簡易檢索 / 詳目顯示

回結果列表
研究生： 趙璟琛
Chao, Ching-Chen
論文名稱： 利用強化學習提升化工產業生產排程之實證研究
An Empirical Study on Enhancing Production Scheduling in the Chemical Industry through Reinforcement Learning
指導教授： 王宏鍇
Wang, Hung-Kai
學位類別： 碩士
Master
系所名稱： 電機資訊學院 - 製造資訊與系統研究所
Institute of Manufacturing Information and Systems
論文出版年： 2024
畢業學年度： 112
語文別： 英文
論文頁數： 74
中文關鍵詞： 強化學習 、基因演算法 、排程最佳化 、PSO 演算法 、永續生產
外文關鍵詞： Reinforcement learning, GA, PSO, Scheduling, Sustainable production
相關次數： 點閱：103  下載：0 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 本研究旨在解決化學工廠生產排程問題，並以強化學習為方法，旨在最小化浪費，提升生產效益。我們提出了一個基於強化學習的模型，以學習最佳的生產排程策略。同時，我們將該方法與基因演算法（GA）和粒子群優化演算法（PSO）進行比較，以深入瞭解各種最佳化方法在化學工廠排程問題上的表現。
    首先從相關文獻中獲取化學工廠生產排程相關知識，並明確定義了問題，包括考慮的變數、目標函數（最小化浪費）以及約束條件。然後，通過數據收集，我們獲得了化學工廠的相關數據，包括生產設備運轉情況、產品生產需求和原料供應情況。並且我們建立了基於強化學習的模型，使用強化學習方法進行訓練，以學習最佳的生產排程策略。同時，我們實施了基因演算法和 PSO 演算法以及傳統派工法則作為對比方法，並在實驗中比較了四種方法的性能。
    實驗結果顯示，基於強化學習的模型在最小化浪費方面取得了優異的效果，優於基因演算法和 PSO 演算法，能夠排除最多浪費量以達到永續生產的需求。這表明強化學習在化學工廠生產排程問題上的優越性，有望成為一種有效的優化方法。最後我們進行了對比方法之間的深入分析，討論了各種方法的優勢和劣勢，並提出了未來研究的建議，包括模型擴展、更多算法的比較以及實際工廠的應用等。本研究為化學工廠生產排程最佳化提供了有價值的參考，同時豐富了強化學習在工業應用中的應用範疇。

    This study uses reinforcement learning to optimize production scheduling in chemical factories, aiming to minimize waste and enhance efficiency. We propose a reinforcement learning-based model and compare it with genetic algorithm (GA) and particle swarm optimization (PSO). Initially, we gather relevant knowledge on chemical factory production scheduling and define the problem, including variables, the objective function (minimizing waste), and constraints.
    Through data collection, we obtain relevant data, including production equipment status, product demands, and raw material supply. We build the reinforcement learningbased model and train it to learn the optimal scheduling strategy, fine-tuning it using interpolation. We also implement GA and PSO as comparative methods. Experimental results show that the reinforcement learning model significantly reduces waste, outperforming GA and PSO. It demonstrates the potential of reinforcement learning in chemical factory scheduling for sustainable production.
    Additionally, we conduct a thorough analysis of the comparative methods, discussing their strengths and weaknesses and proposing suggestions for future work, such as model expansion and more algorithm comparisons. This study provides valuable insights for optimizing chemical factory production scheduling and extends the application of reinforcement learning in industrial settings.

    中文摘要 I Abstract II Acknowledgements III Table of Contents IV List of Tables VI List of Figures VII Chapter 1 Introduction 1 1.1 Background, Motivation and Importance 1 1.2 Research Objectives 4 1.3 Structure of the Thesis 6 Chapter 2 Literature Review 7 2.1 The Introduction of Polypropylene 7 2.1.1 Evolution of Polypropylene 8 2.1.2 Manufacturing Processes of PP 9 2.2 Category of Job Types in Scheduling 13 2.3 Reinforcement Learning for Scheduling Problem 17 Chapter 3 Research Methodology 23 3.1 Research Framework 24 3.2 Q-Learning 31 3.2.1 State Representation 34 3.2.2 Action Space 36 3.2.3 Reward Design 37 3.2.4 Q-table 40 3.2.5 Termination Condition 43 Chapter 4 Empirical Study and Results Analysis 44 4.1 Data Analysis and Parameter Settings 45 4.1.1 Data Analysis 46 4.1.2 Experimental Parameter Settings 49 4.1.3 Experimental Environment 50 4.2 Result Comparison 51 4.2.1 Short-term scale 52 4.2.2 Long-term scale 57 Chapter 5 Conclusion and Future Research 61 5.1 Conclusion 61 5.2 Future Research 62 References 64

