钛合金相变点预测模型的构建和评估

发布时间：2018-05-12 10:46

  本文选题：钛合金 + 相变点　； 参考：《钛工业进展》2016年06期

【摘要】：基于西北有色金属研究院实际生产中统计的321组钛合金铸锭化学成分与相变点数据,构建了预测钛合金(α+β)/β相变点的人工神经网络模型和多元线性回归模型,并对模型的准确性进行了评价分析。结果显示,多元线性回归模型的训练值及预测值与(α+β)/β相变点实际值的相关性系数分别为0.761 05和0.809 93,而人工神经网络模型的相关性系数分别为0.927 21和0.818 51,具有更好的相关性。人工神经网络模型的平均绝对误差为4.02℃,相比多元线性回归模型(平均绝对误差为5.11℃)具有更高的精度,可以更好地描述合金元素与钛合金(α+β)/β相变点之间的非线性关系。
[Abstract]:An artificial neural network model and a multivariate linear regression model for predicting the phase transition point of titanium alloy were established based on 321 sets of data of chemical composition and phase transition point of titanium alloy ingot from the actual production of Northwest Nonferrous Metals Research Institute. The accuracy of the model is evaluated and analyzed. The results show that the correlation coefficients between the training value and the predicted value of the multivariate linear regression model and the actual value of the phase transition point (伪 尾) are 0.761 05 and 0.809 93, respectively, while the correlation coefficients of the artificial neural network model are 0.927 21 and 0.818 51, respectively. The average absolute error of the artificial neural network model is 4.02 鈩，

本文编号：1878326