Method and apparatus for designing co-crystal high-entropy alloys based on machine learning and uncertainty assessment
By combining machine learning and uncertainty assessment methods, the complexity of high-entropy alloy composition design was solved, enabling accurate prediction and rapid development of eutectic high-entropy alloys, and producing eutectic high-entropy alloys that meet the requirements.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHANGHAI UNIV
- Filing Date
- 2024-12-06
- Publication Date
- 2026-06-09
AI Technical Summary
The composition design of high-entropy alloys is complex, and existing technologies make it difficult to systematically summarize their phase formation rules and empirical parameters, which leads to difficulties in identifying eutectic composition and crystal structure. Machine learning has not yet been fully utilized in the design of high-entropy alloys.
By combining machine learning classification models and uncertainty assessment, the composition and crystal structure data of high-entropy alloys are collected, machine learning models are trained, virtual composition spaces are designed, low-uncertainty eutectic high-entropy alloy compositions are screened out, and alloys are prepared by vacuum arc furnace melting.
It enables accurate prediction and rapid development of eutectic high-entropy alloys, and produces eutectic high-entropy alloys that meet the requirements, thereby improving the accuracy and efficiency of design.
Smart Images
Figure CN122177293A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of metallic material design technology, specifically to a method and apparatus for designing eutectic high-entropy alloys based on machine learning and uncertainty assessment. Background Technology
[0002] High-performance metallic materials can meet the demands of high-end manufacturing sectors for strength and plasticity, providing strong support for the development of aerospace and automotive manufacturing, and occupying an important strategic position in the field of new materials. Eutectic high-entropy alloys are eutectic alloys composed of multiple main elements, possessing a balance of high strength and high plasticity in their mechanical properties. In recent years, they have attracted widespread attention due to their unique microstructure, excellent performance, and potential for further development.
[0003] Due to the presence of multiple key elements, high-entropy alloys exhibit an exponentially increasing combination pattern. Furthermore, current research on the performance mechanisms of high-entropy alloys is incomplete, making it impossible to systematically summarize their phase formation rules and the influence of empirical parameters on phase formation. The vast compositional space and lack of comprehensive design standards further complicate the design of high-entropy alloys, making the prediction of phase formation exceptionally difficult. Therefore, the systematic identification of eutectic composition and crystal structure in specific alloy systems remains challenging.
[0004] With the emergence of artificial intelligence, machine learning technology, through data-driven alloy design, has become a promising solution to overcome these obstacles. Based on material data, machine learning can build models of specific material properties, gain a deep understanding of the relationship between material structure and physical properties, and ultimately establish a model that can quickly predict the target properties of undeveloped materials. Currently, machine learning is being effectively utilized in the design of high-entropy alloys and is becoming a powerful tool for materials innovation. For example, Chinese patent CN117236189A discloses a method and apparatus for rapidly designing L12 phase-strengthened single-crystal high-entropy alloys based on machine learning. By establishing a machine learning model, the composition of L12 phase-strengthened high-entropy alloys is predicted, leading to the development of high-performance single-crystal high-entropy alloys. Chinese patent CN110010210A discloses a multi-component alloy composition design method based on machine learning and oriented towards performance requirements. It utilizes machine learning methods to construct an implicit and complex relationship between "composition and performance," achieving the goal of accurately and rapidly designing alloy compositions according to performance requirements. Summary of the Invention
[0005] This invention provides a method and apparatus for designing eutectic high-entropy alloys based on machine learning and uncertainty assessment, aiming to combine machine learning classification models and uncertainty assessment to design eutectic high-entropy alloys.
[0006] The above-mentioned technical objective of this invention is achieved through the following technical solution: a eutectic high-entropy alloy design method based on machine learning and uncertainty assessment, comprising the following steps:
[0007] S1: Collect composition data and crystal structure data of high-entropy alloys. The composition includes the elements that make up the alloy and their content. Through data cleaning and classification, establish a high-entropy alloy crystal structure dataset.
[0008] S2: Using the high-entropy alloy crystal structure dataset, with the composition of the high-entropy alloy as the input data and the crystal structure as the output data, a machine learning classification algorithm is used for training to obtain a trained machine learning classification model;
[0009] S3: Design a virtual composition space and use a machine learning classification model to predict it to obtain the predicted crystal structure; the virtual composition space is a sample group composed of high-entropy alloys to be designed that conform to the composition within a predetermined range.
