Combinatorial optimization of resource distribution processes in organizational systems based on neural networks

Combinatorial optimisation is now widely used in science, industry and expert circles. Decisions on resource allocation, experiment design, task planning, or choosing the most effective path often have to be made from a multitude of options. In this class of problems, combinatorial optimisation methods have significant potential be-cause they can help find optimal or near-optimal solutions, contributing to more informed and rational decisions. With this in mind, this article is devoted to the study of issues related to the application of combinatorial optimisation methods for the design of production processes with limited resources. The research describes a design approach that takes into account the needs of a specific production process, the availability of resources, the volume of their use, and the sequence of their use. Special attention is paid to the selection and justification of the objective function. Based on the results obtained, the article formalises a combinatorial optimisation algorithm implemented on the basis of the HyperGNN neural network. The proposed algorithm has been tested on the example of a production chain of a machine-building enterprise. During testing, real, clearly defined resources were considered: a milling machine, a welding machine, grinding, heat treatment, and assembly. All tasks had a specific duration, logical sequence, and fixed requirements. The data obtained demonstrate the high practical significance of the developed algorithm. During its comparative analysis with the greedy resource allocation method, it was found that the proposed approach significantly outperforms traditional planning methods in a number of key criteria, including a reduction in the total production cycle time, a decrease in the number of conflicts over resources, a more even load on equipment, and a reduction in downtime between operations.