自適應神經模糊推理系統（ANFIS）的車間作業調度表觀遲延成本修正（ATC）的先行參數預測方法

Look Ahead Parameter Prediction Method of Modified Apparent Tardiness Cost (ATC) for Job-Shop Scheduling by Using Adapted Neuro-Fuzzy Inference System (ANFIS)

艾廷佑(Adityo Murdiarto)

指導教授 ： 鍾文仁

中原大學/工學院/機械工程研究所/碩士(2020年)

https://doi.org/10.6840/cycu202000287

摘要

使拖延率最小化可以使製造公司受益於減少因遲到罰款而造成的損失利潤或增加客戶評估。表觀遲滯成本（ATC）是一種啟發式方法，已廣泛使用並且表現良好，可以使拖延時間減少。此方法具有一個稱為超前參數（k）的參數，它將改變ATC的特性。先前的研究已經研究了ATC的k預測，但是這些案例僅使用單機或併行機製造系統。通常使用固定的k值來簡化調度過程，但是很難確定它是否是最佳值。
這項研究提出使用自適應神經模糊推理系統（ANFIS）解決Job-shop調度情況下的k搜索。修改後的ATC用於處理每台計算機上發生的工作量，然後將其用作ANFIS的輸入。最後，ANFIS生成每台計算機的k值並定義每個作業的優先級索引。
計算序列程序表明，所提出的方法主要以其最佳性能工作。一些性能測試顯示出顯著的結果，可以將遲到率降低多達50％。

關鍵字

調度 ； 表觀拖延成本 ； ANFIS ； 超前參數

並列摘要

Minimizing tardiness can benefit a manufacturing company to decrease the loss profit due to lateness penalty or increase the customer assessment. Apparent tardiness cost (ATC) is one of heuristic method that widely used and performs well to make the tardiness time decrease. This method has a parameter called look-ahead parameter (k) that will change the characteristic of the ATC. The previous researches already studied the k predicting of the ATC, but only for cases with single or parallel machine manufacturing system. A fixed value of k is commonly used to simplify the scheduling process, but the optimum is still questionable.
This research proposes the use of Adapted Neuro-Fuzzy Inference System (ANFIS) to search the k value in Job-shop scheduling case. A modified ATC is used to analyze the workload of each machine or workstation, then uses it as the input of the ANFIS. Finally, the ANFIS generates the k value of each machine and define the priority index of each job.
The computational sequencing program shows the proposed method majorly works in its best performance. Some performance tests show significant results which can reduce the tardiness up to 50%.

並列關鍵字

Scheduling ； Apparent Tardiness Cost ； ANFIS ； look-ahead parameter

參考文獻

[1] V. Manthou and M. Vlachopoulou, ‘Agile Manufacturing Strategic Options’, in Agile Manufacturing: The 21st Century Competitive Strategy, A. Gunasekaran, Ed. Oxford: Elsevier Science Ltd, 2001, pp. 685–702.

[2] J. Du and J. Y.-T. Leung, ‘Minimizing Total Tardiness on One Machine Is NP-Hard’, Math. Oper. Res., vol. 15, no. 3, pp. 483–495, 1990.

[3] T. F. Morton and R. M. V. Rachamadugu, ‘Myopic Heuristics for the Single Machine Weighted Tardiness Problem’, p. 39, Nov. 1982.

[4] M. L. Pinedo, Planning and Scheduling in Manufacturing and Services. New York, NY: Springer New York, 2009.

[5] K. Li, W. Xiao, and S. Yang, ‘Minimizing total tardiness on two uniform parallel machines considering a cost constraint’, Expert Syst. Appl., vol. 123, pp. 143–153, Jun. 2019, doi: 10.1016/j.eswa.2019.01.002.