本文以精實六個標準差(Lean Six Sigma)手法為主軸，利用科學方法的應用、資訊的取得與統計數據的呈現進行印刷電路板(Printed circuit board)製程良率改善的研究。 本研究的目的以印刷電路板焊錫製程為背景，在各製程中找出影響良率最大之製程別為改善對象，在降低成本的主要考量下，其中進行產品良率的提升，可直接地降低生產成本與後續衍生出的退貨、客訴問題。 定義階段，針對印刷電路板焊錫製程defect ratio進行柏拉圖分析，並選定功能性測試工作站點做為關鍵品質要素(Critical To Quality，CTQ)，並藉由SIPOC Chart 找出主要製成站點及專案成立之關單位。 量測階段，評估量測系統利用量測系統分析手法(Measurement Systems Analysis)得的結果指出量測系統是良好的。並利用製成能力分析(Analysis of Process Capability)及製程品質不良比對找出欲改善品質異常項目為主要改善方向。 分析階段，利用實驗設計法(Design of Experiment)，分析出主要影響製程良率因子及區分影響因子是否顯著，並加以佐證影響因子是否呈現顯著效果，列出待改善因子進行改善。 改善階段，將顯著因子透過響應分析手法(Response Curve)，利用各因數的機台極限對應產品特性限制，設計實驗測試條件，經多項實驗驗證，影響產品良率的因數皆被找出，亦經過修正後驗證品質為良好的，以及進行批量的驗證，以確保對策是否有效。 控制階段，針對製程條件主要影響因數，找出最佳製程參數，最終經由批量的驗證，並再次利用製成能力分析進行驗證。 本研究結果證實藉由精實六個標準差的五個階段(DMAIC)的實施能使製程平均DPMO由1,792降為至780改善成本效益一年約美金3萬元，證明良率有明顯提升改善，其優點在於有明確的方法與架構，使得改善方案容易實踐。並結合統計工具，快速分析，得以在最短時間內確認針對改善方案之對策是否有成效。;The Research uses Lean Six Sigma as principle to apply science method, information acquisition and presentation of Statistical data for PCB process yield improvement. Purpose- Find out the biggest impact for each process stage of PCB Welding, lower production cost by improving the product’s yield at each stage and reducing return of goods as well as customer complaints. Definition- Use PCB process’s defect ratio to perform the Pareto analysis and select functional test station as Critical to Quality (CTQ) to find out the critical stage and project established unit by SIPOC Chart. Measurement- Use Measurement Systems Analysis (MSA), Analysis of Process Capability and comparison of defective product to find out the root cause of quality issues, then to improve the quality. Analysis- Determine the significant and improvement factors by applying Design of Experiment (DOE). Improvement- Determine the key impact factors and characteristic limitation of the products by applying Response Curve of DOE. Control-Determine the key process parameters/factors for batch verification & validation. This research verifies the DAMIC method of Lean Six Sigma could save 30,000 US dollars a year by reducing average process DPMO from 1792 to 711. The effectiveness of yield improvement analysis can be verified in a short period-of-time by using statistic tools with well-developed workflow.