[0010] S4: The predicted crystal structure is evaluated in conjunction with an entropy-based uncertainty model, and the composition of the eutectic high-entropy alloy with low uncertainty is selected based on the calculation results;
[0011] The calculation formula for the uncertainty model is as follows:
[0012] H = -∑ i,j P i p i (j)logp i (j), where i represents the i-th crystal structure combination, j represents the j-th high-entropy alloy composition in the virtual composition space, P i p represents the probability of generating the i-th crystal structure combination. i (j) represents the probability that the j-th high-entropy alloy composition produces the ith crystal structure combination, and H represents the entropy of the j-th high-entropy alloy composition;
[0013] The term "low uncertainty" refers to H being lower than a preset value.
[0014] According to a specific embodiment of the present invention, the crystal structures of the high-entropy alloy crystal structure dataset are divided into four categories: face-centered cubic (FCC), body-centered cubic (BCC), multiphase eutectic crystal structure, and multiphase non-eutectic crystal structure.
[0015] According to a specific embodiment of the present invention, in step S2, multiple machine learning classification algorithms are used for training, and the training results of multiple learning models are compared to obtain the machine learning classification model with the best training effect. For example, the machine learning classification algorithm is a Support Vector Machine (SVM) algorithm, a Random Forest (RF) algorithm, or a Multilayer Perceptron (MLP) algorithm.
[0016] According to a specific embodiment of the present invention, in step S2, the method for comparing the training results of multiple learning models includes: evaluating the training results of multiple learning models based on the area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, and F1 score.
[0017] According to a specific embodiment of the present invention, in step S3, the composition of the high-entropy alloy to be designed is Al. x Co y Cr z Fe m Ni n , where x = 0.1 - 3.0, y = 1, z = 1, m = 1, n = 0.1 - 2.0.
[0018] Furthermore, the eutectic high-entropy alloy design method based on machine learning and uncertainty assessment further includes: experimentally verifying the composition of the selected eutectic high-entropy alloy with low uncertainty.
[0019] The present invention also provides a device for designing eutectic high-entropy alloys based on machine learning and uncertainty assessment, comprising:
[0020] An acquisition module is used to acquire training data, wherein the training data includes input data and output data, wherein the input data is the composition of the high-entropy alloy and the output data is the crystal structure of the high-entropy alloy;
[0021] The training module is used to input the input data into a selected machine learning classification algorithm for training, so as to obtain a trained machine learning classification model.
[0022] The prediction module is used to input a sample group of high-entropy alloys to be designed that meet the predetermined composition range into a trained machine learning classification model to predict the crystal structure and obtain the predicted crystal structure.
[0023] The evaluation module is used to evaluate the predicted crystal structure in conjunction with an entropy-based uncertainty model, and to screen out the composition of eutectic high-entropy alloys with low uncertainty based on the calculation results.
[0024] The calculation formula for the uncertainty model is as follows:
[0025] H = -∑ i,j P i p i (j)logp i (j), where i represents the i-th crystal structure combination, j represents the j-th high-entropy alloy composition in the virtual composition space, P i p represents the probability of generating the i-th crystal structure combination. i(j) represents the probability that the j-th high-entropy alloy composition produces the ith crystal structure combination, and H represents the entropy of the j-th high-entropy alloy composition;
[0026] The term "low uncertainty" refers to H being lower than a preset value.
[0027] Furthermore, the eutectic high-entropy alloy design device based on machine learning and uncertainty assessment also includes: a verification module for experimentally verifying the composition of the eutectic high-entropy alloy with low uncertainty selected.
[0028] The present invention also provides a method for preparing a eutectic high-entropy alloy, the method comprising: preparing a eutectic high-entropy alloy based on the composition of a low-uncertainty eutectic high-entropy alloy selected by a eutectic high-entropy alloy design method based on machine learning and uncertainty assessment as described above.
[0029] According to a specific embodiment of the present invention, a eutectic high-entropy alloy is prepared by a vacuum electric arc furnace melting method.
[0030] The present invention also provides a eutectic high-entropy alloy, which is prepared by the aforementioned method for preparing a eutectic high-entropy alloy.
[0031] According to a specific embodiment of the present invention, a eutectic high-entropy alloy comprises the following elements by mass percentage:
[0032] Al: 8.03%, Co: 19.48%, Cr: 17.18%, Fe: 18.45% and Ni: 36.86%;
[0033] Or Al: 7.87%, Co: 19.11%, Cr: 16.86%, Fe: 18.10% and Ni: 38.06%.
[0034] Compared with existing technologies, this invention has the following beneficial technical effects: by combining machine learning classification models and uncertainty, accurate prediction of eutectic high-entropy alloys can be achieved, accelerating the development of advanced eutectic high-entropy alloys. Based on the predicted composition of the eutectic high-entropy alloy, a eutectic high-entropy alloy meeting the requirements can be prepared through smelting. Attached Figure Description
[0035] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings are provided to further explain the invention and constitute a part of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention. In the drawings:
[0036] Figure 1 This is a flowchart of the design method in an embodiment of the present invention;
[0037] Figure 2This is an evaluation graph of the performance of three machine learning classification models;
[0038] Figure 3 These are scanning electron microscope (SEM) images of the microstructure of the eutectic high-entropy alloy according to an embodiment of the present invention: wherein, Figure 3 (a) and Figure 3 (b) and are microstructure images of the eutectic high-entropy alloy 8.03Al-19.48Co-17.18Cr-18.45Fe-36.86Ni; Figure 3 (c) and Figure 3 (d) is a microstructure image of the 7.87Al-19.11Co-16.86Cr-18.10Fe-38.06Ni eutectic high-entropy alloy. Detailed Implementation
[0039] To make the technical solution and technical effects of the present invention clearer, the following description is provided in conjunction with the appendix. Figure 1-3 The present invention will be further described in detail below with reference to the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them; all other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0040] This invention provides a method for designing eutectic high-entropy alloys based on machine learning and uncertainty assessment, such as... Figure 1 As shown, the method includes the following steps:
[0041] S1. Establishing a High-Entropy Alloy Crystal Structure Dataset: A literature review of high-entropy alloys was conducted to collect compositional and crystal structure data. Through data cleaning and classification, a high-entropy alloy crystal structure dataset was established. The crystal structures in the dataset are divided into four categories: face-centered cubic (FCC), body-centered cubic (BCC), multiphase eutectic crystal structure, and multiphase non-eutectic crystal structure. The composition includes the elements composing the alloy and their content.
[0042] S2, Training and evaluating machine learning classification models based on different algorithms: Using the high-entropy alloy crystal structure dataset, machine learning classification models were trained using Support Vector Machine (SVM), Random Forest (RF), and Multilayer Perceptron (MLP) algorithms. The composition of the high-entropy alloy was used as input data, and four crystal structures—FCC, BCC, multiphase eutectic crystal structure, and multiphase non-eutectic crystal structure—were used as output data. The classification model performance was evaluated based on the area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, and F1 score. The machine learning classification model with the best overall performance was selected for the next step.
[0043] S3, Design a virtual composition space and use a machine learning classification model for prediction: Design a virtual composition space and use a machine learning classification model to predict it to obtain the predicted crystal structure; the virtual composition space is a sample group composed of high-entropy alloys to be designed that conform to the composition within a predetermined range.
[0044] S4. Evaluate using an uncertainty model to select eutectic high-entropy alloys: Evaluate the predicted crystal structure using an entropy-based uncertainty model, and select the composition of eutectic high-entropy alloys with low uncertainty based on the calculation results.
[0045] Entropy quantifies the uncertainty in phase prediction caused by compositional variations, and the optimal alloy composition between the Cr content and the ideal eutectic high-entropy alloy microstructure is determined using entropy-based uncertainty. The calculation formula for the uncertainty model is as follows:
[0046] H = -∑ i,j P i p i (j)logp i (j), where i represents the i-th crystal structure combination, j represents the j-th high-entropy alloy composition in the virtual composition space, P i p represents the probability of generating the i-th crystal structure combination. i (j) represents the probability that the j-th high-entropy alloy composition produces the ith crystal structure combination, and H represents the entropy of the j-th high-entropy alloy composition;
[0047] The term "low uncertainty" refers to H being lower than a preset value.
[0048] Example 1
[0049] This invention provides a method for designing eutectic high-entropy alloys based on machine learning and uncertainty assessment, the method comprising the following steps:
[0050] S1: A literature review was conducted on high-entropy alloys in the Al-Co-Cr-Fe-Ni system, collecting compositional and crystal structure data for 256 samples. Through data cleaning and classification, a high-entropy alloy crystal structure dataset was established. The crystal structures in the dataset were divided into four categories: face-centered cubic (FCC), body-centered cubic (BCC), multiphase eutectic crystal structure, and multiphase non-eutectic crystal structure.
[0051] S2: The dataset is divided into an 80% training set and a 20% test set. The elemental compositions of Al, Co, Cr, Fe, and Ni are used as input data, and four crystal structures—FCC, BCC, multiphase eutectic crystal structure, and multiphase non-eutectic crystal structure—are used as output data. A machine learning classification model is trained using Support Vector Machine (SVM), Random Forest (RF), and Multilayer Perceptron (MLP) algorithms. The performance of the classification model is evaluated based on the Area Under the Receiver Operating Characteristic (AUC), accuracy, precision, recall, and F1 score. Figure 2 This is a performance evaluation chart of three machine learning classification models. The multilayer perceptron (MLP) machine learning classification model with the best overall performance will be selected for the next step.
