An analytics-based framework for optimizing resource allocation and preemptive scheduling in manufacturing

Abstract

Production scheduling is critical in manufacturing operations, requiring the optimal assignment of limited resources. This paper introduces a novel generalization of the Unrelated Parallel Machine (UPM) problem, addressing key real-world complexities: sequence- and machine-dependent setup times, resource assignment constraints, and preemptive scheduling. These extensions, particularly workforce assignments where specific qualifications and availability schedules determine employee eligibility, represent a significant step forward in industrial scheduling research. A Mixed Integer Linear Programming (MILP) model and three constraint-specific variations were developed to evaluate performance and scalability and isolate preemption and resource constraints. Extensive computational experiments demonstrated a trade-off between model applicability and computational efficiency. The proposed model achieved realistic job distribution across machines but encountered scalability challenges due to the combinatorial complexity introduced by what we term dense eligibility matrices, representing a high proportion of potential employee-machine assignments. The preemption-only model optimized makespan effectively, while the resource-focused model provided more practical solutions at the cost of higher processing times. The baseline UPM with sequence-dependent setup times (UPMS) model exhibited computational efficiency but lacked applicability to dynamic industrial environments. This study highlights the impact of preemption and resource assignment on scheduling optimization and underscores the importance of sparsity reduction techniques to enhance scalability. By bridging gaps in workforce management and operational flexibility, the proposed framework provides a robust foundation for addressing complex industrial scheduling challenges.