AI技術快速演進帶動AI伺服器產業蓬勃發展，其製造過程涉及高客製化、多工序與技術密集等特性，對排程與派工決策形成極大挑戰。傳統的排程方法難以應對高變異訂單與有限資源的現實情境，因此本研究聚焦於派工法則於AI伺服器生產系統中的應用，並運用Pro-Model模擬軟體進行建模與實證分析，以探討不同派工策略對生產績效之影響。 本研究首先針對AI伺服器製程特性進行問題分析，包括訂單多樣化所引發的排程複雜性、技術密集工序的流程瓶頸，以及人力與設備資源限制等核心議題。隨後，以Pro-Model模擬軟體建立實際產線之模擬模型，納入各製程站點的加工時間、產能限制、訂單到達間隔與模擬時間範圍等參數設定。 本研究比較四種常見派工法則：先進先出（FIFO）、最短處理時間（SPT）、最早交期（EDD）與隨機選取（RS）。透過模擬分析，評估各法則在平均流程時間、平均延遲時間、站點利用率、產出數量與工單完成率等指標上的表現差異。模擬結果顯示，SPT法則在降低流程時間上效果顯著，而EDD法則可有效減少延遲並提升達交率；FIFO執行簡易但整體表現中等，RS雖公平分配但穩定性較差。 綜合模擬結果，本研究驗證不同派工策略對AI伺服器生產效能具有顯著影響，亦突顯Pro-Model模擬在多工序製造系統模擬與排程決策支援上的實用性。研究成果可為製造業者提供具體派工策略依據，未來亦可擴展至即時資料導向的智慧派工系統，以因應日益變動的高階製造需求。 ;The rapid growth of artificial intelligence (AI) technology has fueled demand in the AI server industry, where manufacturing is characterized by high customization, complex multi-stage processes, and limited resources. These conditions make scheduling and dispatching particularly challenging. This study investigates the application of dispatching rules in AI server production and utilizes Pro Model simulation to assess their impact on manufacturing performance. Key production challenges—such as order variability, process bottlenecks, and constrained resources—are first analyzed. A simulation model replicating real production scenarios is developed using Pro Model, incorporating parameters such as processing times, station capacities, and order arrivals. Four dispatching rules—First-In-First-Out (FIFO), Shortest Processing Time (SPT), Earliest Due Date (EDD), and Random Selection (RS)—are compared. Results show SPT minimizes flow time, EDD enhances due-date performance, FIFO performs moderately, and RS is less stable. This study demonstrates that dispatching strategies significantly affect production efficiency and highlights the value of simulation as a decision-support tool. The use of Pro Model not only enables performance comparison under controlled conditions but also provides visual insights for identifying bottlenecks and optimizing scheduling logic in complex manufacturing systems.