[0052] S3: A virtual composition space was constructed to explore potential novel alloys. This virtual composition space has the molecular formula Al. x Co y Cr z Fe m Ni n A sample group consisting of high-entropy alloys was constructed, where x ranged from 0.1 to 3.0, n ranged from 0.1 to 2.0, the increments of x and n were 0.1, and the values of y, z, and m were 1. This method generated a virtual composition space containing 600 potential high-entropy alloys, and predicted their crystal structures using the established multilayer perceptron (MLP) machine learning classification model.
[0053] S4: The predicted crystal structure is evaluated using an entropy-based uncertainty model, the calculation formula of which is as follows:
[0054] H = -∑ i,j P i p i (j)logp i (j), where i represents the i-th crystal structure combination, j represents the j-th high-entropy alloy composition in the virtual composition space, P i p represents the probability of generating the i-th crystal structure combination. i (j) represents the probability that the j-th high-entropy alloy composition produces the ith crystal structure combination, and H represents the entropy of the j-th high-entropy alloy composition;
[0055] The term "low uncertainty" refers to H being lower than a preset value.
[0056] Based on the calculation results, eutectic high-entropy alloy compositions with low uncertainty multiphase eutectic crystal structures were selected, such as the following two: 8.03Al-19.48Co-17.18Cr-18.45Fe-36.86Ni (i.e., Al 0.9 CoCrFeNi 1.9) and 7.87Al-19.11Co-16.86Cr-18.10Fe-38.06Ni (i.e., Al 0.9 CoCrFeNi2). 8.03Al-19.48Co-17.18Cr-18.45Fe-36.86Ni, comprising the following elements by mass percentage: Al: 8.03%, Co: 19.48%, Cr: 17.18%, Fe: 18.45%, and Ni: 36.86%. 7.87Al-19.11Co-16.86Cr-18.10Fe-38.06Ni, comprising the following elements by mass percentage: Al: 7.87%, Co: 19.11%, Cr: 16.86%, Fe: 18.10%, and Ni: 38.06%.
[0057] Experimental verification:
[0058] 1. Alloys 8.03Al-19.48Co-17.18Cr-18.45Fe-36.86Ni and 7.87Al-19.11Co-16.86Cr-18.10Fe-38.06Ni were prepared using a vacuum electric arc furnace melting method. Specifically, high-purity Al, Co, Cr, Fe, and Ni were used as raw materials (purity 99.99 wt.%). The raw materials were melted into alloys in an argon atmosphere according to the mass percentage of each element in the alloy. The molten alloy was then cast into a water-cooled copper mold and the melting process was repeated eight times.
[0059] 2. X-ray diffraction (XRD) revealed that both the 8.03Al-19.48Co-17.18Cr-18.45Fe-36.86Ni and 7.87Al-19.11Co-16.86Cr-18.10Fe-38.06Ni alloys contain both the FCC(L12) phase and the B2 phase. The microstructure and elemental composition of the alloys were analyzed using a scanning electron microscope (SEM) equipped with a backscattered electron detector (BSE) and an energy dispersive spectroscopy (EDS) instrument. Figure 3 (a) Figure 3 (b) and Figure 3 (c) Figure 3As shown in (d), both alloys exhibit a distinct alternating two-phase distribution. Besides the relatively uniform, typical petal-shaped eutectic structure, there are also scattered eutectic structures composed of regular, straight, plate-like lamellae. Surface scanning spectroscopy was then used to observe the elemental distribution. The surface scan images showed that the matrix was enriched in Al and Ni; the coarse lamellae in the eutectic structure were enriched in Fe and Cr, tending to form the FCC phase, while the fine lamellae were enriched in Al and Ni, tending to form the B2 phase. Finally, transmission electron microscopy was used to analyze the crystal structure and microstructure of the alloys. The results showed that both the 8.03Al-19.48Co-17.18Cr-18.45Fe-36.86Ni alloy and the 7.87Al-19.11Co-16.86Cr-18.10Fe-38.06Ni alloy possessed FCC (L12) and BCC (B2) crystal structures. The above microstructure characterization shows that the 8.03Al-19.48Co-17.18Cr-18.45Fe-36.86Ni and 7.87Al-19.11Co-16.86Cr-18.10Fe-38.06Ni alloys are both eutectic high-entropy alloys with an FCC(L12)+BCC(B2) crystal structure.
