Empirical Analysis of Deep Neural Networks for Classifying and Identifying 316L and AZ31B Mg Metal Surface Morphology

Deep neural networks (DNNs) have demonstrated remarkable capabilities in image classification and pattern recognition, making them well-suited for material surface morphology analysis. This study presents an empirical analysis of DNNs for classifying and identifying surface morphology of 316L stainless steel and AZ31B magnesium alloy. High-resolution microscopic images of both metals were obtained and preprocessed to enhance feature extraction. Multiple DNN architectures, including convolutional neural networks (CNNs), were trained and evaluated to determine their efficacy in distinguishing surface textures influenced by processing methods such as machining, etching, and corrosion. Performance metrics such as accuracy, precision, recall, and F1-score were analyzed to assess model effectiveness. The results indicate that DNN models can successfully differentiate between the two metal surfaces with high accuracy, providing a robust framework for automated material characterization. These findings contribute to the advancement of intelligent material inspection systems, reducing manual effort and improving the reliability of surface morphology assessments in industrial applications.

深度神经网络（Deep Neural Networks，DNNs）已在图像分类与模式识别领域展现出卓越性能，使其非常适用于材料表面形貌分析。本研究针对316L不锈钢与AZ31B镁合金的表面形貌分类与识别任务，开展了深度神经网络的实证分析。研究采集了两种金属的高分辨率显微图像并进行预处理，以强化特征提取效果。本研究训练并评估了多种深度神经网络架构，其中包括卷积神经网络（Convolutional Neural Networks，CNNs），以探究其在区分受机械加工、蚀刻及腐蚀等工艺影响的表面纹理时的性能表现。通过分析准确率、精确率、召回率与F1分数等性能指标，对模型的有效性进行评估。实验结果表明，深度神经网络模型能够以较高准确率有效区分两种金属的表面形貌，为自动化材料表征提供了稳健的框架。本研究成果推动了智能材料检测系统的发展，可减少人工工作量，提升工业应用中表面形貌评估的可靠性。