A method for preparing laser micro-grooves on the surface of AZ31B magnesium alloy based on machine learning optimization

Abstract

The application discloses a kind of based on machine learning optimization's AZ31B magnesium alloy surface laser micro-groove preparation method, the method includes the following steps: step 1, constructs AZ31B magnesium alloy nanosecond laser micro-groove processing parameter and performance index dataset;Step 2, magnesium alloy nanosecond laser processing micro-groove geometric feature machine learning prediction model training and fusion;Step 3, NSGA-II optimization processing parameter;Step 4, constructs LDH film and is combined with stearic acid modification.The application process flow is clear, can be directly implemented on the basis of existing nanosecond laser and hydrothermal equipment, parameter can be quickly optimized by machine learning, with good repeatability and stability.AZ31B magnesium alloy surface prepared simultaneously has superhydrophobicity and high corrosion resistance, significantly prolongs the service life of component, is suitable for the engineering application in the field such as automobile lightweight parts, aerospace structural parts, ocean engineering equipment and 3C electronics.

Description

Technical Field

[0001] This invention relates to a method for preparing laser microgrooves on the surface of magnesium alloys, specifically a method for preparing laser microgrooves on the surface of AZ31B magnesium alloys based on machine learning optimization. Background Technology

[0002] Magnesium alloys, due to their low density, high specific strength, and excellent biocompatibility, are widely used in aerospace, automotive manufacturing, and biomedical implants. However, their high electrochemical activity and susceptibility to corrosion severely limit their practical engineering applications. For example, corrosion of magnesium alloy components inside aircraft cabins can not only affect the tightness of connections but also trigger crack propagation in structural components, posing a threat to flight safety. For helicopters and spacecraft, corrosion not only increases maintenance costs but can also lead to mission failure. Therefore, improving the corrosion resistance of magnesium alloys in aerospace service environments has become a key issue restricting their further application.

[0003] Currently, conversion coating technology, coating and composite coating protection technology, and superhydrophobic surface anti-corrosion technology based on micro / nano structures are commonly used methods to improve the corrosion resistance of magnesium alloys. Conversion coatings form a stable film on the alloy surface through chemical or electrochemical processes, thereby blocking corrosive media and reducing the corrosion rate. These films can be used not only as independent protective layers but also as pretreatment layers for subsequent organic coatings or electroplating, enhancing adhesion and overall corrosion resistance. Coating and composite coating protection technology works by forming a dense barrier layer on the metal substrate surface, preventing direct contact between corrosive media and the magnesium alloy substrate, thus delaying electrochemical corrosion. Commonly used technologies include epoxy resin coatings, polyurethane coatings, fluoropolymer coatings, sol-gel coatings, and ceramic coatings. Superhydrophobic surface anti-corrosion technology based on micro / nano structures mainly stems from the "air cushion effect." When water droplets contact the surface, the micro / nano structure traps air, forming a stable solid-gas-liquid three-phase interface, significantly reducing the contact area between the corrosive media and the substrate, hindering electron transport and ion diffusion, and thus delaying the corrosion process. However, current methods suffer from limitations in protective performance, which are susceptible to defects in the film and contamination from the service environment. They also rely heavily on the consistency and stability of the micro / nano structure geometry, and the complex coupling relationships between various processing parameters during microstructure fabrication make it difficult to achieve stable and controllable structure construction. Therefore, it is necessary to develop a machine learning-based method for optimizing the processing parameters of laser microgrooving on the surface of AZ31B magnesium alloy, and combine it with LDH film and low surface energy molecular modification to construct a composite protective layer that combines superhydrophobicity and corrosion resistance. Summary of the Invention

[0004] This invention provides a machine learning-optimized method for fabricating laser-guided microgrooves on the surface of AZ31B magnesium alloy. A predictive model is constructed using Support Vector Regression (SVR), Random Forest (RF), and Lightweight Gradient Boosting Machine (LGBM), and fused within an ensemble learning framework to achieve high-precision prediction of the groove depth-to-width ratio (DVR), with a prediction error of less than 4%. Combined with Non-Dominated Sorting Genetic Algorithm II (NSGA-II), the optimal combination of processing parameters can be obtained to maximize the DVR of the fabricated grooves. In experimental verification, constructing an LDH thin film on the AZ31B magnesium alloy surface after processing with optimized parameters and then performing low surface energy modification significantly improves its superhydrophobicity and corrosion resistance. This provides a new surface protection solution for the long-term service of magnesium alloys in the automotive, aerospace, 3C electronics, and marine engineering fields, and has promising prospects for industrial application.

