| 要約: | The accuracy of hydraulic support load prediction in deep-well mining environments is critical for ensuring the safe operation of coal mines. Existing hydraulic support load prediction methods for deep-wells are usually passive, which affects the real-time accuracy of load prediction. Therefore, this paper proposes a novel method for predicting hydraulic support loads based on digital twin technology and multi-agent systems. Firstly, we introduce architecture for a digital twin-based predictive system that integrates multi-agent capabilities. This system facilitates interactive mapping and synchronous feedback between the physical entity of the hydraulic support and its corresponding digital twin through data-driven processes. We propose a model-driven multi-agent application system architecture that enables agents to make decisions and process complex tasks based on model-driven methodologies. Then, we constructed a hybrid prediction model called DPDF-Net and utilized it to predict the support loads in multi-support zones of the working face. The accuracy of our proposed model's predictions is verified through experimental results. Furthermore, ablation experiments are conducted to assess the contribution of different modules to overall model performance. Comparative analyses reveal that our proposed model achieves an R² value exceeding 0.97 in multi-zone load predictions, significantly outperforming the comparison model. This provides theoretical insights for rock burst prevention in deep mining.
|
|---|
