Scholarlink

<p>Post-mortem fractographic analysis plays an important role in analyzing the fracture of brittle specimens. The formation of surface features results from the dynamic crack propagation process and could be quantitatively correlated to the fracture strength. However, the accuracy in conventional fractographic approaches relies on the subjective interpretation of the surface features. In this study, the fracture surface of single-crystal silicon on the (110) cleavage plane was studied. Various imaging methods were used to collect the fracture surface information, and later, a convolutional neural network model was developed to correlate the fracture strength and fracture surface features. The proposed approach was capable of analyzing optical images under various illumination conditions and scanning electron microscope images, while minimizing the effect of subjectivity in the fractographic approach. The strength estimates by the proposed model on single-crystal silicon with two crystalline systems were obtained within 10% error.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Quantitative fractographic analysis on (110) fracture surface of single-crystal silicon via convolutional neural network

  • 期刊论文

  • 发表日期：

    2026年7月14日

查看全文
  • Tao Wang,
  • Haoran Song,
  • Jianzhong Zhao,
  • Huamiao Wang,
  • Lingyue Ma

摘要

Post-mortem fractographic analysis plays an important role in analyzing the fracture of brittle specimens. The formation of surface features results from the dynamic crack propagation process and could be quantitatively correlated to the fracture strength. However, the accuracy in conventional fractographic approaches relies on the subjective interpretation of the surface features. In this study, the fracture surface of single-crystal silicon on the (110) cleavage plane was studied. Various imaging methods were used to collect the fracture surface information, and later, a convolutional neural network model was developed to correlate the fracture strength and fracture surface features. The proposed approach was capable of analyzing optical images under various illumination conditions and scanning electron microscope images, while minimizing the effect of subjectivity in the fractographic approach. The strength estimates by the proposed model on single-crystal silicon with two crystalline systems were obtained within 10% error.