<![CDATA[Data-driven prognostics and health management (PHM) for turbofan engines requires a Health Index (HI) that is learnable from multivariate telemetry and credible as a basis for maintenance decisions. This study presents a deep learning-based HI modelling framework on the N-CMAPSS benchmark that converts operating conditions and sensor streams into a bounded HI and, subsequently, into decision-oriented outputs for predictive maintenance. A baseline convolutional model is benchmarked against a residual dilated CNN to capture multi-scale degradation signatures from fixed-length temporal windows. To preserve evaluative integrity, health-zone thresholds are calibrated on validation predictions and then fixed, producing a three-zone taxonomy (critical, warning, healthy) for rapid field triage, alongside a continuous risk score that induces a rank-ordered maintenance priority list from most critical to most healthy. The selected model achieves HI regression performance of RMSE = 0.1266, MAE = 0.0720, and R² = 0.7241, while the calibrated zone mapping attains accuracy = 0.8688 and macro-F1 = 0.6124. The main contribution is a leakage-aware, decision-coupled pipeline that delivers both interpretable health zoning and risk-ranked prioritization, strengthening the operational linkage between predictive modelling and maintenance triage within PHM-oriented Informatics.]]>
Copyrights © 2026
