Abstract

In this work, a new attempt has been made using machine learning algorithms for assessing failure mode of austempered ductile iron perforated plates. This aims at providing some insights into these problems by comparing the performance of machine learning models which are part of artificial intelligence. The ballistic performance could be assessed by k-nearest neighbors (KNN), support vector machine (SVM), logistic regression, and decision tree (DT) algorithms. Precision of KNN, SVM, logistic regression and DT models is found to be 0.75, 0.75, 0.8, and 1, respectively. F1 score of KNN, SVM, logistic regression and DT models is found to be 0.86, 0.86, 0.89, and 1, respectively for smooth bulge formation. Eventually, the DT model is established and the optimal prediction model is derived by fine-tuning the parameters.

摘要

在这项工作中，使用机器学习算法评估等温淬火球墨铸铁穿孔板的失效模式进行了新的尝试。这旨在通过比较作为人工智能一部分的机器学习模型的性能来提供对这些问题的一些见解。弹道性能可以通过 k 近邻 (KNN)、支持向量机 (SVM)、逻辑回归和决策树 (DT) 算法进行评估。KNN、SVM、逻辑回归和 DT 模型的精度分别为 0.75、0.75、0.8 和 1。发现 KNN、SVM、逻辑回归和 DT 模型的 F1 分数分别为 0.86、0.86、0.89 和 1，用于平滑凸起形成。最终建立DT模型，通过微调参数推导出最优预测模型。