LSTM VS. Feed-Forward Autoencoders for Unsupervised Fault Detection in Hydraulic Pumps
This addresses early fault detection for industrial hydraulic pumps to prevent unplanned failures and costs, but it is incremental as it compares existing methods on a specific dataset.
The paper tackled unsupervised fault detection in hydraulic pumps by comparing feed-forward and LSTM autoencoders trained on healthy data, achieving high reliability without fault samples during training.
Unplanned failures in industrial hydraulic pumps can halt production and incur substantial costs. We explore two unsupervised autoencoder (AE) schemes for early fault detection: a feed-forward model that analyses individual sensor snapshots and a Long Short-Term Memory (LSTM) model that captures short temporal windows. Both networks are trained only on healthy data drawn from a minute-level log of 52 sensor channels; evaluation uses a separate set that contains seven annotated fault intervals. Despite the absence of fault samples during training, the models achieve high reliability.