A budget-aware simheuristic framework for stochastic single-machine scheduling
COMPUTERS & INDUSTRIAL ENGINEERING, cilt.221, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 221
- Basım Tarihi: 2026
- Doi Numarası: 10.1016/j.cie.2026.112288
- Dergi Adı: COMPUTERS & INDUSTRIAL ENGINEERING
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, ABI/INFORM, Aerospace Database, Applied Science & Technology Source, Compendex, INSPEC, DIALNET, Business Source Ultimate (EBSCO), Engineering Source (EBSCO), Technology Collection (ProQuest)
- Eskişehir Osmangazi Üniversitesi Adresli: Evet
Özet
This study addresses the stochastic single-machine scheduling problem with sequence-dependent setup times and uncertain processing times. An integrated simheuristic framework is proposed that combines a Genetic Algorithm (GA) with Monte Carlo simulation, OCBA-based replication allocation, and a cumulative simulation memory mechanism. The memory structure enables incremental refinement of solution estimates by accumulating simulation statistics across generations, thereby reducing redundant re-evaluation. A Random Forest (RF) screening layer is additionally incorporated to pre-filter the OCBA candidate pool. Experimental results show that the RF layer reduces simulation replication consumption while maintaining statistically equivalent solution quality compared to the memory-assisted variant without RF screening. Computational experiments compare four method variants across instances of varying sizes and variability levels. Results indicate that memoryassisted OCBA consistently improves solution quality and enhances the consistency of search-time performance estimates under fixed simulation budgets. The proposed variants are further benchmarked against an adapted Simulated Annealing based simheuristic (SimSA) baseline on large-scale, high-variance instances. Both GA + mOCBA and GA + RF + mOCBA achieve statistically significantly better expected tardiness values while consuming substantially fewer simulation replications. Nonparametric tests confirm the significance of observed differences. The framework is presented as a practical simulation-optimization architecture for uncertaintyaware scheduling, with potential future applicability in related stochastic scheduling and decision-support environments, although its applicability beyond the studied problem setting remains to be validated.