A Variable Neighborhood Search Using Very Large Neighborhood Structures For The 3-Staged 2-Dimensional Cutting Stock Problem

In this work we consider the 3-staged 2-dimensional cutting stock problem, which appears in many real-world applications such as glass and wood cutting and various scheduling tasks. We suggest a variable neighborhood search (VNS) employing "ruin-and-recreate"-based very large neighborhood searches (VLNS). We further present a polynomial-sized integer linear programming model (ILP) for solving the subproblem of 2-staged 2-dimensional cutting with variable sheet sizes, which is exploited in an additional neighborhood search within the VNS. Both methods yield significantly better results on about half of the benchmark instances from literature than have been published before.