[0005] The objective of this invention is achieved through the following technical solution:

[0006] A machine learning-optimized method for fabricating laser microgrooves on the surface of AZ31B magnesium alloy includes the following steps:

[0007] Step 1: Construct a dataset of nanosecond laser microgrooving parameters and performance indicators for AZ31B magnesium alloy:

[0008] Step 1-1: Process the AZ31B magnesium alloy plate into small square plates and polish them to a surface roughness of 5~15 nm.

[0009] Steps 1-2: After ultrasonic cleaning of the sample with anhydrous ethanol, the irradiation power, number of irradiations and irradiation speed are selected as key factors for prediction and optimization. Microgroove structures with different groove widths and depths are prepared on a nanosecond laser micromachining system. The irradiation power is controlled at 10~18 W, the number of irradiations is 80~160 times and the scanning speed is 100~500 mm / s.

[0010] Steps 1-3: After processing, clean again and store in a vacuum container;

[0011] Steps 1-4: Use a white light interferometer to accurately measure the depth and width of the microgrooves in each group of samples. Each group of samples is tested three times. The data obtained will provide support for subsequent optimization of processing parameters based on machine learning.

[0012] Step 2: Training and fusion of machine learning prediction models for the geometric features of microgrooves processed by nanosecond lasers on magnesium alloys:

[0013] Step 2-1: Using the laser process parameters (irradiation power, number of irradiations, and irradiation rate) and experimental data on groove width and depth as the training set, optimize the hyperparameters using random search combined with five-fold cross-validation, and use RMSE and R² values. 2 Evaluate the generalization performance of the model;

[0014] Step 2-2: Select three regression models—LGBM (Lightweight Gradient Boosting Machine), RF (Random Forest), and SVR-rbf (Support Vector Regression Based on Radial Basis Function)—to establish the nonlinear mapping between laser process parameters and groove width and groove depth;

[0015] Steps 2-3: Construct a Voting ensemble model that integrates three base learners. Perform a weighted ensemble of the prediction results from multiple models. The results show that the Voting model achieves the lowest RMSE and highest R-value in both slot width and slot depth prediction. 2 It significantly outperforms each individual model and can be used for new sample recommendation and subsequent multi-objective optimization.

[0016] Step 3: Optimize machining parameters using NSGA-II:

[0017] Step 3-1: Use the NSGA-II multi-objective optimization algorithm, with slot width and slot depth as dual objectives, and combine it with the Voting ensemble model to construct an integrated prediction-optimization framework;

[0018] Step 3-2: Implement and iteratively solve the problem in Python to obtain the Pareto optimal frontier with a reasonable distribution through non-dominated sorting and crowding strategies;

[0019] Step 3-3: Select the five sets of laser processing parameters with the largest aspect ratio from many Pareto optimal solutions, providing a reliable reference for process optimization;

[0020] Step 4: Construct an LDH membrane and modify it with stearic acid:

[0021] Step 4-1: By optimizing the processing parameters, a nanosecond laser is used to etch a grid groove on the surface of AZ31B magnesium alloy, and then a MgAl-LDH film is constructed by hydrothermal method, and then a low surface energy modification is performed on it with stearic acid ethanol solution.

[0022] Step 4-2: Conduct contact angle and electrochemical tests on the prepared surface to verify its superhydrophobic and corrosion-resistant properties.

[0023] Compared with the prior art, the present invention has the following advantages:

[0024] 1. Utilizing Machine Learning for Efficient Optimization Design of Magnesium Alloy Laser Microgrooving Structures: SVR, RF, and LGBM machine learning algorithms were integrated into a Voting model for optimizing the processing parameters of nanosecond laser microgrooving on AZ31B magnesium alloy. A data-driven prediction system based on "process parameters—groove width / depth—depth-to-width ratio" was constructed, with a prediction error of less than 4%. On this basis, the NSGA-II multi-objective optimization algorithm was coupled to obtain a reasonably distributed Pareto optimal solution set, and the parameter combination with the largest depth-to-width ratio was selected, realizing the transformation of magnesium alloy surface structure design from empirical parameter selection to intelligent optimization design.

