Fatigue life prediction method for thin micro-arc oxidation coating aluminum alloy

Abstract

The invention provides a method for predicting the fatigue life of a thin micro-arc oxidation coating aluminum alloy. And a physical equation for predicting the fatigue life of the aluminum alloy with the thin micro-arc oxidation coating is obtained by utilizing an image processing technology of machine learning, a Murakami and crack propagation theory, an adversarial generative neural network and a random forest algorithm. According to the method, the micropore size of the ceramic coating can be accurately extracted, the fatigue life of the coating aluminum alloy under the conditions of small samples, different micropore distributions and cyclic stress loads can be accurately predicted, and the method is suitable for predicting the fatigue life of micro-arc oxidation of an aluminum alloy material of air-space-earth equipment. And a basic support is provided for solving the problems of poor fatigue life and low reliability of the coating aluminum alloy.

Description

Technical Field

[0001] The present invention relates to the field of materials science and engineering, and specifically to a fatigue life prediction method for thin micro-arc oxidation coated aluminum alloy. The system combines advanced machine learning algorithms with material fatigue damage theory, analyzes and models the fatigue behavior of thin micro-arc oxidation aluminum alloy under different stress conditions, and uses machine learning image processing technology to obtain the diameter of micropores on the coating surface, thereby achieving high-precision prediction of its fatigue life. This technology is particularly suitable for life assessment and design optimization of key components of thin micro-arc oxidation aluminum alloys based on surface performance improvement in aerospace, automobile manufacturing and other high-performance industrial fields, aiming to improve the reliability of coating materials and the safety of engineering applications. Background Art

[0002] Thin micro-arc oxidation coatings are widely used in high-performance industrial fields such as aerospace, automobile manufacturing, and electronic equipment because of their significant advantages in improving the surface hardness, wear resistance, corrosion resistance, and aesthetics of aluminum alloys. Micro-arc oxidation (MAO) technology forms a dense and uniform matrix oxide ceramic coating in situ on the surface of aluminum alloys through electro-hydraulic-chemical action, which effectively enhances the surface properties of the matrix material. However, although thin micro-arc oxidation coatings can significantly improve the surface properties of aluminum alloys, the difficulty in predicting their fatigue life is still one of the important factors affecting their practical application. Thin micro-arc oxidation coatings have micropores, and thin coatings will not introduce interface defects into the matrix. Therefore, under the action of cyclic stress, micropores are the key to initiating fatigue crack formation, and the coating cracks extend into the interior of the substrate, eventually leading to fatigue fracture of the coated aluminum alloy, seriously affecting the reliability and service life of the components.

[0003] Traditional fatigue life prediction methods mainly rely on experimental tests and theoretical analysis based on fracture mechanics. These methods usually require a large number of tensile or fatigue tests, which are time-consuming, labor-intensive and costly. In addition, the amount of experimental data is limited and is restricted by the complexity of the experimental conditions, making it difficult for the prediction accuracy based on empirical formulas and simple statistical models to meet actual engineering needs. Especially in the fatigue life prediction of thin micro-arc oxidation coated aluminum alloys, the coating has small micropores, wide distribution and diverse characteristics, and the coating preparation is expensive, making the fatigue life of thin micro-arc oxidation coated aluminum alloys difficult to predict.

[0004] With the rapid development of artificial intelligence and machine learning technologies, data-driven prediction methods have shown great potential in the field of materials science. However, the current research on machine learning prediction systems for fatigue life of thin micro-arc oxidation coated aluminum alloys is still insufficient, especially in data expansion, coating surface pore statistics, model optimization and physical parameter solution. There are still many challenges. At present, how to effectively integrate experimental data, machine learning algorithms and physical models to build an efficient and accurate fatigue life prediction system is still a research topic that needs to be solved urgently. Therefore, the development of a fatigue life prediction system for thin micro-arc oxidation coated aluminum alloys based on machine learning can not only solve the problems of insufficient data volume and low prediction accuracy in traditional methods, but also identify micropore characteristics and count the micropore diameters on the coating surface. It can also significantly improve the prediction efficiency and reliability, which has important theoretical significance and broad application prospects. Summary of the invention

