詳細書目資料

7
0
0
0

基於YOLO的紅墨水測試錫球斷面影像辨識技術及其效率提升應用 = YOLO-based Image Recognition of Solder Ball Cross-Section for Red Dye Penetration Testing and Its Application for Efficiency Improvement / 鄭博仁.

館藏資訊

摘要註

在科技快速發展的背景下，電子元件的內部設計變得日益複雜，同時更容易受到環境因素（如溫度、濕度）和機械衝擊的影響，這些因素可能導致焊點劣化，進而影響球柵陣列封裝（BGA）的可靠性。近十年來，BGA結構在多個產業中得到廣泛應用，因此對其可靠性的檢測具有重要的學術和實用價值。本研究採用影像處理和YOLO技術，對使用紅墨水試驗後的BGA斷面進行自動辨識和分析。相比於傳統方法，這種自動化流程具有多項顯著優勢。首先，它大幅減少人工判定所需的時間和人力成本：在涉及大量錫球的情況下，人工判定通常需要超過一小時，而本研究採用的YOLO模型僅需少量時間即可完成，顯著降低人力投入時間。其次，自動生成的錫球腳位圖解決拔起晶片後缺乏對應腳位圖的問題，從而提高整個檢測流程的效率和專業性。這一進展也消除從客戶處獲取相應腳位圖的需要，減輕客戶的負擔。最後，深度學習技術不僅加速檢測流程，還提高後續人工判定的準確性。本研究希望能利用YOLO提供一個具有學術和實用價值的解決方案，能夠大幅提高BGA檢測的效率和準確性，並且減少人力訓練成本，使專業人員能更有效地從事其他有產值的工作，並且提升在可靠度實驗可融入AI的整體意識使產業進步。. Against the backdrop of rapid technological advancement, the internal designs of electronic components are becoming increasingly complex and susceptible to environmental factors such as temperature, humidity, and mechanical impacts. These influences can degrade solder joints, thereby affecting the reliability of Ball Grid Array (BGA) packaging. Over the past decade, BGA structures have found widespread applications across multiple industries, underscoring the significant academic and practical value in their reliability testing,this study employs image processing and YOLO (You Only Look Once) technology to automatically identify and analyze cross-sections of BGAs treated with red ink testing, compared to traditional methods, this automated process offers several notable advantages. Firstly, it drastically reduces the time and labor costs associated with manual inspection: in scenarios involving a large number of solder balls, manual inspection typically exceeds an hour, whereas the YOLO model used in this research completes the task in significantly less time, thereby reducing manpower requirements. Secondly, the automated generation of the solder ball footprint diagram addresses the challenge of lacking corresponding diagrams after chip removal, thereby enhancing the efficiency and professionalism of the entire inspection process. This advancement also eliminates the need to obtain such diagrams from clients, alleviating their burden. Lastly, deep learning technology not only accelerates the inspection process but also enhances the accuracy of subsequent manual evaluations, this research aims to provide a solution of both academic and practical value using YOLO, significantly improving the efficiency and accuracy of BGA inspection while reducing training costs for personnel. This allows professionals to engage more effectively in other value-added tasks, contributing to an enhanced awareness of integrating AI into reliability experiments and fostering industrial

相關作品

回到最上