Learns to Read a Jet Engine's Remaining Lifespan Through Noise and

A new research paper in Results in Engineering proposes a deep learning framework to estimate an aircraft jet engine’s remaining useful life while explicitly accounting for uncertainty. Authored by Peng Peng, Xueqiang Fan, Bing Lin, Xixun Sun, Zhiping Yin and Zhongyi Guo of Hefei University of Technology, the approach is called UP2-GTN: a physics-informed uncertainty-aware patch-level graph temporal network. The model targets the reality that sensors in engines face noise, drift, electrical interference and changing operating conditions. During training, it corrupts inputs using engineered Gaussian, impulse, scaling and drift noise. It also segments time-series sensor history into short sections to reduce error propagation. The study is positioned as a step toward more reliable maintenance predictions.