[0005] The present invention provides a fatigue life prediction method for thin micro-arc oxidation coated aluminum alloy, aiming to improve the fatigue life prediction accuracy and efficiency of coated aluminum alloy parts in practical applications through a systematic and scientific method. The system comprehensively utilizes fatigue life data obtained from fatigue experiments, graph neural network (GNN) for data expansion, machine learning image processing technology to obtain the average diameter of micropores, combined with the defect equivalent Murakami theory and the physical model constructed by crack growth rate and the random forest prediction model, to achieve accurate prediction of the fatigue life of thin micro-arc oxidation coated aluminum alloy under different working conditions. The specific implementation steps are as follows: S1. Dataset creation and expansion The system first conducts a series of standardized fatigue experiments to obtain fatigue life data of thin micro-arc oxidation coated aluminum alloy under different stress levels (e.g., 200 MPa, 250 MPa, 300 MPa, etc.) and cycle numbers. At least 5 repeated tests are performed under each set of experimental conditions to ensure the statistical reliability of the data. During the experiment, external factors such as ambient temperature and humidity are strictly controlled to ensure the consistency and repeatability of the data. The original fatigue life data obtained will be used as the initial data set to provide a basis for subsequent data expansion and model training. In order to make up for the problem of insufficient experimental data, the system uses a graph neural network (GNN) to expand the initial data set. The GNN structure has at least 3 hidden layers, each layer contains 64 neurons, and the activation function uses ReLU (Rectified Linear Unit), and is trained using the Adam optimization algorithm, with the learning rate set between 0.001 and 0.01. Through the trained GNN model, the original data is diversified and the data set is expanded to three times the original data volume to improve the training effect and generalization ability of the subsequent machine learning model.

[0006] S2. Physical model construction After the data set is established and expanded, the system combines the defect equivalent Murakami theory and crack growth rate to build a physical model that reflects the fatigue failure behavior of coated aluminum alloys and derives the physical expression of fatigue life: (1) in S e is the fatigue limit of micro-arc oxidation coated aluminum alloy, m eff and C eff is the material constant, a 0 is the initial defect length. According to Murakami theory and coating defect characteristics, a 0 is approximately the average diameter of micropores on the coating surface, which is calculated by the machine learning open cv module. This physical model takes into account the effect of thin micro-arc oxidation coating on the fatigue life of the substrate, and combines it with the defect equivalent theory to ensure that the prediction results are not only statistically accurate, but also physically reasonable.

[0007] S3. Random forest prediction model establishment and evaluation Using the expanded data set, the system constructs a random forest prediction model. The random forest model includes 200 decision trees, the maximum depth of each tree is limited to 20 layers, and the minimum number of sample splits is set to 5 to prevent the model from overfitting. The input features of the random forest model and the expanded data features are output as fatigue life prediction values. In order to ensure the prediction performance of the model, the system trains and verifies the random forest model through the ten-fold cross validation method. By calculating the coefficient of determination (R 2 ) and mean square error (MSE) to comprehensively evaluate the prediction accuracy and stability of the model. Ideally, the model should achieve (R 2 ) value is not less than 0.90, and the MSE value is not higher than 50.

[0008] S4. Parameter solution and model optimization Based on the prediction results of the random forest model, the system solves the key parameters in the physical model S e , m eff and C eff . The specific method is to use the fitting relationship between the random forest prediction value and the actual fatigue life data to solve the unknown parameters in the physical model. During the optimization process, the optimal solution is gradually approached through iterative optimization algorithms (such as the Levenberg-Marquardt algorithm). The optimized parameters are substituted into the physical model to form an accurate fatigue life prediction expression. This expression can achieve high-precision prediction of the fatigue life of aluminum alloys under different stress conditions, combined with the microscopic defect characteristics of the material.

[0009] S5. System application and verification Finally, the system applies the optimized fatigue life prediction expression to the life prediction and design optimization of actual thin micro-arc oxidation aluminum alloy parts. The accuracy and practicality of the system prediction are further confirmed by comparing and verifying with the fatigue life data in actual use. During the verification process, actual thin micro-arc oxidation aluminum alloys under different working conditions are selected, and the system is used to predict the life, and compared and analyzed with the actual observation data to ensure the reliability and applicability of the prediction model.

[0010] By integrating experimental data, advanced machine learning algorithms and physical models, this invention not only solves the problems of insufficient data and low prediction accuracy in traditional fatigue life prediction methods, but also significantly improves prediction efficiency and reliability. Its wide applicability enables it to be applied to the life prediction of thin micro-arc oxidation aluminum alloy parts in multiple high-performance industrial fields such as aerospace and automobile manufacturing, with significant economic benefits and social value. Beneficial Effects

[0011] This invention uses advanced machine learning techniques such as graph neural networks and random forests, combined with a physical model constructed by material defect theory and crack growth rate, to achieve high-precision prediction of the fatigue life of thin micro-arc oxidation coated aluminum alloy. Specific advantages include: 1. Strong data expansion capability: GNN is used to effectively expand experimental data, which solves the problem of insufficient experimental data and improves the training effect and generalization ability of the model.

[0012] 2. Solid physical model foundation: Combining Murakami theory and crack growth rate, a fatigue life model with physical significance is constructed to ensure the rationality and accuracy of the prediction results.

[0013] 3. High accuracy of prediction model: The random forest model can capture the complex nonlinear relationships in the data through integrated learning of multiple decision trees, significantly improving the accuracy and stability of fatigue life prediction.

[0014] 4. Wide range of applications: The system is suitable for a variety of thin micro-arc oxidation coated aluminum alloys, has good promotion and application potential, and can meet the needs of high-performance industrial fields such as aerospace, automobile manufacturing, etc.

[0015] In summary, the thin micro-arc oxidation coating aluminum alloy fatigue life prediction system based on machine learning described in the present invention has the advantages of efficient data processing, high prediction accuracy, and a wide range of applications. It can significantly improve the reliability and service life of aluminum alloy components in practical applications, and has important theoretical significance and broad application prospects.

[0016] 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

[0017] 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.

[0018] Figure 1 It is a system flow chart of the present invention.

[0019] Figure 2 This is the open cv porosity statistics chart.

[0020] Figure 3 This is a comparison chart of the random forest predicted value and the actual value.

[0021] Figure 4 It is a line chart comparing the predicted value and the actual value. DETAILED DESCRIPTION

[0022] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the specific implementation mode of the present invention will be described in detail in conjunction with the accompanying drawings. The principle and operation process of the present invention are explained through examples (the thin micro-arc oxidation coating 2024-T3 aluminum alloy is selected as a case), as well as other aspects, features and advantages of the present invention. In the referenced drawings, the detailed workflow and result display of the fatigue life prediction system of the thin micro-arc oxidation coating aluminum alloy based on machine learning provided by the present invention are shown through different charts. In conjunction with the accompanying drawings, the specific implementation mode of the fatigue life prediction system of the thin micro-arc oxidation coating aluminum alloy based on machine learning of the present invention is described in detail below.

[0023] S1. Dataset establishment First, collect the data set through experiments. The specific steps are as follows: Tensile test data acquisition: Standardized tensile-tensile fatigue tests are performed on thin micro-arc oxidation coated aluminum alloy samples at different stress levels (such as 200 MPa, 250 MPa, 300 MPa, etc.) and number of cycles. Data preprocessing: The collected input feature data is normalized to ensure that the data range of different features is consistent and to avoid weight bias caused by scale differences during model training. At the same time, the target variable (fatigue life) is logarithmically transformed to meet the training requirements of the machine learning model and improve the training efficiency and prediction accuracy of the model. In order to solve the problem of insufficient experimental data, the system uses a graph neural network (GNN) to expand the initial data set. The GNN structure as described in claim 2 includes at least 3 hidden layers, each layer includes 64 neurons, the activation function uses ReLU (Rectified Linear Unit), and is trained using the Adam optimization algorithm, and the learning rate is set between 0.001 and 0.01. After the training is completed, the GNN model generates an extended data set, expanding the original data volume to three times the original amount to enhance the training effect and generalization ability of the subsequent machine learning model.

