AI是如何发现新型钴基高温合金的？

新型γ/γ' 钴基高温合金由于其独特的双相组织和更高的承温能力在下一代航空航天热端部件材料中被认为是传统镍基高温合金的最佳候选。但在面对数以千万计钴基合金体系的搜索中，传统方法经常面临开发周期长和实验试错成本高的问题。

来自中国厦门大学许伟伟和哈尔滨工业大学（深圳）刘兴军团队，提出了将DFT计算与机器学习模型相结合的搜索方法，可用DFT计算的能量数据来构建模型，从而提取有关能量和γ'相关系的完整信息，包括影响因素和竞争相，以及对超150000的合金体系进行快速的搜索，寻找出更多潜在的合金体系。

他们发现了更多新型的γ/γ'钴基高温合金，并通过实验合成了两种新型合金，验证了机器学习模型在快速搜索新型合金的可行性以及模型预测的可靠性。

Fig. 3 | Comparison of accuracy (R2) and error (MAE) of 13 machine learning models.

作者除了通过可靠的预测模型获取到更多新型钴基合金体系的信息外，还揭示了影响相合成和稳定性的因素，以及可能的竞争相，初步揭示了部分添加元素对γ'相的影响机制，与现有的部分研究结果一致：

Fig. 4 | Importance weights analysis of features from Random Forest and feature elimination by using RFECV5.

1)训练的随机森林模型实现了形成能(H_f)预测精度98.07%和97.05%的分解能(H_d)预测精度。

2)Ni、Nb、Ta、Ti和V等元素增强了γ'相的稳定性，而Mo、W和Al通过增加分解能(H_d)对稳定性产生负面影响。

3) 确定了1,049种有前途的候选物，主要分布在11个含铝和25个非铝合金体系中。实验表征了其中两种最佳体系，两种合金的 γ' 相稳定性超出了预期，即使在高温和长期时效处理下也能保持稳定。两种合金的最小密度约为7.90 g/cm³，优于大多数现有的钴基合金。

Fig. 7 | Experimental verification of the U01 and U02 includes CALPHAD evaluation, X-ray diffraction (XRD), and electron microscopy images (SEM).

该研究展示了机器学习在合金设计中的优越性，可以极大的加快新型γ/γ'钴基高温合金体系的发现。相关论文近期发布于npj Computational Materials 10: 259 (2024)。手机阅读原文，请点击本文底部左下角“阅读原文”，进入后亦可下载全文PDF文件。

Editorial Summary

New γ/γ' Cobalt-based superalloys are emerging as promising candidates for high-temperature components in next-generation aerospace applications, offering unique dual-phase microstructures and higher temperature resistance than traditional Nickel-based alloys. However, the vast number of potential Cobalt alloy systems--literally in the tens of millions--has posed significant challenges, with traditional development approaches often bogged down by lengthy timelines and costly trial-and-error processes.

A team led by Weiwei Xu from Xiamen University and Xingjun Liu from Harbin Institute of Technology, China, has introduced an innovative approach that combines Density Functional Theory (DFT) calculations with machine learning models. This method leverages DFT energy data to build model that can extract critical insights into energy relationships and the γ' phase. As a result, they achieved rapid screening of over 150,000 potential alloy systems, uncovering new candidates with the potential to outperform existing superalloys.Theyidentified several novel γ/γ' cobalt-based alloys and successfully synthesized two of these in experiments, validating the machine learning model’s effectiveness in quickly discovering new alloy candidates and confirming the accuracy of its predictions. In addition to identifying new cobalt-based alloy systems through a robust predictive model, this study provides critical insights into factors affecting phase synthesis, stability, and potential competing phases, revealing key mechanisms by which certain alloying elements influence the γ' phase. Consistent with some existing studies: (a) the Random Forest model achieved prediction accuracies of 98.07% for formation energy (H_f) and 97.05% for decomposition energy (H_d); (b) Elements like Ni, Nb, Ta, Ti, and V were found to enhance γ' phase stability, while Mo, W, and Al had a destabilizing effect by increasing the decomposition energy (H_d); (c) the study identified 1,049 promising candidates, including 11 aluminum-containing and 25 aluminum-free alloy systems. Two of the most optimal systems were experimentally characterized, with their γ' phase stability exceeding expectations—even after prolonged high-temperature aging. 

