隨著全球製造業競爭日益激烈，許多企業愈加重視生產流程的內部化，特別是從原材料採購到產品製造的全過程，以降低成本並提升品質。半成品的保質期對最終產品品質至關重要，因此，客戶格外關注半成品是否能在時效性內完成加工過程。基於此，本研究聚焦於包含流水車間（Flow Shop）和作業車間（Job Shop）的環境，探索如何同時考量半成品的時效性與訂單截止日期，實現生產調度的優化。在多個流線型生產與工作坊生產系統中，各系統之間的互動使整體生產計劃變得更加複雜且難以預測。半成品在流線型生產與工作坊生產之間的交付時間對整體生產至關重要，缺乏有效協作可能導致訂單延遲或半成品過期。此外，考慮加工時間具有隨機性，我們需要一種能動態應對不確定性的方法，確保排程解的可行性與優性。為解決上述問題，本研究提出了元導向萊維優化演算法（MetaGuided Levy Optimization, MGLO）。該演算法在解空間中進行廣泛探索，以避免陷入局部最優解，並結合元模型（Metamodel）對解快速分類，聚焦於高質量解決方案，從而顯著減少模擬資源的消耗。實驗結果顯示，相較於其他演算法，MGLO 能有效找出更優質的解並顯著降低生產成本。研究成果為企業生產計劃的制定提供了重要的管理啟示，並為後續相關研究奠定了基礎。

As global manufacturing competition intensifies, companies are increasingly
emphasizing the internalization of production processes, focusing particularly on the entire workflow from raw material procurement to product manufacturing. This approach aims to reduce costs and improve quality. The shelf life of semi-finished products is crucial to the quality of final products. Consequently, customers are highly concerned about whether semi-finished products can be processed within their timeliness constraints. Based on this, this study focuses on the environments with flow shop and job shop, exploring how to simultaneously consider the timeliness of semifinished products and order deadlines to optimize production scheduling. In systems
involving multiple flow lines and job shop, interactions between systems significantly increase the complexity and unpredictability of overall production planning. The delivery time of semi-finished products between flow shops and job shop is critical to the overall production process. A lack of effective collaboration may result in order delays or the expiration of semi-finished products. Moreover, considering the stochastic nature of processing times, a method is required to dynamically address uncertainties and ensure the feasibility and optimization of scheduling solutions. To address these challenges, this study proposes the Meta-Guided Levy Optimization (MGLO)
algorithm. This algorithm conducts extensive exploration in the solution space to avoid local optima and incorporates a metamodel to quickly classify solutions, focusing on high-quality solutions and significantly reducing the consumption of simulation resources. Experimental results demonstrate that, compared to other algorithms, MGLO effectively identifies superior solutions and significantly reduces production costs. These findings provide valuable managerial insights for production planning and lay a solid foundation for future research.