[0060] Therefore, the design method of this invention can achieve accurate prediction of eutectic high-entropy alloys, thus accelerating the development of advanced eutectic high-entropy alloys.
Claims
1. A design method for eutectic high-entropy alloys based on machine learning and uncertainty assessment, characterized in that, Includes the following steps: S1: Collect composition data and crystal structure data of high-entropy alloys. The composition includes the elements that make up the alloy and their content. Through data cleaning and classification, establish a high-entropy alloy crystal structure dataset. S2: Using the high-entropy alloy crystal structure dataset, with the composition of the high-entropy alloy as the input data and the crystal structure as the output data, a machine learning classification algorithm is used for training to obtain a trained machine learning classification model; S3: Design a virtual composition space and use a machine learning classification model to predict it to obtain the predicted crystal structure; the virtual composition space is a sample group composed of high-entropy alloys to be designed that conform to the composition within a predetermined range. S4: The predicted crystal structure is evaluated in conjunction with an entropy-based uncertainty model, and the composition of the eutectic high-entropy alloy with low uncertainty is selected based on the calculation results; The calculation formula for the uncertainty model is as follows: H = -∑ i,j P i p i (j)logp i (j), where i represents the i-th crystal structure combination, j represents the j-th high-entropy alloy composition in the virtual composition space, P i p represents the probability of generating the i-th crystal structure combination. i (j) represents the probability that the j-th high-entropy alloy composition produces the ith crystal structure combination, and H represents the entropy of the j-th high-entropy alloy composition; The term "low uncertainty" refers to H being lower than a preset value.
2. The design method according to claim 1, characterized in that, The crystal structures in the high-entropy alloy crystal structure dataset are divided into four categories: face-centered cubic structure, body-centered cubic structure, multiphase eutectic crystal structure, and multiphase non-eutectic crystal structure.
3. The design method according to claim 1, characterized in that, In step S2, multiple machine learning classification algorithms are used for training, and the training results of multiple learning models are compared to obtain the machine learning classification model with the best training effect.
4. The design method according to claim 3, characterized in that, The machine learning classification algorithm is a support vector machine algorithm, a random forest algorithm, or a multilayer perceptron algorithm.
5. The design method according to claim 3, characterized in that, In step S2, the method for comparing the training results of multiple learning models includes: evaluating the training results of multiple learning models based on the area under the receiver operating feature curve, accuracy, precision, recall, and F1 score.
6. The design method according to claim 1, characterized in that, In step S3, the composition of the high-entropy alloy to be designed is Al. x Co y Cr z Fe m Ni n , where x = 0.1 - 3.0, y = 1, z = 1, m = 1, n = 0.1 - 2.
0.
7. The design method according to any one of claims 1-6, characterized in that, Also includes: The composition of the eutectic high-entropy alloy with low uncertainty was experimentally verified.
8. A device for designing eutectic high-entropy alloys based on machine learning and uncertainty assessment, characterized in that, include: An acquisition module is used to acquire training data, wherein the training data includes input data and output data, wherein the input data is the composition of the high-entropy alloy and the output data is the crystal structure of the high-entropy alloy; The training module is used to input the input data into a selected machine learning classification algorithm for training, so as to obtain a trained machine learning classification model. The prediction module is used to input a sample group of high-entropy alloys to be designed that meet the predetermined composition range into a trained machine learning classification model to predict the crystal structure and obtain the predicted crystal structure. The evaluation module is used to evaluate the predicted crystal structure in conjunction with an entropy-based uncertainty model, and to screen out the composition of eutectic high-entropy alloys with low uncertainty based on the calculation results. The calculation formula for the uncertainty model is as follows: H = -∑ i,j P i p i (j)logp i (j), where i represents the i-th crystal structure combination, j represents the j-th high-entropy alloy composition in the virtual composition space, P i p represents the probability of generating the i-th crystal structure combination. i (j) represents the probability that the j-th high-entropy alloy composition produces the ith crystal structure combination, and H represents the entropy of the j-th high-entropy alloy composition; The term "low uncertainty" refers to H being lower than a preset value.
9. The design apparatus according to claim 8, characterized in that, Also includes: The verification module is used to experimentally verify the composition of the selected eutectic high-entropy alloys with low uncertainty.
10. A eutectic high-entropy alloy, characterized in that, The eutectic high-entropy alloy comprises the following elements by mass percentage: Al: 8.03%, Co: 19.48%, Cr: 17.18%, Fe: 18.45% and Ni: 36.86%; Or Al: 7.87%, Co: 19.11%, Cr: 16.86%, Fe: 18.10% and Ni: 38.06%.