DLC coating doping optimization method fusing support vector machine and genetic algorithm
By integrating support vector machines and genetic algorithms to optimize the doping of DLC coatings, the problems of insufficient database and incomplete modeling in DLC coating design are solved. This method achieves efficient and low-cost multi-objective performance optimization and produces DLC coatings with high hardness, low friction coefficient and excellent corrosion resistance.
Patent Information
- Application Number
- CN202510960919.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-12
- Publication Date
- 2025-10-31
AI Technical Summary
Existing technologies lack high-quality database support, feature engineering, and modeling frameworks in DLC coating doping design, making it difficult to achieve multi-objective performance optimization, resulting in low design efficiency and high cost.
By integrating support vector machines and genetic algorithms, a coating doping optimization method is constructed. Through experimental databases, feature engineering, performance prediction models, and genetic algorithm optimization, the rapid and accurate design of doping elements is achieved, and DLC coatings with high hardness, low friction coefficient, and excellent corrosion resistance are prepared.
It enables the rapid screening of high-performance dopant element combinations based on partial experimental or simulation calculations, reducing design costs, improving the overall performance of coatings, and has significant engineering application prospects.
Smart Images
Figure CN120877982A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of intelligent design and performance prediction technology of thin film materials, specifically involving a DLC coating doping optimization method that integrates support vector machine and genetic algorithm. Background Technology
[0002] Diamond-like carbon (DLC) coatings are a type of amorphous carbon coating that combines the properties of diamond and graphite. Due to their high hardness, low coefficient of friction, good chemical stability, and excellent corrosion resistance, they are widely used in high-end fields such as aerospace, automotive manufacturing, medical devices, electronic packaging, and precision machining. To further expand the application boundaries of DLC coatings, researchers have proposed a method of doping with various metallic or non-metallic elements (such as Si, N, O, F, Ti, Cu, etc.) to achieve synergistic control over the coating's microstructure, mechanical properties, electrical properties, and chemical stability.
[0003] However, different dopants play significantly different roles in DLC coatings, influenced by factors such as electronegativity, chemical activity, and physical behavior during deposition. Furthermore, the performance of DLC coatings is constrained by multiple process variables, including gas ratios, power supply parameters, deposition time, and operating pressure. Traditional coating design methods often rely on experimental trial-and-error strategies or limited parameter space scanning, which are not only time-consuming and costly but also struggle to obtain optimal solutions under multiple performance requirements.
[0004] In recent years, with the accumulation of materials data and the development of computational methods, machine learning has shown great potential in the field of materials design. By constructing a mapping relationship between coating formulations and properties, machine learning can efficiently model and predict complex systems, potentially overcoming the bottlenecks of traditional methods in performance control and design efficiency. Existing research has shown that data-driven methods can optimize targets such as the hardness and electrochemical performance of high-entropy coatings.
[0005] However, current research on machine learning-based doped DLC coating design is still in its early stages and suffers from the following shortcomings: (1) a lack of high-quality, structured experimental databases for doped DLCs; (2) the absence of a generalizable feature engineering and modeling framework; and (3) an incomplete coupling feedback mechanism between performance prediction and experimental verification. Therefore, it is urgent to establish a systematic machine learning-driven coating design method to improve the ability to predict specific targets and iterate rapidly in DLC coatings, thereby promoting the practical application and engineering implementation of intelligent coating design technology. Summary of the Invention
[0006] The purpose of this invention is to overcome the above-mentioned defects in the background technology and propose a DLC coating doping optimization method that integrates support vector machine and genetic algorithm. It does not rely on a large number of repeated experiments and trial and error, nor does it require additional complex process control equipment. The modeling and optimization algorithm process is simple, the performance prediction speed is fast, and the multi-objective optimization anti-bias capability is strong. It can quickly and accurately achieve the optimal design of dopant element type and concentration, and obtain DLC coating with high hardness, low friction coefficient and excellent corrosion resistance.
[0007] To achieve the above objectives, the present invention provides the following technical solution:
[0008] A method for optimizing the doping of DLC coatings by integrating support vector machines and genetic algorithms includes the following steps:
[0009] Step S1: Construct an experimental database of doped DLC coatings. The data in the experimental database includes the types and concentrations of doping elements in the coating, as well as the mechanical, electrochemical, and tribological properties of the coating, namely the ratio of hardness to elastic modulus, coefficient of friction, and charge transfer resistance.