[0025] 2. Pre-treatment with laser microgrooves before LDH film construction to form a multi-scale micro / nano composite structure: Before constructing the LDH film, optimized nanosecond lasers are used to precisely etch grid microgrooves on the surface of AZ31B magnesium alloy, enabling the substrate to obtain a controllable microscale geometry beforehand. Then, the MgAl-LDH film is grown in situ using a hydrothermal method. This process achieves the coupling of laser micromachining and LDH growth, constructing a multi-scale rough interface of "microgrooves + LDH nanosheets," providing a key structural foundation for the subsequent formation of superhydrophobic surfaces.

[0026] 3. Significantly Enhanced Superhydrophobicity and Corrosion Resistance of AZ31B Magnesium Alloy: Through a synergistic design of "laser microgroove structure + MgAl-LDH film + stearic acid low surface energy modification," a composite interface with multi-scale roughness and low surface energy is constructed on the surface of AZ31B magnesium alloy. This significantly improves the surface contact angle and droplet rolling properties, achieving a stable superhydrophobic state, with the contact angle increasing from 62° in the substrate to 153°. Simultaneously, the ion barrier and corrosion inhibition effects of the LDH film, combined with the shielding effects of the air layer and hydrophobic layer, significantly reduce the corrosion current and increase the interfacial impedance, thereby substantially improving the corrosion resistance of AZ31B magnesium alloy. The corrosion current density is reduced from 2.477 × 10⁻⁶ in the substrate. -5 A·cm -2 Reduced to 9.08 × 10 -6 A·cm -2 .

[0027] 4. The process flow of this invention is clear and can be directly implemented based on existing nanosecond laser and hydrothermal equipment. Parameters can be quickly optimized using machine learning, exhibiting good repeatability and stability. The prepared AZ31B magnesium alloy surface possesses both superhydrophobicity and high corrosion resistance, significantly extending the service life of components. It is suitable for engineering applications in fields such as lightweight automotive parts, aerospace structural components, marine engineering equipment, and 3C electronics. Attached Figure Description

[0028] Figure 1 A flowchart for machine learning and surface preparation;

[0029] Figure 2 This is a comparison chart of predictions and actual values ​​based on the LGBM model.

[0030] Figure 3 This is a comparison chart of predictions and actual values ​​based on the RF model.

[0031] Figure 4 This is a comparison chart of predictions and actual values ​​based on the SVR-rbf model.

[0032] Figure 5 This is a comparison chart of predictions and actual results based on an ensemble learning model.

[0033] Figure 6 A comparison chart of R² and RMSE for four machine learning models;

[0034] Figure 7 Pareto front plot;

[0035] Figure 8 This is a SEM image of the surface after laser etching.

[0036] Figure 9 SEM image of the surface after laser etching-LDH film-stearic acid modification;

[0037] Figure 10 This is a comparison diagram of the contact angle before and after surface treatment;

[0038] Figure 11 This is a comparison of polarization curves before and after surface treatment. Detailed Implementation

[0039] The technical solution of the present invention will be further described below with reference to the accompanying drawings, but it is not limited thereto. Any modifications or equivalent substitutions to the technical solution of the present invention that do not depart from the spirit and scope of the technical solution of the present invention should be covered within the protection scope of the present invention.

[0040] This invention provides a machine learning-optimized method for fabricating laser microgrooves on the surface of AZ31B magnesium alloy. The method uses optimized processing parameters to construct an LDH thin film on the AZ31B magnesium alloy surface and further modifies it with low surface energy, which can significantly improve its superhydrophobicity and corrosion resistance. Figure 1 As shown, the specific steps are as follows:

[0041] Step 1: A nanosecond laser with a pulse width of 20 ns and a center wavelength of 1064 nm was used to fabricate microgrooves on a small AZ31B magnesium alloy square plate (20 mm × 20 mm × 2 mm) with a polished surface roughness of approximately 10 nm. The sample was fixed on a motion platform. By changing three parameters—irradiation power (10–18 W), number of irradiations (80–160 times), and scanning speed (100–500 mm / s)—a total of 5 × 5 × 5 = 125 process combinations were designed to fabricate microgroove structures with different groove widths and depths. Before and after processing, the samples were ultrasonically cleaned with anhydrous ethanol and vacuum-preserved to reduce oxidation and secondary contamination.

