ABSTRACT
Optimization of manufacturing schedules is of great theoretical and practical significance, especially when high-tech products are produced in large manufacturing networks with global supply chains. In this paper, a novel approach is described how discrete-event simulations shall be enhanced by Artificial Intelligence (AI) to optimize manufacturing schedules for complex job shop manufacturing networks in the high-tech energy industry.
This work attempts to close the gap between the growing expectations on AI-based discrete-event simulations for manufacturing schedule optimization on the one hand and the current limitations of data quality and Information Technology (IT) capabilities in existing Enterprise Resource Planning (ERP) systems on the other hand. In order to deliver the expected benefits, business targets of schedule optimization need to be understood, the solution approach has to be defined and verified in a complex manufacturing environment. Performance and reliability of the implemented solution are validated with two real data sets derived from ongoing business planning activities.
An actor-critic architecture is proposed which uses a multi-stage schedule compression algorithm for modifying the start dates of production orders according to defined business targets and a genetic algorithm which selects production orders for automated make or buy decisions. The first research result is the provision of a manufacturing schedule for 375 product types, more than 180 manufacturing resources, 367 process variants (routings), and a total of 1.293 production orders at about 80% average capacity utilization of bottleneck machinery in a 12 months planning horizon, which is optimized according to lead times of customer products. The second research result is the generation of a manufacturing schedule for 3.657 production orders in a 24 months planning horizon, where the optimal outsourcing portion is calculated according to defined extensions of standard lead times within the same supply and manufacturing network.
- Stricker, N./ Kuhnle, A./ Sturm, R./ Friess, S., Reinforcement learning for adaptive order dispatching in the semiconductor industry, CIRP Annals -- Manufacturing Technology 67, 2018, 511--514.Google ScholarCross Ref
- Sotskov, Y./ Egorova, Single Machine Scheduling Problem with Interval Processing Times and Total Completion Time Objective, in: Werner, F./ Burtseva, L./ Sotskov, Y., Algorithms for Scheduling Problems, MDPI, 2018, 21--40.Google Scholar
- Vakhania, N., Scheduling a Single Machine with Primary and Secondary Objectives, in: Werner, F./ Burtseva, L./Sotskov, Y., Algorithms for Scheduling Problems, MDPI, 2018, 41--56.Google Scholar
- Fuchigami, H. Y./ Sarker, R./ Rangel, S., Near-Optimal Heuristics for Just-In-Time Jobs Maximization in Flow Shop Scheduling, in: Werner, F./ Burtseva, L./ Sotskov, Y., Algorithms for Scheduling Problems, MDPI, 2018, 57--73.Google Scholar
- Drugan, M. M., Reinforcement learning versus evolutionary computation: A survey on hybrid algorithms, Swarm and Evolutionary Computation 44, 2019, 228--246.Google ScholarCross Ref
- Rooyani, D./ Defersha, F. M., An Efficient Two-Stage Genetic Algorithm for Flexible Job-Shop Scheduling, Science Direct, IFAC PapersOnLine 52-13, 2019, 25192524.Google Scholar
- Waschnek, B./ Reichstaller, A./ Belzner, L./ Altenmüller, T./ Bauernhansl, T./ Knapp, A./ Kyek, A., Optimization of global production scheduling with deep reinforcement learning. In 51st CIRP Conference on Manufacturing Systems, Science Direct, Procedia CIRP 72 (2018), 1264--1269.Google ScholarCross Ref
- Ghavamzadeh, M./ Mannor, S./ Pineau, J./ Tamar, A., Bayesian Reinforcement Learning: A Survey, now Publishers, 2015, 359--483. http://dx.doi.org/10.1561/2200000049Google ScholarDigital Library
- Ghavamzadeh, M./ Engel, Y./ Valko, M., Bayesian policy gradient and actor-critic algorithms, Journal of Machine Learning Research 17, 2016, ResearchGate, 1--53.Google Scholar
- Kasie, F. M./ Bright, G./ Walker, A., Decision support systems in manufacturing: a survey and future trends. Journal of Modelling in Management, Vol. 12 No. 3, 2017, 432--454. https://doi.org/10.1108/JM2-02-2016-0015.Google ScholarCross Ref
- Lassig, L. Mazzer, F. Nicolich. Poloni, C., Hybrid flow shop management: multi objective optimization, ScienceDirect, Procedia CIRP 62, 2017, 147--152.Google ScholarCross Ref
- Such, F. P./ Madhavan, V./ Conti, E./ Lehman, J./ Stanley, K. O./ Clune, J, Deep Neurorevolution: Genetic Algorithms are a Competitive Alternative for Training Deep Neural Networks for Reinforcement Learning, Uber AI Labs, 2018, arXiv:1712.06567v3.Google Scholar
- Palombarini, J. A./ Martinez, E. C., Closed-Loop Rescheduling using Deep Reinforcement Learning, ScienceDirect, IFAC PapersOnline 52-1, 2019, 231--236.Google ScholarCross Ref
- Sheikh, H. U./ Bölöni, L., Multi-Agent Reinforcement Learning for Problems with Combined Individual and Team Reward, Cornell University, 2020, arXiv:2003.10598v1.Google ScholarCross Ref
Index Terms
- AI-Based Discrete-Event Simulations for Manufacturing Schedule Optimization
Recommendations
Open-source discrete-event simulation software for applications in production and logistics: An alternative to commercial tools?
AbstractDiscrete-event simulation is an established method to support decision making for planning tasks in production and logistics. However, there are still many enterprises, especially smaller companies that do not use discrete-event simulation because ...
Manufacturing planning and control in batch manufacturing industries: a multidimensional construct
Manufacturing activity plays a major role in the industrial and economic development of a country. Effective planning and control of manufacturing is the key to performance of the manufacturing firms. The research question addressed in this paper is ...
KI-Net: AI-Based Optimization in Industrial Manufacturing—A Project Overview
Computer Aided Systems Theory – EUROCAST 2022AbstractArtificial intelligence (AI) is a crucial technology of industrial digitalization. Especially in the production industry, a great potential is present in optimizing existing processes, e.g., concerning resource consumption, emission reduction, ...
Comments