[0010] Step S2: Based on the experimental database from Step S1, perform feature engineering and data preprocessing to extract feature parameters, doping element properties, process variables, and test results to construct input and output datasets.
[0011] Step S3: Construct a performance prediction model. The input and output datasets obtained in step S2 are used as training sets. A machine learning algorithm is used, and feature parameters are extracted as input features of the machine learning model to train the coating performance prediction model and obtain the trained coating performance prediction model.
[0012] Step S4: Construct an optimization objective function. The predicted coating performance indicators, including the ratio of hardness to elastic modulus, friction coefficient, and charge transfer resistance, are output by the coating performance prediction model trained in Step S3 as target variables. A genetic algorithm is then introduced to optimize and obtain the optimal dopant element content.
[0013] Step S5: Using DLC coating deposition technology, a doped DLC coating is prepared based on the optimal doping element content in step S4. The performance is verified, and the verification results are used to update the database and iteratively optimize the model and design results.
[0014] Furthermore, the types of doping elements mentioned in step S1 include Si, N, Ag, F, Ti, Cu, B, Cr, W, and Mo.
[0015] Furthermore, the characteristic parameters in step S2 include the atomic radius, atomic number, first ionization energy, electronegativity, atomic volume, and dopant content of the dopant element. The dopant element types and concentrations of the coating in the experimental database of step S1 are used as the input dataset, and the mechanical, electrochemical, and tribological performance indicators of the coating, namely the ratio of hardness to elastic modulus, friction coefficient, and charge transfer resistance, are used as the output test set.
[0016] Furthermore, the machine learning algorithm described in step S3 is the support vector machine method, and the training parameters include 100 training rounds, a learning rate of 0.03, an L2 regularization coefficient of 5, and five-fold cross-validation is used to evaluate the model performance.
[0017] Furthermore, the optimization objective function in step S4 is to maximize the ratio of hardness to elastic modulus, maximize the charge transfer resistance, and minimize the friction coefficient and corrosion current density. A genetic algorithm is used for global search, and the specific optimization objective function is as follows:
[0018] S=αH / E+βR ct +γμ+ωI corr
[0019] Where α, β, γ, and ω are weighting coefficients, H / E is the ratio of hardness to elastic modulus, and R... ct I is the charge transfer resistance value, μ is the coefficient of friction, and I is the resistance value. corr This represents the corrosion current density.
[0020] Furthermore, the DLC coating deposition technology used in step S5 is a DC pulsed PECVD process, in which doping elements are introduced in the form of precursor gas or target material.
[0021] Furthermore, the doped DLC coating prepared in step S5 has a hardness higher than 15 GPa, a friction coefficient lower than 0.15, and a corrosion current density lower than 10 in a 3.5 wt% NaCl solution. -7 A / cm 2 .
[0022] Compared with the prior art, the present invention has the following beneficial effects:
[0023] (1) The DLC coating doping optimization method that integrates support vector machine and genetic algorithm described in this invention constructs a multi-performance prediction model based on machine learning, which realizes the performance prediction and screening of common doping elements in the periodic table based on only some experiments or simulation calculations, which greatly reduces the time and computational cost required for material design.
[0024] (2) The DLC coating doping optimization method integrating support vector machine and genetic algorithm described in this invention is based on the influence of doping elements on the mechanical properties, tribological properties and electrochemical properties of DLC coating, and selects a series of element combinations (such as Si, Ti, W, Cr, N, etc.) that help improve the overall performance of the coating, thus guiding the high-performance doping design of DLC coating.
[0025] (3) The DLC coating doping optimization method integrating support vector machine and genetic algorithm described in this invention introduces a multi-objective optimization strategy, specifically aiming to improve the hardness / elastic modulus ratio H / E and the charge transfer resistance R. ct Reduce friction coefficient and corrosion current density I corr To achieve the design goals, the DLC coating's "high hardness - low friction - strong corrosion resistance" performance was synergistically optimized.
[0026] (4) The DLC coating doping optimization method integrating support vector machine and genetic algorithm described in this invention, combined with experimental verification results, shows that the optimized recommended doping conditions can reduce the friction coefficient of the DLC coating to below 0.15 and the corrosion current density to as low as 10. -7 The order of magnitude, while maintaining a hardness of no less than 15 GPa, fully demonstrates its high-performance advantages.
