| 何洁1， 徐平华1， 袁子舜1*， 陆振乾2， 徐望3.基于机器学习的UD布弹道冲击有限元结果分析[J].轻工机械,2024,42(4):7-15 |
| 基于机器学习的UD布弹道冲击有限元结果分析 |
| Finite Element Results Analysis of UD Fabric Ballistic Impact Based on Machine Learning |
| |
| DOI：10.3969/j.issn.1005 2895.2024.04.002 |
| 中文关键词: 机器学习 单向织物 k Means聚类算法 有限元分析 应力分析 弹道冲击 |
| 英文关键词:machine learning UD fabric k Means clustering algorithm FEA（Finite Element Analysis） stress analysis ballistic impact |
| 基金项目:浙江理工大学科研基金（22072134 Y）；江苏省高等学校自然科学基金（23KJA430017）。 |
|
| 摘要点击次数: 120 |
| 全文下载次数: 92 |
| 中文摘要: |
| 由于目前主要采用的分析复合材料有限元模型图像结果的定性对比观察法分析不够深入和全面，可能会忽略某些重要信息，故而需要一种能够快速进行图像量化处理的客观分析方法。课题组结合机器学习，提出了基于k Means聚类算法的有限元图像结果分析方法；以分析Dyneema单向（unidirectional，UD）布以普通对齐堆叠和准各向同性堆叠的弹道冲击有限元模型的应力分布结果为例，利用k Means聚类算法对所截取的应力云图进行基于颜色特征的像素点聚类和区域分割，分割的区域可以实现快速统计和面积计算。结果表明：该方法能够快速准确地量化不同应力范围的面积差距，从而得出更为客观明了的结果，便于深入地分析。该方法还可以推广应用于其他需要分析云图结果的领域。 |
| 英文摘要: |
| Due to the insufficient depth and comprehensiveness of the qualitative comparative observation method currently used to analyze the image results of composite material finite element models, some important information may be overlooked. Therefore, an objective analysis method that can quickly quantify the image is needed. Combined with machine learning, the method of finite element image result analysis based on k Means clustering algorithm was proposed. Taking the stress distribution results of the ballistic impact finite element model of Dyneema UD laminate with ordinary aligned stacking and quasi isotropic stacking as an example, the k Means clustering algorithm was utilized to cluster and segment the captured stress cloud image by pixel points based on color features. The segmented area enabled fast statistics and area calculations. The results show that this method can efficiently quantify the area difference of different stress ranges, and can obtain more objective and clearer results which is convenient for in depth analysis. The method can applied to other fields in which cloud image results need to be analyzed. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |
