A digital twin-driven installation resource dynamic scheduling method of fully mechanized mining faces

Abstract

Existing resource scheduling methods for fully mechanized mining face installation fail to account for variations in resource demand during the installation process, resulting in resource shortages or redundancies that cause delays and disrupt normal production continuity. Here, we propose a digital twin-driven dynamic installation resource scheduling approach, integrating digital twin technology, neural network-based demand forecasting, and multi-objective optimization. A hierarchical digital twin system was constructed, comprising the twin model layer, data layer, and physical entity layer, which collaboratively facilitated real-time, adaptive scheduling of installation resources. The mechanism was elucidated through two phases: Pre-installation resource configuration and in-process dynamic scheduling, achieving closed-loop optimization of the installation resource management. Additionally, a real-time mapped twin model layer was developed to reflect the current state of the physical installation environment. The demand forecasting model employed an Adam-LSTM neural network to quantify installation resource requirements, generating demand predictions. The dynamic scheduling model utilized multi-objective optimization to balance physical constraints and operational goals, integrating demand forecasts to simulate the physical resource allocation process and derive optimal scheduling schemes. Application of this methodology to actual installation operations demonstrated that, compared to traditional approaches, the dynamic scheduling scheme reduced total costs by 75.66%, potential revenue loss by 73.35%, and inventory holding costs by 93.58%, offering a novel solution for rapid relocation of fully mechanized mining faces.