[0042] Step 2: Using a white light interferometer, the width and depth of the microgrooves in each group of samples were precisely measured. Each group of data was tested three times to ensure the reliability and statistical significance of the results. These experimental data constitute the dataset for the subsequent machine learning model.

[0043] Step 3: In the machine learning modeling part, irradiation power, number of irradiations, and scanning speed are first used as input features, and groove width and groove depth are used as prediction targets to construct an internal dataset. Hyperparameter tuning is performed using random search combined with five-fold cross-validation, and the results are evaluated through RMSE and R... 2 The generalization performance of the models was evaluated. The selected models included LightGBM (LGBM), Random Forest (RF), and Radial Basis Kernel Support Vector Regression (SVR-rbf), and a Voting ensemble learning model was further designed to fuse multiple base learners. Figure 2-5 The charts show a comparison of the predicted and actual values ​​of slot depth and slot width for LGBM, RF, SVR-rbf, and the ensemble learning model, respectively. Compared to the three basic models, the ensemble learning model's scatter points closely revolve around the Y=T line, exhibiting strong linear consistency, high prediction accuracy, and small error. Figure 6 This compares the coefficients of determination (R²) and root mean square errors (RMSE) of four machine learning models. The ensemble learning model has the highest coefficient of determination and the smallest error. (The text also mentions the coefficient of determination R² for slot width.) 2 The root mean square error (RMSE) is 0.8086, and the coefficient of determination for the trench depth is R0. 2 The value is 0.9795, and the root mean square error (RMSE) is 1.2286.

[0044] Step 4: Employ the NSGA-II multi-objective genetic algorithm, using both trench width and depth as optimization objectives, and combine it with the aforementioned Voting ensemble model to construct an integrated "prediction-optimization" framework. Through multiple rounds of iterative evolution, a series of Pareto front solutions are obtained, from which representative process parameter combinations with a high depth-to-width ratio are selected. Following the above process, multiple rounds of iterative calculations successfully yielded several Pareto front optimal solutions, such as... Figure 7 As shown.

[0045] Step 5: Using optimized process parameters, a mesh-like microgroove structure was formed on the magnesium alloy substrate using laser processing. The resulting sample was then ultrasonically cleaned in anhydrous ethanol and stored in a vacuum environment. The SEM image of the sample surface after laser processing is shown below. Figure 8 As shown. Prepare 100 mL of 0.04 mol / L aluminum nitrate solution, and add sodium hydroxide (NaOH) to adjust the pH of the solution to 11. Transfer the laser-processed and pretreated magnesium alloy sample together with the prepared mixture into a hydrothermal reactor, and place the reactor in a constant temperature drying oven for hydrothermal reaction at 120℃ for 24 h. After the hydrothermal reaction is completed, allow it to cool naturally to room temperature, remove the sample, and thoroughly clean it with deionized water.

[0046] Step 6: Prepare 100 ml of a 0.5 mol / L stearic acid-ethanol solution. Place the sample with the constructed LDH membrane and the stearic acid-ethanol solution in a hydrothermal reactor and heat at 70°C for 6 hours in an oven. After modification, thoroughly clean the sample surface with anhydrous ethanol, then place it in a vacuum drying oven and dry at 80°C for 1 hour. After laser processing and LDH membrane construction, the sample surface is further modified with stearic acid. The SEM image of its surface morphology is shown below. Figure 9 As shown.

[0047] Step 7: Using an optical contact angle meter and deionized water as the test liquid, the surface of the magnesium alloy sample was tested using the static contact angle measurement method. The volume of each droplet added was 2 μL, and the testing environment was room temperature (approximately 20°C). Optical photographs of the contact angle between the magnesium alloy substrate and the prepared sample are shown below. Figure 10 As shown, the contact angle of the untreated magnesium alloy substrate surface is 62°, while the contact angle of the treated surface is 153°, which significantly improves the superhydrophobicity of the magnesium alloy sample.

