<p>Predicting the remaining useful life (RUL) of aircraft engines is critical for maintaining safety, optimizing maintenance planning, and minimizing lifecycle expenses. While deep learning models have shown strong potential in predicting RUL, their effectiveness is fundamentally constrained by the scarcity of labeled training data under real-world operational conditions. Although active learning (AL) presents a viable solution to alleviate data annotation burdens, existing research predominantly focuses on single-instance query scenarios while neglecting the practical requirements for batch-mode AL in industrial applications. This paper proposes a cluster edge-based active learning with batch queries method, which establishes geometric distance metrics in the feature space to systematically select the farthest boundary samples from cluster centroids during each iteration. The proposed method optimizes the synergy between annotation resources and model generalization capability by maximizing geometric dispersion within homogeneous clusters. Experiments conducted using the NASA C-MAPSS dataset have demonstrated that our approach outperforms other baselines in terms of prediction accuracy, particularly when the number of training samples is limited. The results establish a new state-of-the-art for data-efficient prognostic modeling, providing significant implications for intelligent maintenance systems in aerospace engineering.</p>