Abstract: This paper considers an optimization problem of scheduling operations of main production units in a continuous chemical plant. Although it does not include auxiliary units, it can be viewedasacoreproblem,suchthatitssolutionservesasastartingpoint to construct a detailed schedule. The main features of the problem are the presence of different modes to produce a certain product and sequence-dependent setup times. For its solution, a parallel evolutionary heuristic is developed and implemented for execution on a Graphics Processing Unit (GPU). The key parts of the heuristic are the Iterated Local Search and the “Go with the winners” replacement scheme for selection and sequencing of operations, and the simplified Alternating Direction Method of Multipliers algorithm to find the continuous values of operation durations. This approach is rather simple in structure and naturally fits particular features of a GPU. The experimental evaluation has demonstrated a significant advantage compared to the general-purpose solver Gurobi and a new GPU based solver cuOpt in terms of the solution accuracy and the running time. A comparison with the previously developed Genetic Algorithm has shown an order of magnitude speedup while providing solutions of similar or better quality.

Cite: Borisovsky P.
A GPU-accelerated “Go with the winners” algorithm for solving a large-scale production scheduling problem
Journal of Heuristics. 2026. V.32. 31 :1-24. DOI: 10.1007/s10732-026-09605-5 WOS Scopus OpenAlex

Dates:

Submitted:	Apr 24, 2026
Accepted:	Sep 2, 2026
Published print:	Sep 18, 2026
Published online:	Sep 18, 2026

Identifiers:

≡ Web of science:	WOS:001878687800001
≡ Scopus:	2-s2.0-105051169338
≡ OpenAlex:	W7213552503

Altmetrics: