基于机器学习的挤压铸造铝合金力学性能预测

Prediction of Mechanical Properties of Squeeze-cast Aluminum Alloys Based on Machine Learning

基于现有挤压铸造研究数据,以不同合金元素及其含量下铝合金的力学性能作为训练数据,结合带有因子分解机(Factorization Machine,FM)的多项式回归模型,通过机器学习算法,以梯度下降策略对模型进行训练学习。然后,以合金的元素含量作为输入条件,预测该成分下合金的力学性能,并与试验力学性能作对比验证。结果表明,该模型能较好地预测不同元素含量铝合金的抗拉强度、屈服强度、硬度和伸长率等力学性能指标。

A polynomial regression model with factorization machine(FM)was applied for predicting the mechanical properties of squeeze-cast aluminum alloys.The data of mechanical properties with different element contents from reported researches was used as training samples based on a gradient descent algorithm in the machine learning model.Then the element contents were input to predict the mechanical properties of aluminum alloys,which was also validated with the experiment data.The validation reveals that the model resulted from machine learning can well predict tensile strength,yield strength,hardness and elongation of alloy with different Al content.