背鑽品質影響因子之資料分析與研究-以G公司為例;Data Analysis and Investigation of Quality Impact Factors in Back-Drill: A Case Study of Company G

NCUIR > School of Management at National Central University > Executive Master of Industrial Management > Electronic Thesis & Dissertation > Item 987654321/97310

Please use this identifier to cite or link to this item: https://ir.lib.ncu.edu.tw/handle/987654321/97310

Title:	背鑽品質影響因子之資料分析與研究-以G公司為例;Data Analysis and Investigation of Quality Impact Factors in Back-Drill: A Case Study of Company G
Authors:	黃郁姍;Huang, Yu-Shan
Contributors:	工業管理研究所在職專班
Keywords:	背鑽製程;統計資料分析;決策樹;隨機森林;擬合最小化平方法;關鍵製程變因篩選;Back-drill Process;Statistical Data Analysis;Decision Tree;Random Forest;Fit Least Squares Method;Key Process Variable Selection
Date:	2025-07-15
Issue Date:	2025-10-17 11:07:11 (UTC+8)
Publisher:	國立中央大學
Abstract:	隨著電路板設計逐漸朝向高密度、多層堆疊與高速傳輸發展，背鑽技術已成為確保訊號完整性的重要製程環節，其品質穩定性對最終產品良率與系統效能具有關鍵影響。在伺服器、5G通訊設備及高速運算等應用場域中，背鑽品質的細微差異即可能對系統運作造成訊號干擾或失真。然而，背鑽製程涉及多項具交互關係的參數變化，包括板厚、層壓結構、鑽孔深度控制等，導致傳統單點監控或單因分析方法難以有效辨識潛在問題來源。本研究以 G公司 2024 年實際生產數據為基礎，針對背鑽加工樣本進行資料結構化與統計分析，透過多層次的變異比較與品質指標檢視，探討影響良率變異的關鍵製程因素。研究結果顯示，板厚穩定度與 K 值的統計離散性為與品質表現高度相關的核心因子，並在不同條件與抽樣架構下均展現出一致性的趨勢。進一步的分析亦協助識別異常樣本的特徵分布，提供品質異常預警的參考依據。綜合研究成果，本文提出具實務應用價值的變因排序結果與可運用於製程監控的分群條件，期望協助企業建構更具前瞻性的數據驅動品質改善機制，提升生產品質穩定性與風險控管能力。 ;As printed circuit board designs continue evolving toward high density, multilayer stacking, and high-speed signal transmission, back drilling has become a critical process for maintaining signal integrity. Its quality stability significantly influences product yield and system performance in applications such as servers and 5Gcommunication equipment. However, the process involves multiple interacting parameters, making it difficult to identify root causes using conventional monitoring approaches. This study analyzes back-drill process data from G Company in 2024 to investigate the key factors affecting yield variation. Through structured data preprocessing and statistical analysis, the research identifies board thickness stability and the statistical dispersion of K-values as consistently influential variables. These factors show stable trends across various sampling conditions and analysis perspectives. Based on the results, the study further proposes a prioritized list of contributing variables and classification conditions that can support early warning systems and quality control strategies. The findings are expected to assist process engineers in identifying hidden risks and improving process robustness through data-driven decision making.
Appears in Collections:	[Executive Master of Industrial Management] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	50	View/Open

社群 sharing

Loading...