昝工具：人工智慧用於錫球枕頭效應瑕疵檢測

TsanKit: Artificial Intelligence for Solder Ball Head-In-Pillow Defect Inspection

昝亭甄(Ting-Chen Tsan)

指導教授 ： 傅楸善

國立臺灣大學/電機資訊學院/資訊工程學研究所/碩士(2019年)

https://doi.org/10.6342/NTU201901318

摘要

本論文提出一個人工智慧(Artificial Intelligence)解決方案，用於錫球枕頭效應 (Head-In-Pillow, HIP)瑕疵檢測。HIP 瑕疵會影響錫球的導電性，進而導致間歇性故 障。由於 HIP 產生的位置與形狀千變萬化，傳統電腦視覺演算法無法完全檢測到 瑕疵。近年來，卷積類神經網路(Convolutional Neural Network)在影像識別與分類 上具有出色的性能，但是當資料量不足時，容易產生過度擬合的問題。因此我們 結合 CNN 和機器學習演算法——支持向量機(Support Vector Machine)，設計一套 檢測流程。參考幾種最先進模型的優點後，我們提出一個三維 CNN 模型，並採用 Focal Loss和Triplet Loss來解決因為缺少瑕疵資料造成的資料不平衡問題。與數 種經典的 CNN 模型和深度學習檢測軟體 SuaKIT 相比，我們的檢測方法具有最佳 的性能和快速的測試速度。

關鍵字

錫球 ； 枕頭效應 ； 瑕疵檢測 ； 人工智慧 ； 卷積類神經網路

並列摘要

In this thesis, we propose an AI (Artificial Intelligence) solution for solder ball HIP (Head-In-Pillow) defect inspection. HIP defect will affect the conductivity of the solder balls leading to intermittent failures. Due to the variable location and shape of HIP defect, traditional machine vision algorithms cannot solve the problem completely. In recent years, Convolutional Neural Network (CNN) has an outstanding performance in image recognition and classification, but it is easy to cause overfitting problem due to insufficient data. Therefore, we combine CNN and the machine learning algorithm Support Vector Machine (SVM) to design our inspection process. Referring to the advantages of several state-of-the-art models, we propose our 3D CNN model and adopt focal loss as well as triplet loss to solve the data imbalance problem caused by rare defective data. Our inspection method has best performance and fast testing speed compared with several classic CNN models and the deep learning inspection software SuaKIT.

並列關鍵字

Solder Ball ； Head-In-Pillow ； Defect Inspection ； Artificial Intelligence ； Convolutional Neural Networks

參考文獻

[1] Agilent Technologies, “Chemical Analysis, Life Sciences, and Diagnostics - Agilent,” https://www.agilent.com, 2019.

[2] A. Castellanos, et al. “Head in Pillow X-ray Inspection at Flextronics,” https://smtnet.com/library/files/upload/head-in-pillow-inspection-flextronics.pdf, 2019.

[3] C. Cortes, and V. Vapnik, “Support-Vector Networks,” Machine Learning, vol. 20, no. 3, pp. 273-297, 1995.

[4] L. Deng and D. Yu, “Deep Learning: Methods and Applications,” Foundations and Trends® in Signal Processing, vol. 7, no. 3–4, pp. 197-387, 2014.

[5] ESCATEC, “ESCATEC Invests €150,000 in 2.5D X-Ray Inspection System to Further Improve Quality Control,” https://myemail.constantcontact.com/ESCATEC-PR--150-000-Euro-investment-in- X-ray-inspection-system-.html?soid=1108958702593&aid=Bx02AvE5mV8, 2013.