[0024] Figure 1 The overall process of the system is demonstrated, including the main steps of data acquisition, data expansion, physical model construction, prediction model establishment and evaluation, parameter optimization, and system application and verification. The steps are connected by arrows, reflecting the logical order of the system and the data flow.

[0025] S2, initial defect length a 0Get like Figure 2 The results are shown in the statistical analysis of the porosity of the micropores on the coating surface using OpenCV, and the average diameter of the pores on the coating surface is statistically calculated ( a 0=2.2μm).

[0026] S3. Random forest prediction model establishment and evaluation Using the expanded data set, the system constructs a random forest prediction model. The random forest model, as described in claim 3, includes 200 decision trees, the maximum depth of each tree is limited to 20 layers, and the minimum number of sample splits is set to 5 to prevent the model from overfitting. The input features of the model include stress level, material parameters (such as alloy composition, coating thickness, etc.) and expanded data features, and the output is the fatigue life prediction value. The random forest model is trained and validated using the ten-fold cross validation method. By calculating the determination coefficient (R 2 ) and mean square error (MSE) to evaluate the predictive performance of the model. Figure 3 , the comparison chart of the predicted value of the random forest model and the actual experimental value and Figure 4 It shows that the model has a high accuracy. 2 ≥0.90 and MSE less than 50, reflecting good prediction accuracy and stability.

[0027] S4. Parameter solution and model optimization Based on the prediction results of the random forest model, the key parameters in the physical model are calculated and obtained. m eff =3.53, C eff =1.2×10 -12 , S e = 155. An accurate fatigue life prediction model was obtained.

[0028] S5. System application and verification The optimized fatigue life prediction expression is applied to the life prediction and design optimization of actual aluminum alloy parts. By selecting actual aluminum alloy parts under different working conditions, the system is used to predict the life and compared with the actual observation data. The verification results show that the system can accurately predict the fatigue life of actual parts.

Claims

1. A fatigue life prediction method for thin micro-arc oxidation coating aluminum alloy, characterized in that: The following steps are involved: Obtain fatigue life data of aluminum alloys at different stress levels through fatigue experiments; The fatigue life data was expanded using graph neural network (GNN) to generate an extended data set. A physical model of fatigue life was constructed based on Murakami theory and crack growth rate. A random forest prediction model was established to predict the fatigue life of aluminum alloy using the extended data set. The model performance was evaluated by evaluating the coefficient of determination (R²) and mean square error (MSE) of the prediction model. Based on the output results of the prediction model, the key parameters in the physical model were solved and the fatigue life prediction expression was optimized.

2. The system according to claim 1, wherein the graph neural network (GNN) has at least 3 hidden layers, each layer contains 32 to 128 neurons, the activation function adopts ReLU, and is trained using the Adam optimization algorithm, and the learning rate is set between 0.001 and 0.

01.

3. The system according to claim 1, wherein the random forest model includes 100 to 500 decision trees, the maximum depth of each tree is limited to 10 to 50 layers, and the minimum sample split number is set to 2 to 10 to prevent overfitting.

4. The system according to claim 1, wherein the physical model is based on Murakami and crack growth theory, and the specific expression is as follows: （1） in S e is the fatigue limit of micro-arc oxidation coated aluminum alloy, m eff and C eff is the material constant, a 0 is the equivalent initial defect length. According to Murakami theory and coating defect characteristics, a 0 is approximately the average diameter of the micropores on the coating surface, which is calculated based on the electron microscope scanning image using the machine learning open cv module.

Images