軍事用途產品種類繁多，其用途橫跨陸、海、空及通訊電子產品，生產時屬於「少量多樣化」的生產，當軍用品碰上了大量製造的印刷電路板時，批次生產站常因量能未達經濟生產量生產，造成產能的浪費及生產成本增加；因此本研究提出利用合併訂單生產的方式，在不影響產品品質的前提下，提升站點的產能利用率。 群集分析的目的是透過觀察樣本間的特性，利用資料間的共同性高低進行分類，形成不同組別的方法；本研究利用群集分析設立併單公式，設立公式時需同時由製程面、產品特性面及品質面進行多方考量，將生產時的製作條件以及品質結果列為關鍵因子，可以確保合併後的產品品質無虞，並利用相似度給分高低的方式表現出產品間適合合併生產與否，為了考量產品群中的極端值情況，不讓產品因合併生產而品質下降，因此只有相似度高的產品才可分為同一群集中並合併生產，故本研究選用「完全連結法」(Complete linkage method) 連結各產品。另外在考量產品品質面及製程面時，部分關鍵因子非單純的定量型或定性型資料，而是需要透過實際生產時的狀況並輔以電路板產業的背景知識，給予不同產品不同的相似度給分；因此如何制定併單公式及產品間的相似度，端看各使用者所面臨的生產狀況進行調整。 本研究針對三個批次生產站進行研究，利用群集分析所設立併單公式，讓投料人員下料時能抓出相似性高的料號進行併單生產，結果發現併單後的站點產能利用率可提升約兩倍，隨著訂單量愈多合併的效果愈好。;Military electronic products have a wide range of applications in different fields. They are small-volume production of a wide range of different items. However, when it comes to printed circuit board manufacturing which are famous for mass production, the problem of low volume at batch production stations would cause a waste of capacity and increase costs. Therefore, this study proposes a method of combining orders for production to improve the utilization rate of the stations without affecting the product quality. The purpose of cluster analysis is to observe the characteristics of the samples and use the commonality between the data to classify them and form different groups. In this study, cluster analysis is used to set up the single formulas. The formulas must be composed of the manufacturing process, product quality and product characteristics at the same time. The production conditions and quality results are listed as key factors to ensure the quality of the merged product. The degree of similarity can indicate whether the products are suitable for combined production or not. In order to consider the extreme values in the product group, complete linkage method is used to link products. Under this method, only products with a high degree of similarity can be categorized into the same cluster and combined production. In addition, when considering product quality and processes, some of the key factors are not simply qualitative or quantitative data. Instead, they need to provide different similarities to different products through the actual production conditions and background knowledge of the circuit board industry. Thus, how to formulate the similarity between single formulas and products depends on the production situation that each user faces. This study was conducted on three batches of production stations. By using cluster analysis to establish a single formula, people can decide whether to combine or not. The station productivity with combined production was positive. The utilization rate increased approximately two times. Moreover, the larger the order volume was. the greater the effect of order combination would be.
