<p>Aircraft engine reliability is paramount to ensuring safe flight operations and cost-effective air transport. Although numerous studies have employed simulated datasets for prognostic modeling, the use of real engine flight data remains underutilized despite its potential to capture the full complexity of in-service conditions. This study proposes an advanced prognostic health management framework that leverages supervised machine learning techniques applied to real sensor data from F100‐PW‐229 turbofan engines. Central to our approach is the introduction of a novel Engine Health Index (EHI), which fuses two critical performance parameters—Performance Margin (PMAR) and Specific Fuel Consumption Margin (SMAR)—to classify engine states into three discrete conditions: “Safe,” “Middle,” and “Unsafe.” Our methodology integrates robust data preprocessing, sophisticated feature engineering, and a cost-sensitive classification strategy designed to penalize dangerous misclassifications. We evaluate a diverse suite of learning algorithms using MATLAB with rigorous fivefold and tenfold cross-validation. Experimental results reveal that tree-based models (Fine, Medium, Coarse, and Bagged Trees) without dimensionality reduction achieve near‐perfect accuracy and zero misclassification cost on test datasets. In contrast, applying Principal Component Analysis (PCA) degrades performance—especially in detecting unsafe conditions. The proposed framework is readily integrable into existing maintenance decision support systems, promising significant improvements in proactive maintenance, flight safety, and resource optimization.</p>