當今製程技術不斷進步，個人電腦的各項需求近乎滿足，市場日趨飽和，產業逐漸走下坡。個人電腦已滿足了許多種可能的需求，例如各種功能的有無與其強度，反觀工業電腦，較強調各產品的穩健性，且高度客製化，因而維持了高毛利財務數字。工業電腦本質上一樣扮演電腦的角色，但必須在嚴苛的應用環境下正常運作，對不同的功能的穩定強度有不同的需求。為了達到穩定的保證，廠商經常花費大量時間研發測試，造成產品開發時間較常。同時，工業用電腦須長期運轉，產品生命週期較常。在高度客製化與長生命週期的情況下，產品架構、原始研發的考量等知識變得更不容易保存。如果能在相似產品或數代產品之間，找出合適的產品平台、模組，成為一項可能解決方案。同時我們期望在釐清可能的模組化時，在運轉環境限制下，電腦模組化的彈性如何保存，也一併呈現了出來。設計結構矩陣是一套能方便且清楚呈現系統的方法。在對子系統們做排序或群集後，以二維方陣呈現產品架構，子系統之間的相依性、模組關係一目了然，因此很適合發展模組與複雜產品的產品平台，與隔代產品知識的傳遞。本研究首先探討設計結構矩陣的使用時機與前提，並展示如何在適當的流程中使用設計結構矩陣，針對穩定性，求得適當地模組、發展平台。最後，我們以工業電腦為例，展示方法與分析，然後下結論。 When process technology continues to progress, demands of personal computer have been almost fully fulfilled, markets are becoming more saturated, and there is a trend for decline in this industry. Personal computers have fulfilled many possible demands no matter in all kinds of functions and the scales of its performance. On the other hands, industrial computers stress on reliability and are highly customized, as a result, industrial computer firms kept a relative high gross margin.Still as a role of computer, industrial computer further has to work in some much extremer environments and to fulfill functions of different performance and robustness. To ensure robustness, firms usually take much of time testing, which lead to more development time. At the same time, industrial computer has to be working within a long period and has a long product life cycle. Under highly customization and long life cycle, it is not easy to preserve the knowledge of product architecture and reasons for decision made. So, deriving modules and platforms for generations for product can be a possible solution.Design Structure Matrix (DSM) is a simple way to clearly representing a system. By sequencing or clustering subsystems it clearly represent the dependencies between subsystems in a two dimension matrix. And it is apt to developing modules or platforms for complex products. This thesis first discusses the context for using DSM, and demonstrates the proper steps to incorporate DSM to derive appropriate modules and plat forms regarding robustness. Lastly, we take industrial PC as a case for demonstration and analysis and then conclude.
