Turbofan Engine Prognostics using Regression and Hidden Markov Models on the "New C-MAPSS" Dataset

This thesis presents a comprehensive approach to Turbofan Engine Prognostics using Regression and Hidden Markov Models (HMMs) applied to the New Commercial Modular Aero-Propulsion System Simulation (N-CMAPSS) dataset. The N-CMAPSS dataset, an evolution of the widely used C-MAPSS benchmark, provides high-fidelity, continuous sensor data from simulated turbofan engines under realistic flight conditions. The dataset includes multiple failure modes, operational settings, and auxiliary data, enabling robust validation of the proposed prognostic framework. The work focuses on Prognostics and Health Management (PHM) for aerospace applications, where accurate Remaining Useful Life (RUL) estimation is critical for maintenance optimization and operational safety. The methodology integrates time-series modeling with Hidden Markov Models (HMMs) to capture the degradation patterns of turbofan engines. Key steps include data preprocessing, feature extraction, health indicator generation, and HMM training with Mixture of Gaussians (MoG) for emission modeling. The approach leverages clustering techniques to segment operational states and online detection algorithms to predict RUL in real-time. Results demonstrate the effectiveness of the combined regression-HMM approach in detecting degradation trends and estimating RUL. This work contributes to the broader field of predictive maintenance by providing a framework for real-time health monitoring of complex systems, such as aviation and industrial asset management.