«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《电镀与精饰》[ISSN:1001-3849/CN:12-1096/TG]

卷:

期数:

2025年09

页码:

13-21

栏目:

出版日期:

2025-09-30

文章信息/Info

Title:

Research on accurate detection of bulking fault in gold plating of aluminum alloy parts under intergranular corrosion stress

作者:

田雪英¹; 陈佳鹏²; 齐艳珂³; 李艳丽^4*

故障精准检测研究田雪英1,陈佳鹏2,齐艳珂3,李艳丽4*(1. 郑州电子信息职业技术学院 信息工程学院,河南 郑州 451450 ;2. 上海工程技术大学 微纳制造先进材料研究中心,上海 201620 ;3. 郑州航空工业管理学院 计算机学院,河南 郑州 450046 ;4. 信阳师范大学 计算机与信息技术学院,河南 信阳 464000)

Author(s):

Tian Xueying1;  Chen Jiapeng2;  Qi Yanke3;  Li Yanli4*

(1. School of Information Engineering, Zhengzhou Electronic Information Vocational and Technical College, Zhengzhou 451450, China; 2. Research Center for Advanced Micro-/Nano- Fabrication Materials, Shanghai University of Engineering Science, Shanghai 201620, China; 3. School of Computer Science, Zhengzhou University of Aeronautics, Zhengzhou 450046, China; 4. School of Computer Science and Technology, Xinyang Normal University, Xinyang 464000, China)

关键词:

晶间腐蚀; 铝合金零件; 镀金层; 鼓包故障; 可分离卷积

Keywords:

intergranular corrosion;  aluminum alloy parts;  gold plating;  bulge failure;  separable convolution

分类号:

TP391

文献标志码:

A

摘要:

铝合金零件表面的镀金层遭受晶间腐蚀时，金属晶粒间的连接强度会被削弱，腐蚀区域会产生应力集中，促使镀层更快分离并形成鼓包。在形状简单的零件上鼓包较易发现，但零件结构复杂或鼓包位于结合面附近时，传统卷积操作提取的镀金层特征图像素信息多，只能粗略判断鼓包所在层，无法精确检测其位置和形态。为此，提出一种精准检测铝合金零件镀金层鼓包故障的新方法。将铝合金零件镀金层的图像数据转化为灰度数据。优化亚历克斯网络架构(AlexNet Architecture)，以深度可分离卷积替代传统卷积操作来提取灰度图像数据特征，引入全局平均池化技术来取代全连接层，将卷积层输出的特征图直接映射为关键特征点，大幅减少数据量。将关键特征向量输入分类层，判断是否存在鼓包故障。利用交叉熵损失函数评估模型检测结果与真实结果的差异，并据此训练和优化模型。选取一家铝合金零件制造商的电镀零件作为样本进行鼓包故障检测实验。实验结果显示：在训练次数高达700次时，该检测方法的交并比高达0.90，每秒检测帧数仅降低了5.5%；检测准确率一直处于98.28%以上，且能够精准检测出镀金层鼓包的位置和形态。

Abstract:

When the gold plating layer on the surface of aluminum alloy parts suffered from intergranular corrosion, the connection strength between metal grains was weakened. Stress concentration occurred in the corroded area, which promoted the faster separation of the plating layer and the formation of bulges. Bulges were relatively easy to detect on parts with simple shapes. However, when the part structure was complex or the bulges were located near the bonding surface, the feature images of the gold plating layer extracted by traditional convolution operations contained a large amount of pixel information. Only the layer where the bulge was located could be roughly determined, and its position and shape could not be accurately detected. Therefore, a new method for accurately detecting bulge faults in the gold plating layer of aluminum alloy parts was proposed. The image data of the gold plating layer of aluminum alloy parts were converted into grayscale data. The AlexNet architecture was optimized. Depthwise separable convolution was used to replace traditional convolution operations for extracting features of grayscale image data. Global average pooling technology was introduced to replace the fully connected layer. The feature maps output by the convolution layer were directly mapped into key feature points, greatly reducing the amount of data. The key feature vectors were input into the classification layer to determine whether there was a bulge fault. The cross-entropy loss function was used to evaluate the difference between the model detection results and the real results, and the model was trained and optimized accordingly. Electroplated parts from an aluminum alloy parts manufacturer were selected as samples for bulge fault detection experiments. The experimental results showed that when the training times are as high as 700 times, the intersection ratio of the detection method is as high as 0.90, the detection frame number per second is only reduced by 5.5%, the detection accuracy rate is always above 98.28%, and the location and shape of the gold-coated bulges can be accurately detected

常用功能

更新日期/Last Update: 2025-09-11