A Deterministic Annealing Neural Network Algorithm for Optimal Coal Mining Resource Allocation via Minimum Bisection Problem

Abstract

This paper proposes a deterministic annealing neural network algorithm to address critical resource partitioning challenges in coal mining, such as equipment scheduling, safety zone division, and logistics optimization. By integrating a novel square-root barrier function within a temperature-controlled annealing framework, this algorithm transforms the NP-hard minimum bisection problem into a tractable convex optimization problem with linear constraints. This formulation ensures convergent solutions while effectively balancing operational efficiency and safety requirements. Theoretical analysis rigorously proves the algorithm’s global convergence to discrete partitions, guaranteeing that resources—such as machinery, zones, and transport nodes—are split into balanced groups with minimized cross-group costs. Numerical experiments demonstrate that this algorithm significantly reduces computation time compared to traditional methods, including the Kernighan-Lin algorithm and Networkx, while achieving objective values nearly reaching the theoretical optimum. Notably, the algorithm exhibits strong scalability and stability, with performance advantages becoming more pronounced as graph size increases. Furthermore, tests in a dynamic scenario simulating node failure confirmed the algorithm's capability for rapid rescheduling, a critical feature for real-time adaptation in mining environments. The low variance observed across multiple runs underscores its reliability for consistent decision-making. This work not only introduces a methodologically innovative optimization tool but also provides a practical bridge between theoretical computer science and industrial engineering by reformulating coal-specific problems into the minimum bisection problem framework. The results underscore the deterministic annealing neural network algorithm’s potential as a reliable and efficient decision-support system for intelligent mining operations.