With the increasing complexity of industrial automation, planetary gearboxes play a vital role in large-scale equipment transmission systems, directly impacting operational efficiency and safety. Traditional maintenance strategies often struggle to accurately predict the degradation process of equipment, leading to excessive maintenance costs or potential failure risks. However, existing prediction methods based on statistical models are difficult to adapt to nonlinear degradation processes. To address these challenges, this study proposes a novel condition-based maintenance framework for planetary gearboxes. A comprehensive full-lifecycle degradation experiment was conducted to collect raw vibration signals, which were then processed using a temporal convolutional network autoencoder with multi-scale perception capability to extract deep temporal degradation features, enabling the collaborative extraction of long-period meshing frequencies and short-term impact features from the vibration signals. Kernel principal component analysis was employed to fuse and normalize these features, enhancing the characterization of degradation progression. A nonlinear Wiener process was used to model the degradation trajectory, with a threshold decay function introduced to dynamically adjust maintenance strategies, and model parameters optimized through maximum likelihood estimation. Meanwhile, the maintenance strategy was optimized to minimize costs per unit time, determining the optimal maintenance timing and preventive maintenance threshold. The comprehensive indicator of degradation trends extracted by this method reaches 0.756, which is 41.2% higher than that of traditional time-domain features; the dynamic threshold strategy reduces the maintenance cost per unit time to 55.56, which is 8.9% better than that of the static threshold optimization. Experimental results demonstrate significant reductions in maintenance costs while enhancing system reliability and safety. This study realizes the organic integration of deep learning and reliability theory in the maintenance of planetary gearboxes, provides an interpretable solution for the predictive maintenance of complex mechanical systems, and promotes the development of condition-based maintenance strategies for planetary gearboxes.

Accurate prediction of remaining useful life serves as a reliable basis for maintenance strategies, effectively reducing both the frequency of failures and associated costs. As a core component of PHM, RUL prediction plays a crucial role in preventing equipment failures and optimizing maintenance decision-making. However, deep learning models often falter when processing raw, noisy temporal signals, fail to quantify prediction uncertainty, and face challenges in effectively capturing the nonlinear dynamics of equipment degradation. To address these issues, this study proposes a novel deep learning framework. First, a new bidirectional long short-term memory network integrated with an attention mechanism is designed to enhance temporal feature extraction with improved noise robustness. Second, a probabilistic prediction framework based on kernel density estimation is constructed, incorporating residual connections and stochastic regularization to achieve precise RUL estimation. Finally, extensive experiments on the C-MAPSS dataset demonstrate that our method achieves competitive performance in terms of RMSE and Score metrics compared to state-of-the-art models. More importantly, the probabilistic output provides a quantifiable measure of prediction confidence, which is crucial for risk-informed maintenance planning, enabling managers to optimize maintenance strategies based on a quantifiable understanding of failure risk.

Remaining useful life (RUL) prediction for complex equipment is a critical technology for ensuring the safe and reliable operation of industrial systems. However, existing data-driven models commonly suffer from limitations such as weak cross-operational condition generalization, insufficient physical interpretability, and unstable training on non-stationary time-series data. To address these challenges, this paper proposes a temporal degradation prediction model that integrates context adaptation and physics-consistent constraints, named the Context-Adaptive Physics-informed Time-aware meta-Network (CAPTAIN). The model incorporates four core components: a Context-Aware Meta-Learning (CAML) module that enables lightweight parameter adaptation to diverse scenarios; Physics-Informed Neural Network (PINN) constraints that uniformly characterize deterministic degradation dynamics and stochastic Wiener process perturbations; a three-layer dynamic stabilization training strategy comprising temporal meta-training, residual adaptive refinement, and exponential moving average to ensure training stability; and a multimodal interpretability framework integrating LIME, GradCAM, GradCAM_LW, Integrated Gradients, and KernelSHAP to enhance prediction transparency. Extensive experiments on the NASA C-MAPSS datasets (FD001–FD004) demonstrate that CAPTAIN achieves state-of-the-art performance under both single/multiple failure modes and steady/varying operating conditions, with an average RMSE of 12.02 ± 0.98 and an average SCORE of 487.50 ± 23.0, outperforming ten advanced baseline models. The model exhibits exceptional generalization capability across different operational conditions and strong robustness in scenarios with coupled multiple faults. Multimodal visualizations and quantitative assessments verify its interpretability advantages, showing high consistency with the physical degradation laws of engines. This work provides a reliable paradigm for RUL prediction of complex equipment, combining the flexibility of data-driven modeling with the credibility of physical modeling.

With the growing deployment of unmanned aerial vehicles (UAVs), reliable engine health state assessment (HSA) requires methods that are interpretable, auditable, and transferable under noisy data and varying operating conditions. This paper proposes an AHP-enhanced, data-driven HSA framework that builds a unified health vector from four indicators—remaining useful life (RUL) health, absolute state, relative degradation, and condition health. Indicator weights are derived using AHP with consistency checking, and the resulting continuous health index is mapped through nonlinear stretching and four-level thresholds to produce actionable health grades. Experiments on the NASA CMAPSS benchmark (FD001) evaluate conventional machine-learning models (e.g., XGBoost, SVM, Random Forest, MLP, Logistic) and temporal deep models (CNN-LSTM, Keras DNN). Results show that injecting AHP indicators consistently improves classification performance across models; in particular, AHP-CNN-LSTM achieves 0.92 accuracy and 0.924 Macro-F1, outperforming the CNN-LSTM baseline (0.88/0.872). SHAP-based analysis further supports the contribution of key indicators/features to separating adjacent degradation levels. Additional experiments on FD002–FD004 demonstrate the framework’s applicability under multi-operating-condition settings, providing practical guidance for UAV engine prognostics and health management.