現今積層陶瓷電容(MLCC)已走向多堆疊數、薄層化的製程技術，意味著需要更細小陶瓷粉體，但是相對的粉體製作技術有相當的困難度，在開發更高電容的產品上需要投入更多的費用且更長的時間而結果不一定是理想的，故堆疊式陶瓷電容(MEGACAP)的開發因應而生，藉由一介質層將金屬片和單顆或多顆電容器進行並聯組裝，可達到電容量加倍的效果。 市面上有此技術且具備量產能力的公司主要以日系大廠村田、TDK和韓國三星及台系大廠信昌、禾伸堂等。而本研究是以P公司所獨家開發的組裝疊層電容器的技術在組裝過程遇到陶瓷體內部暗裂的問題，運用資料探勘中的羅吉斯回歸、決策樹與類神經網路三種演算法進行數據分析及模型比較，可有系統的協助工程師尋找異常的根本原因進行改善而非利用試誤法(Try & error)方式進行參數調整而增加錯誤機率導致多餘廢料和不良品產生。 ;Nowadays, multi-layer ceramic capacitors (MLCC) have moved towards a process technology with multiple stacks and thin layers, which means that finer ceramic powders are required, but the relative powder manufacturing technology has considerable difficulties in developing higher capacitance products. It requires more investment and longer time, and the result is less than ideal, so the development of Multiple stacked ceramic capacitors (MEGACAP) came into being, through a dielectric layer to connect the metal frames and single or multiple capacitors in parallel, which can achieve the effect of doubling the capacitance. Japan′s Murata, TDK, South Korea′s Samsung, and Taiwan′s Prosperity Dielectrics Co. (PDC) and Holystone, these companies have this technology and are capable of mass production. In this study, the new technology developed by P Company encountered the internal crack problem in the ceramic body during the assembly process. Logistic regression, decision tree and neural network in data mining were used. The three algorithms are used for data analysis and model comparison, which can systematically assist engineers to find the root cause of anomalies and improve them instead of using the trial-and-error method to adjust machine settings and parameters. Through this study, it is possible to reduce the probability of errors and lead to excess waste and defective products.
