| 摘要: |
| 低熔点合金(Low-Melting-Point Alloys, LMAs)因其优异的导热性、低蒸气压、高相变潜热等特性,在焊料、低温润滑和冷却系统等领域具有广泛应用。然而,传统的合金设计方法效率低下,难以满足现代复杂应用需求。本研究提出了一种基于机器学习的逆向设计策略,通过目标熔点反推合金成分。以支持向量机模型为核心,结合逆向设计策略(Inverse Design Strategy),实现了对35℃和70℃目标熔点的低熔点合金设计。结果表明,所建模型在训练集和测试集上的R2值均高于0.95,预测精度优异。设计出的合金成分通过模式识别和SHAP分析验证了其合理性和可解释性。其中,30-40℃的合金填补了现有数据的空白,而70℃的不含重金属合金则展现出实际应用潜力。本研究为低熔点合金的高效设计提供了新的路径,并进一步推动了材料设计的智能化发展。 |
| 关键词: 低熔点合金 机器学习 逆向设计 |
| DOI： |
| 分类号: |
| 基金项目:国家自然科学基金项目（面上项目，重点项目，重大项目） |
|
| Machine Learning-Based Inverse Design Strategy for Low-Melting-Point Alloys |
|
LIU Taiang,WU Yanmiao,LU Tian,YUE Baohua,CAO Xioawei
|
|
Shanghai Weijie InformationTechnology Co,Ltd
|
| Abstract: |
| Low-melting-point alloys (LMAs) are widely used in applications such as soldering, low-temperature lubrication, and cooling systems due to their excellent thermal conductivity, low vapor pressure, and high latent heat of phase change. However, traditional alloy design methods are inefficient and cannot meet the demands of modern complex applications. This study proposes a machine learning-based inverse design strategy that infers alloy compositions from target melting points. Using a support vector machine (SVM) model as the core and combining it with Reverse Design Strategy, we have successfully designed LMAs with target melting points of 35°C and 70°C. The results show that the model has an R2 value higher than 0.95 on both the training and testing sets, indicating excellent prediction accuracy. The designed alloy compositions were validated for their rationality and interpretability through pattern recognition and SHAP analysis. Among them, the alloy with a melting point of 30–40°C fills the gap in existing data, while the 70°C alloy free of heavy metals shows potential for practical applications. This study provides a new pathway for the efficient design of LMAs and further promotes the intelligent development of material design. |
| Key words: LMAs Machine learning Inverse Design |
