Predictive Maintenance of Compressor with LSTM-ADF for Multi-Sensor Multi-Condition Interval Forecasting

Abstract Wind-tunnel compressors and other large rotating machines operate under continuously varying loads and speeds, with dynamics and sensor relationships that evolve across operating regimes. We present LSTM-ADF, an LSTM for adaptive dynamic fusion that meets industrial needs for accuracy, reliability, and deployability with interpretable uncertainty. GraphGate learns inter-channel dependencies; a FiLM pathway conditions on power and speed; multi-quantile outputs yield non-crossing prediction intervals; and a DAP branch provides per-channel risk scores for online monitoring. Physics-informed regularization bounds axial displacement, enforces radial-vibration symmetry, and scales uncertainty with load. We release an engineering-grade compressor dataset (1 Hz synchronized vibration/displacement) capturing realistic nonstationarity and transients. On this dataset, LSTM-ADF matches or exceeds point accuracy while delivering tighter, better-calibrated, and more physically consistent intervals and earlier, steadier anomaly indications than LSTM, Transformer, and TCN baselines—emphasizing reliability, interpretability, and deployable robustness.