CruiseBench Improves Aero-Engine RUL Prediction Benchmarking

Researchers have introduced CruiseBench, a specialized benchmark designed to standardize and improve the evaluation of Remaining Useful Life (RUL) prediction models for aero-engines. This benchmark is derived from the N-CMAPSS dataset, which simulates engine degradation using real-flight profiles. While N-CMAPSS offers increased realism by retaining full within-flight time series, its complexity can make direct comparisons between RUL models challenging due to entangled degradation cues and operational variations. CruiseBench addresses this by focusing specifically on the cruise stage of flights, which is identified using a new artifact called CPM-N-CMAPSS (Cruising-Period Mask for N-CMAPSS). This mask isolates stable cruising intervals, allowing for a fixed evaluation protocol that uses scenario descriptors and measured sensors as inputs, while excluding auxiliary data. By standardizing the data preprocessing and evaluation, CruiseBench provides a more controlled environment for comparing RUL model performance. Initial experiments with various models like LSTM, GRU, TCN, and TSMixer demonstrate baseline results, with TSMixer achieving the lowest average RMSE. The study also highlights how factors like flight-stage selection and RUL-cap thresholds can influence reported outcomes, underscoring the need for such a standardized benchmark.