高强、高导铝合金研发的机器学习策略

Development of high-strength,high-conductivity aluminum alloys:A machine learning strategy

利用机器学习框架搭建材料研究设计平台对材料性能进行分析与预测,成为开发新型材料的重要手段。铝合金的导电率和强度往往是互斥的,导电率的提高,伴随着强度的降低。使用SVM、RF、ELM、BP和DNN五种机器学习方法建立6000系铝合金的导电率和强度的机器学习预测模型。发现以热力学数据和加工工艺为特征输入,在合金性能预测模型的构建方面表现出巨大潜力。并最终筛选出精确度高,泛化能力好的深度神经网络预测模型。经过与实验数据验证,证明了所提模型对于铝合金导电率、强度预报的可靠性。

Constructing a material research and design platform using machine learning frameworks to analyze and predict material properties has become an important tool for developing new materials.The electrical conductivity and ultimate tensile strength of aluminum alloys are often mutually exclusive,with an increase in electrical conductivity accompanied by a decrease in strength.Five machine learning methods,SVM,RF,ELM,BP and DNN,were used to develop machine learning prediction models for the electrical conductivity and tensile strength of 6000 series aluminum alloys.It is found that using thermodynamic data and processing processes as feature inputs shows great potential in the construction of alloy performance prediction models.And finally,a deep neural network prediction model with high accuracy and good generalization ability is proposed.After validation with experimental data,the reliability of this model for predicting the electrical conductivity and strength of aluminum alloys is demonstrated.