[0048] Step 8: The corrosion resistance test uses a three-electrode system, including a reference electrode (saturated calomel electrode), a working electrode (1cm... 2 The sample surface and counter electrode (platinum sheet electrode) were etched using a 3.5 wt.% NaCl solution. The sample was connected to a copper wire using double-sided conductive tape and then sealed with UV-cured adhesive, exposing only the microstructure to be tested. The sample was then immersed in the 3.5 wt.% NaCl solution and allowed to stand at the open circuit potential (OCP) for a period of time (e.g., 20–30 min) until the open circuit potential stabilized. Subsequently, potentiodynamic polarization testing was performed, using the stable open circuit potential as the initial reference potential. Testing was conducted within a voltage range of ±1000 mV from the open circuit potential, with a potentiodynamic scan rate of 1 mV / s. The resulting polarization curves were recorded. The obtained polarization curves are shown below. Figure 11As shown, compared with the magnesium alloy matrix, the polarization curve of the prepared sample shows a positive shift in corrosion potential and a significant decrease in corrosion current density, indicating that the prepared sample (laser etching-LDH-SA) improves the corrosion resistance of the magnesium alloy surface.

Claims

1. A method for fabricating laser microgrooves on the surface of AZ31B magnesium alloy based on machine learning optimization, characterized in that... The method includes the following steps: Step 1: Construct a dataset of nanosecond laser microgroove processing parameters and performance indicators for AZ31B magnesium alloy; Step 2: Training and fusion of machine learning prediction models for the geometric features of microgrooves processed by nanosecond lasers on magnesium alloys: Step 2-1: Using experimental data on irradiation power, number of irradiations, irradiation rate, trench width, and trench depth as the training set, optimize the hyperparameters using random search combined with five-fold cross-validation, and use RMSE and R² values. 2 Evaluate the generalization performance of the model; Step 2-2: Select three regression models, LGBM, RF, and SVR-rbf, to establish the nonlinear mapping between laser process parameters and groove width and groove depth; Steps 2-3: Construct a Voting ensemble model that integrates the three base learners and perform weighted ensemble of the prediction results from multiple models; Step 3: Optimize machining parameters using NSGA-II: Step 3-1: Use the NSGA-II multi-objective optimization algorithm, with slot width and slot depth as dual objectives, and combine it with the Voting ensemble model to construct an integrated prediction-optimization framework; Step 3-2: Implement and iteratively solve the problem in Python to obtain the Pareto optimal frontier with a reasonable distribution through non-dominated sorting and crowding strategies; Step 3-3: Select the five sets of laser processing parameters with the largest aspect ratio from the many Pareto optimal solutions; Step 4: Construct an LDH membrane and modify it with stearic acid: By optimizing the processing parameters, a nanosecond laser was used to etch grid grooves on the surface of AZ31B magnesium alloy, and then a MgAl-LDH film was constructed by hydrothermal method, followed by low surface energy modification with stearic acid ethanol solution.

2. The method for fabricating laser microgrooves on the surface of AZ31B magnesium alloy based on machine learning optimization according to claim 1, characterized in that... The specific steps of step 1 are as follows: Step 1-1: Process the AZ31B magnesium alloy plate into small square plates and polish them; Steps 1-2: After ultrasonic cleaning of the sample with anhydrous ethanol, the irradiation power, number of irradiations and irradiation rate are selected as key factors for prediction and optimization. Microgroove structures with different groove widths and depths are prepared on a nanosecond laser micromachining system. Steps 1-3: After processing, clean again and store in a vacuum container; Steps 1-4: Use a white light interferometer to accurately measure the depth and width of the microgrooves in each group of samples. Each group of samples is tested three times. The data obtained will provide support for subsequent optimization of processing parameters based on machine learning.

3. The method for preparing laser microgrooves on the surface of AZ31B magnesium alloy based on machine learning optimization according to claim 2, characterized in that... In step 1-1, the surface is polished to a roughness of 5~15 nm.

4. The method for preparing laser microgrooves on the surface of AZ31B magnesium alloy based on machine learning optimization according to claim 2, characterized in that... In steps 1-2, the irradiation power is controlled to be 10~18 W, the number of irradiations is 80~160 times, and the irradiation rate is 100~500 mm / s.

Images