運用多種啟發式演算法求解多樓層不規則連續式設施佈置問題

設施佈置問題（Facility Layout Problem, FLP）基本上是一種排列組合的問題，如何將各種部門(人員)、機具設備或原物料安置於正確的位置，讓組織的相關活動流程以更低成本、更高效率完成。而多樓層設施佈置問題是單一樓層的延伸，要考慮的因素也比單一樓層複雜，包括樓地板面積、電梯座標、樓層限定、相鄰問題等，不過當所有影響的因素都列入考慮後，也就一併解決單一樓層設施佈置問題。
過往有不少文獻均有提出運用單一或兩三種以內之演算法進行求解，而再針對該文獻研究問題的類型調整參數或組合演算法以尋求較佳且較有效率的解法，鮮少有研究以齊一的方式測試各種演算法在處理多樓層設施佈置的議題上表現的績效進行評比，本研究希望透過一致的做法，在相同的水準之下，以各啟發式演算原始運算邏輯的基礎下進行比較，以期找出最適合用以求解多樓層設施佈置的演算法。
本研究共使用基因演算法、蟻群演算法、禁忌搜尋法、模擬退火法、人工蜂群演算法、粒子群演算法、複製選擇演算法及差分演算法等8種啟發式演算法進行求解，研究結果顯示，模擬退火法、複製選擇演算法及基因演算法在求解多樓層不規則連續式設施佈置問題上有較佳且較穩定的表現，而以座標搜尋為基礎之粒子群演算法則較不適用於此類問題的求解。

A Facility Layout Problem (FLP) is a combinatorial problem that involves arranging different departments (personnel), machines and equipment, or raw materials in optimal locations to enhance cost-effectiveness and operational efficiency. The multi-floor facility layout problem extends beyond single-floor scenarios, introducing additional complexity factors such as floor area, elevator coordinates, floor restrictions, and adjacent issues. However, even with all these influencing factors considered, the single-floor FLP can still be addressed concurrently. Prior studies have focused on employing a single or a few algorithms to address the FLP and subsequently adjusting parameters or combining algorithms to achieve improved and more efficient solutions for specific problem types. Few studies have comprehensively compared various algorithms for handling multi-floor facility layout problems using consistent methodologies and comparable evaluation criteria. This study aims to compare multiple heuristic algorithms on an equal footing, utilizing their original computational logic to identify the most suitable algorithm for solving multi-floor facility layout problems. In this research, eight heuristic algorithms, including Genetic Algorithm, Ant Colony Optimization, Tabu Search, Simulated Annealing, Artificial Bee Colony, Particle Swarm Optimization, Clonal Selection Algorithm, and Differential Evolution, were employed to address the problem. The results indicate that Simulated Annealing, Clonal Selection Algorithm, and Genetic Algorithm demonstrate superior and more stable performance in solving multi-floor irregular continuous facility layout problems. However, Particle Swarm Optimization based on the coordinate search is found to be unsuitable for addressing such issues.