Performant algorithm for spatial scheduling of large structures within computing-intensive simulation projects

Abstract

The production planning of large structures is characterized by a variety of domain-specific aspects. One of the most important is spatial scheduling, because unlike in conventional industries, orders are not allocated to resources on a unit basis. Instead, the allocatable capacity is defined by the available production area and thus the number of objects that can be processed at a time is determined by their dimensions. Therefore, a space-efficient arrangement of objects on the available production area is a key planning task in industries that deal with large structures. Since there is theoretically an infinite number of possibilities for arranging objects on a production area, the planning task is to be seen as an optimization problem, whose solution requires an enormous amount of computing effort. If a large-scale project is to be evaluated over the entire project period as part of a simulation study, this optimization problem often has to be solved multiple thousand, in some cases million times. It is therefore not uncommon for spatial scheduling algorithms to require several hours of cumulative computation time within simulation studies, which significantly slows down the decision-making process. This indicates that the use of a lean algorithm for spatial scheduling is essential to ensure the usability and acceptance of simulation tools for large-scale projects in practice. However, this must not have a significant negative impact on the planning accuracy. In this paper, such a lean algorithm for two-dimensional spatial scheduling problems is presented. The algorithm is explained by means of practical examples and its performance is compared with conventional algorithms in order to evaluate the trade-off between computation time and spatial efficiency. In addition, the impact on the lead time calculated by the simulation tool is analyzed.