[0027] (5) The DLC coating doping optimization method integrating support vector machine and genetic algorithm described in this invention provides an efficient, scalable and iterative intelligent design method for the performance optimization and rapid development of doped DLC coatings by deeply integrating artificial intelligence methods with coating material design. It has important engineering application prospects and industrialization value. Attached Figure Description
[0028] Figure 1 This is a flowchart of the DLC coating doping optimization method that integrates support vector machine and genetic algorithm as described in this invention;
[0029] Figure 2 The image shows the electrochemical performance of corrosion-resistant Si-doped DLC optimized using a machine learning prediction model in Example 1.
[0030] Figure 3 The image shows the tribological properties of wear-resistant W-doped DLC optimized using a machine learning prediction model in Example 2. Detailed Implementation
[0031] 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 specific embodiments. It should be understood that the specific embodiments described herein are only for explaining the invention and are not intended to limit the invention; that is, the described specific embodiments are merely a part of the embodiments of the invention, and not all of them. The components of the specific embodiments of the invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations, and the invention may also have other embodiments.
[0032] Therefore, the following detailed description of specific embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected specific embodiments of the invention. All other specific embodiments obtained by those skilled in the art based on these specific embodiments without inventive effort are within the scope of protection of this invention. To further understand the inventive content, features, and effects of this invention, the following specific embodiments are provided in conjunction with… Figures 1-3 Detailed explanation is as follows:
[0033] A method for optimizing the doping of DLC coatings by integrating support vector machines and genetic algorithms includes the following steps:
[0034] Step S1: Construct an experimental database of doped DLC coatings. The data in the experimental database includes the types and concentrations of doping elements in the coating, as well as the mechanical, electrochemical, and tribological properties of the coating, namely the ratio of hardness to elastic modulus, coefficient of friction, and charge transfer resistance.
[0035] Step S2: Based on the experimental database from Step S1, perform feature engineering and data preprocessing to extract feature parameters, doping element properties, process variables, and test results to construct input and output datasets.
[0036] Step S3: Construct a performance prediction model. The input and output datasets obtained in step S2 are used as training sets. A machine learning algorithm is used, and feature parameters are extracted as input features of the machine learning model to train the coating performance prediction model and obtain the trained coating performance prediction model.
[0037] Step S4: Construct an optimization objective function. The predicted coating performance indicators, including the ratio of hardness to elastic modulus, friction coefficient, and charge transfer resistance, are output by the coating performance prediction model trained in Step S3 as target variables. A genetic algorithm is then introduced to optimize and obtain the optimal dopant element content.
[0038] Step S5: Using DLC coating deposition technology, a doped DLC coating is prepared based on the optimal doping element content in step S4. The performance is verified, and the verification results are used to update the database and iteratively optimize the model and design results.
[0039] Furthermore, the types of doping elements mentioned in step S1 include Si, N, Ag, F, Ti, Cu, B, Cr, W, and Mo.
[0040] Furthermore, the characteristic parameters in step S2 include the atomic radius, atomic number, first ionization energy, electronegativity, atomic volume, and dopant content of the dopant element. The dopant element types and concentrations of the coating in the experimental database of step S1 are used as the input dataset, and the mechanical, electrochemical, and tribological performance indicators of the coating, namely the ratio of hardness to elastic modulus, friction coefficient, and charge transfer resistance, are used as the output test set.
[0041] Furthermore, the machine learning algorithm described in step S3 is the support vector machine method, and the training parameters include 100 training rounds, a learning rate of 0.03, an L2 regularization coefficient of 5, and five-fold cross-validation is used to evaluate the model performance.
[0042] Furthermore, the optimization objective function in step S4 is to maximize the ratio of hardness to elastic modulus, maximize the charge transfer resistance, and minimize the friction coefficient and corrosion current density. A genetic algorithm is used for global search, and the specific optimization objective function is as follows:
[0043] S=αH / E+βR ct +γμ+ωI corr
[0044] Where α, β, γ, and ω are weighting coefficients, H / E is the ratio of hardness to elastic modulus, and R... ct I is the charge transfer resistance value, μ is the coefficient of friction, and I is the resistance value. corr This represents the corrosion current density.
[0045] Furthermore, the DLC coating deposition technology used in step S5 is a DC pulsed PECVD process, in which doping elements are introduced in the form of precursor gas or target material.
[0046] Furthermore, the doped DLC coating prepared in step S5 has a hardness higher than 15 GPa, a friction coefficient lower than 0.15, and a corrosion current density lower than 10 in a 3.5 wt% NaCl solution. -7 A / cm 2 .
[0047] Example 1
[0048] A method for optimizing the doping of DLC coatings by integrating support vector machines and genetic algorithms includes the following steps:
[0049] S1. Construction of experimental data and extraction of periodic table features for doped DLC coatings
[0050] Collect data on the doping concentration, structural parameters (bonding structure, density, film thickness), and performance indicators (hardness, elastic modulus, coefficient of friction, and charge transfer resistance R in electrochemical impedance) of DLC coatings from existing experiments. ct Corrosion current density I corr (etc.). Extract periodic table properties of doped elements, such as atomic radius, electronegativity, and ionization energy, and combine them with element concentration parameters as input features for machine learning models.
[0051] S2. Construction of DLC Coating Performance Prediction Model Based on Support Vector Machine
[0052] The Support Vector Machine (SVM) algorithm was used to train the model, taking elemental properties, process and structural features as inputs and corrosion resistance performance indicators as outputs. During training, cross-validation was used to tune the parameters and evaluate the model's generalization ability.
[0053] S3. Dopant element screening and scheme recommendation based on model prediction
[0054] By inputting the feature vectors of different elements into the trained model, the corrosion resistance of DLC coatings under different concentration conditions is predicted. The optimal concentration range of each doping element is selected, and with improving corrosion resistance as the core, a recommended scheme for corrosion-resistant components of doped DLC coatings is obtained.
[0055] S4. Global Optimization Design of Doping and Process Parameters Based on Genetic Algorithm
[0056] The objective function is set to maximize the charge transfer resistance R. ct Minimize corrosion current density I corr Based on a genetic algorithm, a global search of process parameters was performed, and the optimal doping element was found to be Si with a doping concentration of 8%.
[0057] S5. Experimental Preparation and Performance Verification
[0058] Si-doped DLC coatings of 4–10% were prepared using magnetron sputtering under optimized process conditions. The corrosion resistance of the coatings was tested by electrochemical impedance spectroscopy (EIS) and polarization curves. Figure 2 As shown, Si4-Si10 are Si-DLC coatings with Si doping contents of 4%-10% respectively. It can be observed that the optimized Si-DLC coating has a corrosion current density I0. corr It significantly reduces corrosion and exhibits significantly better corrosion resistance than the YG8 substrate material.
[0059] Example 2
[0060] The difference between this embodiment and Embodiment 1 is that the performance target is tribological performance;
[0061] A method for optimizing the doping of DLC coatings by integrating support vector machines and genetic algorithms includes the following steps:
[0062] S1. Construction of experimental data and extraction of periodic table features for doped DLC coatings
[0063] Collect data on the doping concentration, structural parameters (bonding structure, density, film thickness), and performance indicators (hardness, elastic modulus, coefficient of friction, and charge transfer resistance R in electrochemical impedance) of DLC coatings from existing experiments. ct Corrosion current density I corr (etc.). Extract periodic table properties of doped elements, such as atomic radius, electronegativity, and ionization energy, and combine them with element concentration parameters as input features for machine learning models.
[0064] S2. Construction of DLC Coating Performance Prediction Model Based on Support Vector Machine
[0065] The Support Vector Machine (SVM) algorithm was used for model training, with elemental attributes, process and structural features as inputs, and mechanical performance indicators (hardness / elastic modulus ratio H / E, coefficient of friction) as outputs. During training, cross-validation was used to tune the parameters and evaluate the model's generalization ability.
[0066] S3. Dopant element screening and scheme recommendation based on model prediction
[0067] By inputting the feature vectors of different elements into the trained model, the tribological properties of DLC coatings under different concentration conditions are predicted. The optimal concentration range of each doping element is selected, and with improving tribological properties as the core, a recommended scheme for wear-resistant components of doped DLC coatings is obtained.
