基於可靠度演算法分析系統晶片設計之專案時程可行性

積體電路產業為目前臺灣市場主流，牽動臺灣的經濟命脈，其中因應5G、AI、智能家電的產品需求，客製晶片的研發設計日趨重要，如何讓系統單晶片前端設計於期限及成本內順利完成，是很重要的課題。目前廣泛使用在實務上的專案排程模型，如甘特圖、要徑法，當面對大規模且複雜的工程專案時，其無法事先評估專案可靠度以供參考。且專案人員習慣以經驗法則來作人、機、料的調配，因此常在專案收尾之際，出現超時、超支的損失。Lin等人(2017)提出可反映實務的專案可靠度演算法，藉著成本與工期的限制衍生滿足需求的上下界向量，並根據遞迴不交合法計算，可推估出專案可靠度之上下界限區間數值。另外，此研究法沿用了AOA網絡圖，在箭號上呈現專案內的每一個工作任務，比在節點上標示更具效益，並採用任意隨機機率分配，以取代計劃評核術中的ß機率分配。此對專案管理者更具有參考價值。本研究採用文獻之專案可靠度演算法，應用於個案公司之新產品內部研發專案，針對數據結果進行探討，藉著專案總工期與成本的限制計算可執行機率，則可推估將來專案完工的可靠度，若專案規格難度高於原先定義之模型，即可藉此建議專案管理者調整成本與工期，以提升未來專案完工之有效性。

關鍵字

要徑法；計劃評核術；專案可靠度演算法；晶片設計；時程可行性

並列摘要

The integrated circuit industry is currently the mainstream of the Taiwan. It affects Taiwan's economy a lot. Due to the demand for 5G, AI, and smart home appliances increase, the projects for custom chip design are becoming important than before. The paper is to evaluate the probability of the system chip front-end design completion under the project completion time and cost. At present, the project scheduling models usually used these research, such as Gantt charts, CPM, PERT, they can’t accurately evaluate the exact values for project reliability, when the project is too large and complicate. In the company, the project manager always arranges the resource (include man, machine, and material) by their experience, that’s not reasonable. Therefore, over-time and over-cost often occur by every case. Lin et al.(2017) proposed project reliability algorithm. It can through the upper and lower boundary vectors of cost and duration, then based on the RSDP algorithm, the probability of the upper and lower bounds of the project reliability can be estimated. In addition, this research follows the AOA network diagram, the activities show on the arrow, which is more profitable than on the node, and proposes an arbitrary random probability instead of the Beta distribution. It’s suitable in current situation. This is more helpful for project managers. This paper uses the project reliability algorithm of the research in the schedule of new product projects of the company, and discusses the data results. Through the feasible probability of the project's total duration and cost, it can be estimated as the actual completion probability of the future project. And it can be indicated if the new project spec is higher than this model, the project manager can adjust the appropriate cost and duration in advance, so that the project will enhance project completion more efficiency.