These alloys showed a minimum density around 7.90 g/cm³, surpassing most existing cobalt-based alloys in weight efficiency. This work underscores the power of machine learning in alloy design, substantially accelerating the discovery of novel γ/γ' cobalt-based superalloys. This article was recently published in npj Computational Materials 10: 259 (2024).

原文Abstract及其翻译

Facilitated the discovery of new γ/γ′ Co-based superalloys by combining first-principles and machine learning (机器学习和第一性原理协同加速新型钴基高温合金的发现) 

Abstract Superalloys are indispensable materials for the fabrication of high-temperature components in aircraft engines. The discovery of a novel class of γ/γ′ Co-Al-W alloys has ignited a surge of interest in Co-based superalloys, with the aspiration to transcend the inherent constraints of their Ni-based counterparts. However, the conventional methodologies utilized in the design and advancement of new γ/γ′ Co-based superalloys are frequently characterized by their laborious and resource-intensive nature. In this study, we employed a coupled Density Functional Theory (DFT) and machine learning (ML) approach to predict and analyze the stability of the crucial γ′ phase, which is instrumental in expediting the discovery of γ/γ′ Co-based alloys. A dataset comprised of thousands of reliable formation (H_f) and decomposition (H_d) energies was obtained through high-throughput DFT calculations. Through regression model selection and feature engineering, our trained Random Forest (RF) model achieved prediction accuracies of 98.07% for H_f and 97.05% for H_d. Utilizing the well-trained RF model, we predicted the energies of over 150,000 ternary and quaternary γ′ phases within the Co-Ni-Fe-Cr-Al-W-Ti-Ta-V-Mo-Nb system. The energy analyses revealed that the presence of Ni, Nb, Ta, Ti, and V significantly reduced the H_f and the H_d of γ′, while Mo and W deteriorate the stability by increasing both energy values. Interestingly, although Al reduces the H_f, it increases H_d, thereby adversely affecting the stability of γ′. Applying domain-specific screening based on our knowledge, we identified 1049 out of >150,000 compositions likely to form stable γ′ phases, predominantly distributed across 11 Al-containing systems and 25 Al-free systems. Combining the analysis of CALPHAD method, we experimentally synthesized two new Co-based alloys with γ/γ′ dual-phase microstructures, corroborating the reliability of our theoretical prediction model.

摘要高温合金是制造航空发动机中高温部件不可或缺的材料。近年来，一种新型γ/γ′ Co-Al-W合金的发现引发了对钴基高温合金的极大兴趣，期望能够突破传统镍基高温合金的熔点限制。然而，传统的γ/γ′Co基高温合金的设计和开发依赖试错方法，耗时、费力、周期长。在本研究中，我们采用了结合密度泛函理论(DFT)和机器学习(ML)的方法来预测和分析合金中γ′关键相的稳定性，从而加速γ/γ′钴基合金的发现。通过高通量DFT计算，我们获得了包含数千个可靠的生成能(H_f)和分解能(H_d)能量的数据集。通过回归模型选择和特征工程，我们的随机森林(RF)模型在H_f 预测中达到了98.07%的准确率，在H_d预测中达到了97.05%的准确率。利用训练良好的RF模型，我们预测了在Co-Ni-Fe-Cr-Al-W-Ti-Ta-V-Mo-Nb体系中超过150,000种三元和四元γ′相的能量。能量分析显示，Ni、Nb、Ta、Ti和V的存在显著降低了γ′相的H_f 和H_d，而Mo和W则通过增加能量值降低了其稳定性。值得注意的是，尽管Al降低了H_f ，却增加了H_d，因此对γ′的稳定性产生了不利影响。通过基于领域知识的筛选，我们从超过150,000种成分中确定了1049种可能形成稳定γ′相的成分，这些成分主要分布在11个含铝体系和25个不含铝体系中。结合相图优化与计算（CALPHAD）方法的分析，我们实验合成了两种具有γ/γ′双相显微组织的新钴基合金，验证了理论预测模型的可靠性。