[0068] S4. Global Optimization Design of Doping and Process Parameters Based on Genetic Algorithm
[0069] The objective function was set to maximize the hardness / elastic modulus ratio H / E and minimize the friction coefficient. Based on the genetic algorithm, a global search of the process parameters was performed, and the optimal doping element was found to be W with a doping concentration of 10%.
[0070] S5. Experimental Preparation and Performance Verification
[0071] 6–12% W-doped DLC coatings were prepared using magnetron sputtering under optimized process conditions. The tribological properties of the coatings were tested through linear reciprocating friction experiments, such as… Figure 3As shown, W6-W12 are W-DLC coatings with W doping content of 6%-12% respectively. It can be found that the optimized W-DLC coating has a significantly reduced friction coefficient and its wear resistance is significantly better than that of the substrate YG8 material.
[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A DLC coating doping optimization method integrating support vector machine and genetic algorithm, characterized in that, Includes the following steps: Step S1: Construct an experimental database of doped DLC coatings. The data in the experimental database includes the types and concentrations of doping elements in the coating, as well as the mechanical, electrochemical, and tribological properties of the coating, namely the ratio of hardness to elastic modulus, coefficient of friction, and charge transfer resistance. Step S2: Based on the experimental database from Step S1, perform feature engineering and data preprocessing to extract feature parameters, doping element properties, process variables, and test results to construct input and output datasets. Step S3: Construct a performance prediction model. The input and output datasets obtained in step S2 are used as training sets. A machine learning algorithm is used, and the feature parameters are used as input features of the machine learning model to train the coating performance prediction model, thereby obtaining the trained coating performance prediction model. Step S4: Construct an optimization objective function. The predicted coating performance indicators, including the ratio of hardness to elastic modulus, friction coefficient, and charge transfer resistance, are output by the coating performance prediction model trained in Step S3 as target variables. A genetic algorithm is then introduced to optimize and obtain the optimal dopant element content. Step S5: Using DLC coating deposition technology, a doped DLC coating is prepared based on the optimal doping element content in step S4. The performance is verified, and the verification results are used to update the database and iteratively optimize the model and design results.
2. The DLC coating doping optimization method integrating support vector machine and genetic algorithm according to claim 1, characterized in that, The types of doping elements mentioned in step S1 include Si, N, Ag, F, Ti, Cu, B, Cr, W, and Mo.
3. The DLC coating doping optimization method integrating support vector machine and genetic algorithm according to claim 1, characterized in that, The characteristic parameters in step S2 include the atomic radius, atomic number, first ionization energy, electronegativity, atomic volume, and dopant content of the dopant element. The dopant element types and concentrations of the coating in the experimental database of step S1 are used as the input dataset, and the mechanical, electrochemical, and tribological performance indicators of the coating, namely the ratio of hardness to elastic modulus, friction coefficient, and charge transfer resistance, are used as the output test set.
4. The DLC coating doping optimization method integrating support vector machine and genetic algorithm according to claim 1, characterized in that, The machine learning algorithm described in step S3 is the support vector machine method. The training parameters include 100 training rounds, a learning rate of 0.03, an L2 regularization coefficient of 5, and five-fold cross-validation to evaluate the model performance.
5. The DLC coating doping optimization method integrating support vector machine and genetic algorithm according to claim 1, characterized in that, The optimization objective function described in step S4 is to maximize the ratio of hardness to elastic modulus, maximize the charge transfer resistance, and minimize the friction coefficient and corrosion current density. A genetic algorithm is used for global search. The specific optimization objective function is as follows: S=αH / E+βR ct +γμ+ωI corr Where α, β, γ, and ω are weighting coefficients, H / E is the ratio of hardness to elastic modulus, and R... ct I is the charge transfer resistance value, μ is the coefficient of friction, and I is the resistance value. corr This represents the corrosion current density.
6. The DLC coating doping optimization method integrating support vector machine and genetic algorithm according to claim 1, characterized in that, The DLC coating deposition technology used in step S5 is a DC pulse PECVD process, in which doping elements are introduced in the form of precursor gas or target material.
7. The DLC coating doping optimization method integrating support vector machine and genetic algorithm according to any one of claims 1 to 6, characterized in that, The doped DLC coating prepared in step S5 has a hardness higher than 15 GPa, a friction coefficient lower than 0.15, and a corrosion current density lower than 10 in a 3.5 wt% NaCl solution. -7 A / cm 2 .