本篇論文,以序的最佳化為基礎的方法被提出用來在有限的時間內,針對將延遲成本和保存成本降到最低來求出隨機標準工作生產排程問題一個足夠好的排程解,本篇論文所提議的方法由探索層次和開發層次所構成,在探索層次部分,本篇論文提出使用基因演算法來挑選出一個好的候選解集合,搭配一個已事先訓練好的類神經網路作為未加工模型來評估目標函數值的適應力,在開發層次部份,是以多個子層次所組成,安排分配運算資源進行反覆的運算來挑選出適合的候選解數量,對每一個子層次所剩下的候選解數量執行模擬運算,在對每一個子層次執行模擬運算時,會減少候選解的數量並逐漸增加執行模擬運算的次數,在最後一個子層次所得到的候選解即為本篇論文所要尋找的一個足夠好的排程解,本篇所提出的方法應用在隨機標準工作生產排程問題(SCJSSP),分別針對三種不同的加工時間機率分布函數Truncated normal、Uniform以及Exponential,在測試結果部份證明本篇論文所提出的方法,成功的得到一個足夠好的排程解以及運算效能。
In this paper, an ordinal optimization based approach is proposed to solve for a good enough schedule that minimizes expected sum of storage expenses and tardiness penalties of stochastic classical job shop scheduling problem using limited computation time. The proposed approach consists of exploration and exploitation stage. The exploration stage uses a genetic algorithm to select a good candidate solution set, where the objective function is evaluated with an artificial neural network that is trained beforehand. The exploitation stage composes of multiple substages, which allocate the computing resource and budget by iteratively and adaptively selecting the candidate solutions. At each substage, remaining solutions are simulated and some of them are eliminated, and the solution obtained in the last substage is the good enough schedule that we seek. The proposed approach is applied to a SCJSSP with random processing time in truncated normal, uniform, and exponential distributions. The test results demonstrated that the obtaining good enough schedule is successful in the aspects of solution quality and computational efficiency.