臺灣博碩士論文加值系統

中文摘要
MLCC（Multilayer Ceramic Capacitor，多層陶瓷電容）對電子產品主要作用是進行儲電、旁路、耦合、去耦合、濾雜訊等功能，對電子產品的穩定性和可靠性至關重要。隨著3C產品、AI、車用、物聯網、行動通訊技術蓬勃發展，MLCC小型化的優勢，使得MLCC的運用與需求與日俱增，以車用為例，電動車市佔率增加與輔助駕駛規格的提升，在MLCC的運用上電動智能車比傳統燃油車運用約多3倍的使用量，隨單板運用數量的倍增，MLCC被偵測到不良的機率也隨之增加。MLCC看似是簡單不起眼的電子元件，卻須經複雜的製造工序與冗長的生產流程，生產過程中的製程變異可能造成MLCC可靠度問題，導致運用端客訴的反饋，讓公司付出高昂的成本。
運用測試機將電性不良篩除與經由測試結果進行風險管控，為預防運用端客訴重要防線。本研究以W公司為例，針對小尺寸高容產品0201尺寸1μnF電容值，探討測試製程經由IT與MES系統擷取測試機電測數據，將其資料使用散佈圖將其視覺化分類並使用統計分析t檢定確認各項測試數據因子與Halt結果具有關聯性，再將測試數據以MLP預測Halt檢驗結果，搭配MES系統對測試生產批進行品質分級管控，降低特殊CODE降級比例，降低Halt 批退率與降低客訴風險。

關鍵詞：MLCC、MES、Halt、MLP

Abstract
MLCC (Multilayer Ceramic Capacitor) plays a crucial role in electronic products, including energy storage, bypassing, coupling, decoupling, and filtering noise, which are essential for the stability and reliability of electronic products. With the booming development of 3C products, AI, automotive, IoT, and mobile communication technologies, the advantages of miniaturization of MLCC have led to an increasing demand for its application. For example, in the automotive industry, the use of MLCC in electric and intelligent vehicles is about three times that of traditional fuel vehicles due to the increase in market share of electric vehicles and the improvement of assisted driving specifications. As the number of single-board applications increases, the probability of detecting defective MLCC also increases. Although MLCC seems to be a simple and inconspicuous electronic component, it requires complex manufacturing processes and redundant production processes. Process variations during production may cause reliability issues of MLCC, resulting in feedback from customer complaints and high costs for the company.
Using a testing machine to screen out defective electrical properties and conducting risk management based on the test results is an important line of defense to prevent customer complaints. This study takes Company W as an example and focuses on the 0201 size 1μnF capacitor value of small-sized high-capacity products. The testing process is explored by extracting the test machine's electrical test data through IT and MES systems. The data is visualized and classified using a scatter plot, and statistical analysis t-tests are used to confirm the correlation between various test data factors and Halt results. The test data is then used to predict the Halt inspection results using MLP, and the MES system is used to control the quality of the test production batch by grading and controlling the quality, reducing the proportion of special CODE downgrades, reducing the Halt batch rejection rate, and reducing customer complaint risks.
Keywords：MLCC, MES, Halt, MLP.

目錄
中文摘要................................................................................................................................ i Abstract................................................................................................................................. ii
誌謝...................................................................................................................................... iii
目錄...................................................................................................................................... iv
圖目錄.................................................................................................................................. vi
表目錄................................................................................................................................ viii
第一章 緒論........................................................................................................................ 1
1.1背景與動機...................................................................................................... 1
1.2研究目的.......................................................................................................... 2
1.3研究流程.......................................................................................................... 3
第二章 文獻探討................................................................................................................ 4
2.1 IR特性介紹..................................................................................................... 4
2.2 MLCC可靠度.................................................................................................. 6
2.3 MLP.................................................................................................................. 8
2.4 MES.................................................................................................................. 9
2.5相關文獻探討彙整....................................................................................... 11
第三章 研究方法.............................................................................................................. 17
3.1研究架構........................................................................................................ 17
3.2個案公司介紹................................................................................................ 18
3.3 MLCC測試流程介紹................................................................................... 20
第四章 個案分析.............................................................................................................. 24
4.1資料收集........................................................................................................ 24
4.2資料處理........................................................................................................ 26
4.3探索性分析.................................................................................................... 30
4.4 MLP-深度學習.............................................................................................. 39
4.5 MLP模型優化............................................................................................... 43
4.6 MLCC測試製程結果驗證.......................................................................... 44
第五章 結論與未來研究方向......................................................................................... 45
5.1結論................................................................................................................. 45
5.2未來研究方向................................................................................................ 46
參考文獻............................................................................................................................. 47


















