Eşzamanlı parti büyüklüğü belirleme ve sıralama problemi için sezgisel yöntemler

This study tackles the Capacitated lot-sizing problem (CLSP) enriched with industrial complexities such as sequence-dependent setup times, setup carryovers, secondary resource constraints, and parallel non-identical machines. These characteristics pose significant challenges for traditional optimisation techniques, especially when scaling to large, real-world problem instances. To address these challenges, we develop a novel column generation heuristic framework. The method decomposes the problem into a Restricted Master Problem (RMP) and pricing subproblems, which dynamically generate production patterns based on dual information. Unlike classical exact methods or generic heuristics, our framework incorporates a problem-specific neighbourhood search that guides pattern generation heuristically. This results in high-quality solutions within computational time frames that are acceptable for industrial applications. The proposed Column generation neighbourhood search (CGNS) algorithm is validated on synthetic benchmarks and real-world data from a plastic injection facility. Results show that our method outperforms conventional fix-and-relax heuristics and achieves performance comparable to commercial solvers, while maintaining superior scalability. The modular structure of the algorithm also enables its integration with machine learning-based decision support systems, paving the way for hybrid optimisation approaches implemented as an improvement to CGNS algorithm as Column generation neighbourhoodsearch with neural networks (CGNSNN)).