBoost model optimized by particle swarm algorithm (PSO-XGBoost): Parameter optimization: Genetic algorithm was used to adjust key parameters such as the number of trees (100 to 300) and the maximum depth of the tree (3 to 10), the population size was set to 50, 40 generations were iterated, the learning factors c1 and c2 were both 0.2, and the roulette wheel selection strategy, two-point crossover and uniform mutation were used. Model training and evaluation: After training, the mean square error (MSE) and coefficient of determination (R²) were used to evaluate the model performance.

[0006] S3. Optimization of physical model parameters: 1. Detailed explanation of the model formula. The basic expression used in the physical model is as follows, which combines the fatigue crack growth theory in materials science: (1) in C , m , n , α is the material constant. 2. Parameter optimization Based on the output of the random forest model, optimize the parameters in the physical model C , m , n and α , the goal is to minimize the difference between the predicted lifetime and the experimental lifetime.

[0007] S4. System verification and application: 1. Result verification. Using mean square error (MSE) and determination coefficient (R 2 ) Evaluate the model to verify the accuracy of the prediction results. 2. Parameter adjustment and optimization. Adjust the parameters according to the prediction error, and continue to optimize the model performance through an iterative process to improve the prediction accuracy.

[0008] Beneficial effects of the technical solution: 1. Improve processing efficiency and accuracy: The system-integrated GNN and PSO-XGBoost algorithms optimize the data processing and model parameter adjustment processes, greatly improve prediction efficiency and accuracy, and ensure the consistency and repeatability of results through automated processes.

[0009] 2. Enhanced data analysis capabilities: The use of graph neural networks not only expands the data set, but also enhances the ability to express data features, making the prediction of aluminum alloy fatigue life more accurate, helping to deeply understand the performance of materials under different conditions, and supporting more effective material selection and application decisions.

[0010] 3. Ease of operation: The system is developed based on a modern computing platform, with an intuitive user interface, supports automated operation and batch processing, simplifies the user operation process, lowers the technical threshold, and is suitable for a wide range of applications in industry and scientific research.

[0011] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the following is a detailed description in combination with the preferred embodiments and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the drawings of the embodiment are briefly introduced below.

[0013] Figure 1 It is a system flow chart of the present invention.

[0014] FIG. 2 is a progress diagram of the particle swarm algorithm of the present invention.

[0015] FIG3 is a 45-degree comparison diagram of the predicted value and the actual value of the PSO-XGBoost model of the present invention.

[0016] FIG. 4 is a SN curve diagram of the predicted value based on the PSO-XGBoost model of the present invention. DETAILED DESCRIPTION

[0017] This paper introduces in detail an automated aluminum alloy fatigue life prediction system, which effectively improves data processing efficiency and prediction accuracy by combining advanced data analysis technology and physical model optimization. The following content, combined with the accompanying drawings, shows the workflow and results of the system.

[0018] like Figures 1 to 4 As shown in the figure, the system can accurately predict the fatigue life of aluminum alloy through the following steps: S1. Data processing: Aluminum alloy fatigue data is obtained through standard experimental methods, and graph neural network (GNN) is used for data expansion to form a detailed data set for subsequent analysis.

[0019] S2. Model optimization: Particle swarm optimization (PSO) is used to adjust the XGBoost model parameters to achieve the best prediction effect. Figure 2 ) demonstrate the optimization process and verify the effectiveness and stability of the algorithm; Figure 2 In the example shown, the fitness of the swarm optimization algorithm gradually approaches stability after 80 times, and the optimal effect after particle swarm optimization is obtained; then, the optimized model is used to predict fatigue life, and the comparison chart with the actual value ( Figure 3 ) and mean square error (MSE) and coefficient of determination (R 2 ) to evaluate model performance, such as Figure 3 As shown in the comparison chart of predicted values ​​and true values, all predicted points are within the error line of 1.25, and R 2 The value of 0.96 indicates that the prediction accuracy is very high.

[0020] S3. Optimization of physical model parameters: Using the optimized prediction data, fit the physical model through the L-BFGS-B algorithm, adjust and determine the parameters C, m, n, α to improve the accuracy of the model ( Figure 4 SN curve graph showing predicted values ​​and actual values).

[0021] S4. Result verification and optimization: The system compares the prediction results with the experimental data, adjusts the model parameters according to the differences, and repeats the prediction process to refine the prediction accuracy.

[0022] S5. Report generation: The system automatically generates a report containing prediction results and parameter optimization details, which can be exported to PDF or other formats for easy analysis and sharing.

[0023] By integrating advanced machine learning technology with traditional physical models, this system significantly improves the accuracy and efficiency of predictions, reduces experimental costs, and increases the transparency and credibility of the model. It is suitable for fatigue performance analysis of aluminum alloys in fields such as aircraft, automobiles, high-speed railways and ships, and has broad practical application value.

[0024] The above is only a preferred embodiment of the present invention, which certainly cannot be used to limit the scope of rights of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and changes can be made without departing from the principle of the present invention, and these improvements and changes are also regarded as the protection scope of the present invention.

Claims

1. A method for predicting fatigue life of aluminum alloys by integrating machine learning and physical models, characterized in that: It includes the following modules: Data acquisition module: used to collect fatigue life data of aluminum alloy under different stress levels through experimental methods; Data enhancement module: expands the original data set based on graph neural network (GNN); model optimization module: adjusts the parameters of the XGBoost model through particle swarm optimization (PSO) technology; physical model parameter optimization module: feedback optimizes the key parameters of the physical model through the prediction data output by the XGBoost model; system verification module: used to compare the model prediction results with experimental data.

2. According to the system described in claim 1, the graph neural network (GNN) in the data enhancement module is trained using the Adam optimizer, with a learning rate of 0.001, a batch size of 32, and a training cycle of at least 100 cycles.

3. According to the system described in claim 1, the execution parameters of the particle swarm optimization algorithm in the model optimization module are: the population size is set to 50, the two learning factors c1 and c2 are both set to 0.2, and the total number of iterations is 100 times.

4. According to the system described in claim 1, the optimization of the physical model parameters adopts L-BFGS-B algorithm to fit the physical model formula, and its optimization goal is to minimize the difference between the predicted fatigue life and the experimental results. The physical model formula includes: (1) in C , m , n , α are material constants and these parameters are precisely adjusted by the optimization algorithm.

5. The system described in claim 1, wherein the system verification module verifies the prediction accuracy and reliability of the system based on R² value and mean square error (MSE).

Citation Information

Cited By

  • Foamed aluminum process parameter optimization method based on physical information deep learning

    CN120656620A