圖目錄
圖1-1 研究流程.............................................................................................................. 3
圖2-1 IR 漏電流曲線................................................................................................... 4
圖2-2 Halt 曲線圖......................................................................................................... 7
圖3-1 研究架構............................................................................................................ 17
圖3-2 MLCC產品生產流程圖.................................................................................. 18
圖3-3 測試機外觀及各區說明.................................................................................. 20
圖3-4 上探針及下探針............................................................................................... 20
圖3-5 以MLCC漏電流分佈篩除漏電較大產品.................................................... 22
圖3-6 漏電流曲線圖................................................................................................... 22
圖3-7 測試機PC站別量測過程................................................................................ 22
圖3-8 測試機IR站別量測過程................................................................................. 23
圖4-1 測試機測試完成拋出紀錄至MES伺服器................................................... 24
圖4-2 每Lot各測試因子的數據資料....................................................................... 25
圖4-3 測試機每顆MLCC的量測值......................................................................... 25
圖4-4 C-out去離群前後比較圖................................................................................. 26
圖4-5 DF去離群前後比較圖..................................................................................... 26
圖4-6 IR去離群前後比較圖...................................................................................... 27
圖4-7 NC去離群前後比較圖.................................................................................... 27
圖4-8 PC_Fail charge去離群前後比較圖................................................................ 27
圖4-9 IR1_Fail去離群前後比較圖........................................................................... 27
圖4-10 IR2_Fail去離群前後比較圖........................................................................... 28
圖4-11 IR3_Fail去離群前後比較圖........................................................................... 28
圖4-12 Yield去離群前後比較圖................................................................................. 28
圖4-13 Halt OK與NG部分資料................................................................................. 29
圖4-14 Cout散佈圖................................................................................................. 33
圖4-15 DF散佈圖.................................................................................................... 33
圖4-16 IR散佈圖..................................................................................................... 33
圖4-17 NC散佈圖................................................................................................... 33
圖4-18 PC_Fail Charge散佈圖............................................................................... 34
圖4-19 IR1_Fail Charge散佈圖.............................................................................. 34
圖4-20 IR2_Fail Charge散佈圖.............................................................................. 34
圖4-21 IR3_Fail Charge散佈圖.............................................................................. 35
圖4-22 Yield散佈圖................................................................................................ 35
圖4-23 PC_Fail Charge vs Cout散佈圖.................................................................. 36
圖4-24 PC_Fail Charge vs DF散佈圖..................................................................... 36
圖4-25 PC_Fail Charge vs IR散佈圖...................................................................... 36
圖4-26 PC_Fail Charge vs NC散佈圖..................................................................... 36
圖4-27 PC_Fail Charge vs IR1_ Fail Charge散佈圖............................................... 37
圖4-28 PC_Fail Charge vs IR2_ Fail Charge散佈圖............................................... 37
圖4-29 PC_Fail Charge vs IR3_ Fail Charge散佈圖............................................... 37
圖4-30 PC_Fail Charge vs Yield散佈圖................................................................. 37
圖4-31 0201尺寸1μnF電容值產品Halt LRR趨勢圖......................................... 44







表目錄
表2-1 MIL-PRF-123、MIL-PRF-55681與AEC-Q200的Life testing條件......... 6
表2-2 相關文獻探討彙整一覽表......................................................................... 11
表3-1 測試機量測站別說明................................................................................. 21
表4-1 Cout t檢定表............................................................................................... 30
表4-2 DF t檢定表.................................................................................................. 30
表4-3 IR t檢定表....................................................................................................31
表4-4 NC t檢表......................................................................................................31
表4-5 PC_Fail Charge t檢定表..............................................................................31
表4-6 IR1_Fail Charge t檢定表.............................................................................31
表4-7 IR2_Fail Charge t檢定表.............................................................................32
表4-8 IR3_Fail Charge t檢定表.............................................................................32
表4-9 Yield t檢定表.............................................................................................. 32
表4-10 MLP_PC_Fail charge vs yield模型表......................................................... 39
表4-11 MLP_PC_Fail charge vs Cout模型表......................................................... 39
表4-12 MLP_PC_Fail charge vs DF模型表........................................................... 40
表4-13 MLP_PC_Fail charge vs IR模型表............................................................. 40
表4-14 MLP_PC_Fail charge vs NC模型表........................................................... 40
表4-15 MLP_PC_Fail charge vs IR1_Fail charge模型表....................................... 41
表4-16 MLP_PC_Fail charge vs IR2_Fail charge模型表....................................... 41
表4-17 MLP_PC_Fail charge vs IR3_Fail charge模型表....................................... 41
表4-18 MLP預測模型表(PC_ Fail & IR2_Fail).................................................... 42
表4-19 MLP優化預測模型表(PC_ Fail & IR2_Fail)............................................ 43

