【摘要】 基于文献与试验获得的压铸铝合金“成分-铸态力学性能”数据，利用LASSO回归获得了10种不同元素对屈服强度、抗拉强度、伸长率的影响，并获得了小成分波动范围内的线性预测模型。利用BP神经网络、随机森林、Adaboost以及支持向量机4种非线性机器学习方法对全样本空间进行训练。结果表明，经过超参数优化后，BP神经网络具有最优的屈服强度与伸长率预测能力，随机森林模型具有最优的抗拉强度预测能力。更多还原

【Abstract】 Based on the data of “composition-as-cast mechanical properties” of die-cast aluminum alloy obtained in literature and experiments, influence rules and degrees of different elements on yield strength, tensile strength and elongation were obtained by LASSO regression. The linear prediction model within the small fluctuation range of components was obtained. Nonlinear machine learning methods, including BP neural network, random forest, Adaboost and support vector machine were used to fit the entire dataset. The results indicate that the BP neural network has the desirable prediction ability of yield strength and elongation, and the random forest model has the optimal prediction ability of tensile strength after hyperparameters optimization.更多还原