    Andrew, A. M., 1999 REINFORCEMENT LEARNING: AN INTRODUCTION by Richard S. Sutton and Andrew G. Barto, Adaptive Computation and Machine Learning series, MIT Press (Bradford Book), Cambridge, Mass., 1998, xviii+ 322 pp, ISBN 0-262-19398-1,(hardback,£ 31.95), Robotica vol. 17, (2), 229-235.
    Chen, Y.-F., 2023 Using a Hybrid Genetic Algorithm to Solve Production Scheduling Problems - A Case Study of a Chemical Raw Material Processing Plant, Master's Thesis vol. National Cheng Kung University, Taiwan.
    Chraibi, A., Ben Alla, S., & Ezzati, A., 2021 Makespan Optimisation in Cloudlet Scheduling with Improved DQN Algorithm in Cloud Computing, Sci. Program. vol. 2021, 11.
    Ding, D., Fan, X. C., Zhao, Y. H., Kang, K. X., Yin, Q., & Zeng, J., 2020 Q-learning based dynamic task scheduling for energy-efficient cloud computing, Futur. Gener. Comp. Syst. vol. 108, 361-371.
    Fachantidis, A., Taylor, M., & Vlahavas, I., 2019 Learning to Teach Reinforcement Learning Agents, Mach. Learn. Knowl. Extr. vol. 1, (1), 21-42.
    Ferreira, D., Clark, A. R., Almada-Lobo, B., & Morabito, R., 2012 Single-stage formulations for synchronised two-stage lot sizing and scheduling in soft drink production, Int. J. Prod. Econ. vol. 136, (2), 255-265.
    Hubbs, C. D., Li, C., Sahinidis, N. V., Grossmann, I. E., & Wassick, J. M., 2020 A deep reinforcement learning approach for chemical production scheduling, Comput. Chem. Eng. vol. 141, 22.
    Jeoung, S. K., Choi, J. G., Ko, Y. K., Kim, J. H., & Lee, J. H., 2018 Process Optimization for Manufacturing High Strength Polypropylene Foam Using Homo-polypropylene, Polym.-Korea vol. 42, (3), 364-370.
    Jiang, S. L., Zheng, Z., & Liu, M., 2017 A multi-stage dynamic soft scheduling algorithm for the uncertain steelmaking-continuous casting scheduling problem, Appl. Soft. Comput. vol. 60, 722-736.
    Kim, D., Lee, T., Kim, S., Lee, B., & Youn, H. Y., 2020 Adaptive packet scheduling in IoT environment based on Q-learning, J. Ambient Intell. Humaniz. Comput. vol. 11, (6), 2225-2235.
    Koslovski, G. P., Pereira, K., & Albuquerque, P. R., 2024 DAG-based workflows scheduling using Actor-Critic Deep Reinforcement Learning, Futur. Gener. Comp. Syst. vol. 150, 354-363.
    Lee, C.-Y., Ho, C.-Y., Hung, Y.-H., & Deng, Y.-W., 2024 Multi-Objective Genetic Algorithm Embedded with Reinforcement Learning for Petrochemical Melt-Flow-Index Production Scheduling, Appl. Soft. Comput. 111630.
    Martín-Guerrero, J. D., & Lamata, L., 2021 Reinforcement Learning and Physics, Appl. Sci.-Basel vol. 11, (18), 6.
    Mete, E., & Girici, T., 2020 Q-Learning Based Scheduling With Successive Interference Cancellation, IEEE Access vol. 8, 172034-172042.
    Natta, G., 1955 UNE NOUVELLE CLASSE DE POLYMERES D ALPHA-OLEFINES AYANT UNE REGULARITE DE STRUCTURE EXCEPTIONNELLE, Journal of Polymer Science vol. 16, (82), 143-154.
    Perez-Gonzalez, P., & Framinan, J. M., 2018 Single machine scheduling with periodic machine availability, Comput. Ind. Eng. vol. 123, 180-188.
    Ren, C. X., Xu, H., Yin, C. C., Zhang, L. Y., Chai, C. X., Meng, Q., & Fu, F. F., 2023 Research on Hybrid Scheduling of Shared Bikes Based on MLP-GA Method, Sustainability vol. 15, (24), 23.
    Sajjad, M. H., Naeem, K., Zubair, M., Jan, Q. M. U., Khattak, S. B., Omair, M., & Nawaz, R., 2021 Waste reduction of polypropylene bag manufacturing process using Six Sigma DMAIC approach: A case study, Cogent Eng. vol. 8, (1), 21.
    Su, Y. Y., Han, L. J., Wang, H. M., & Wang, J. N., 2022 The workshop scheduling problems based on data mining and particle swarm optimisation algorithm in machine learning areas, Enterp. Inf. Syst. vol. 16, (2), 363-378.
    Sun, X. Y., & Geng, X. N., 2019 Single-machine scheduling with deteriorating effects and machine maintenance, Int. J. Prod. Res. vol. 57, (10), 3186-3199.
    Wang, H. K., Lin, Y. C., Liang, C. J., & Wang, Y. H., 2023 Multi-subpopulation parallel computing genetic algorithm for the semiconductor packaging scheduling problem with auxiliary resource constraints, Appl. Soft. Comput. vol. 142, 16.
    Wiers, V. C. S., & VanderSchaaf, T. W., 1997 A framework for decision support in production scheduling tasks, Prod. Plan. Control vol. 8, (6), 533-544.
    Xiong, H. J., Chen, J. J., Rong, S., & Zhang, A. W., 2023 Power Battery Scheduling Optimization Based on Double DQN Algorithm with Constraints, Appl. Sci.-Basel vol. 13, (13), 21.
    Zhang, Z. C., Zheng, L., & Weng, M. X., 2007 Dynamic parallel machine scheduling with mean weighted tardiness objective by Q-Learning, Int. J. Adv. Manuf. Technol. vol. 34, (9-10), 968-980.
    Zhao, C., & Deng, N., 2024 An actor-critic framework based on deep reinforcement learning for addressing flexible job shop scheduling problems, Math. Biosci. Eng. vol. 21, (1), 1445-1471.

    無法下載圖示 校內：2029-07-30公開
    校外：2029-07-30公開
    電子論文尚未授權公開，紙本請查館藏目錄
    QR CODE