參考文獻
中文文獻：
[1]鄭文杰 編著，「應用設計思維到記憶體封裝成品測試解決方案之探討」，國立清華大學高階經營管理碩士在職專班碩士論文，2016。
[2]陳育正 編著，「應用六標準差方法改善積體電路成品測試良率」，國立高雄大學電機工程學系碩博士班碩士論文，2018。
[3]陳延昌 編著，「應用類神經網路決定重要IC測試程式組合之決定」，國立清華大學工業工程與工程管理學系碩士在職專班碩士論文，2014。
[4]孔祥竹 編著，「應用類神經網路減少TFT-LCD產品測試項目之研究」，國立交通大學管理學院碩士在職專班工業工程與管理組碩士論文，2007。
[5]陳君豪 編著，「以類神經網路探討晶圓測試良率預測與重測指標值之建立」，國立中央大學工業管理研究所碩士在職專班碩士論文，2007。

英文文獻：
[6]Breeta SenGupta, Dimitar Nikolov, Urban Ingelsson and Erik Larsson, "Test Planning for Core-based Integrated Circuits under Power Constraints," Journal of Electronic Testing volume 33, pages7–23, 2017.
[7]Breeta SenGupta, Dimitar Nikolov, Assmitra Dash and Erik Larsson, "Test Flow Selection for Stacked Integrated Circuits," Journal of Electronic Testing volume 35, pages425–440, 2019.
[8]David Enke and Suraphan Thawornwonog, "The use of data mining and neural networks for forecasting stock market returns," Expert Systems with Applications volume 29, pages927–940, 2005.
[9]Yung Ting, Chih-Ho Chen, Hui-Yi Feng and Shin-Liang Chen, "Apply Computer Vision and Neural Network to Glue Dispenser Route Inspection," 2007 IEEE International Conference on Mechatronics and Automation, 2007.
[10]Anshul Shrivastava, Bhanu Sood, Michael Azarian, Michael Osterman and Michael Pecht, "An investigation into a low insulation resistance failure of multilayer ceramic capacitors," 2010 IEEE 60th Electronic Components and Technology Conference, 2010.
[11]Donhang David Liu, "Insulation Resistance Degradation in Ni–BaTiO3 Multilayer Ceramic Capacitors," 2015 IEEE Transactions on Components Packaging and Manufacturing Technology, 2015.
[12]Alexander Teverovsky, "Cracking Problems in Low-Voltage Chip Ceramic Capacitors," NASA Electronic Parts and Packaging (NEPP) Program, 2018.
[13]Almada-Lobo Francisco, "The Industry 4.0 revolution and the future of Manufacturing Execution Systems (MES)." Journal of innovation management, pages16-21, 2016.
[14]Suzuki Keigo, Okamoto Takafumi, Kondo Hiroyuki and Tanaka Nobuhiko, "Insulation degradation behavior of multilayer ceramic capacitors clarified by Kelvin probe force microscopy under ultra-high vacuum," Journal of Applied Physics 113.6, 2013.
[15]Okamoto Takafumi, Kitagawa Shuji, Inoue Noriyuki and Ando Akira, "Electric field concentration in the vicinity of the interface between anode and degraded BaTiO3-based ceramics in multilayer ceramic capacitor," Applied Physics Letters 98.7, 2011.
[16]David H. Collins, Jason K. Freels, Aparna V. Huzurbazar, Richard L. Warr and Brian P. Weaver, "Accelerated test methods for reliability prediction," Journal of Quality Technology 45.3, pages244-259, 2017.
[17]Pier Francesco Orru, Andrea Zoccheddu, Lorenzo Sassu, Carmine Mattia, Riccardo Cozza and Simone Arena "Machine learning approach using MLP and SVM algorithms for the fault prediction of a centrifugal pump in the oil and gas industry," Sustainability 12.11, 2020.
[18]Taud, Hind and J. F. Mas, "Multilayer perceptron (MLP)," Geomatic approaches for modeling land change scenarios, pages451-455, 2018.
[19]Soujanya Mantravadi and Charles Møller, "An overview of next-generation manufacturing execution systems: How important is MES for industry 4.0?," Procedia manufacturing 30, pages588-595, 2019.
[20]Yuanyuan Wei, Julian Jang-Jaccard, Fariza Sabrina, Amardeep Singh, Wen Xu and Seyit Camtepe, "AE-MLP: A hybrid deep learning approach for DDoS detection and classification," IEEE Access 9, pages146810-146821, 2021.

參考網頁：
[21]AEC Documents, "AEC-Q200: Stress Test Qualification For Passive Components," http://www.aecouncil.com/AECDocuments.html