Bidirectional LSTM-based soft sensor for rotor displacement trajectory estimation

Constant rotor system monitoring enables timely control and maintenance actions that decrease the likelihood of severe malfunctions end product quality deficits. Soft sensors represent a promising branch solutions enhancing monitoring. A soft sensor can substitute malfunctioning physical provide estimates quantity is difficult to measure. This research demonstrates based on bidirectional long short-term memory (LSTM), training procedure for at high sampling frequency varied operating conditions. study adopts large bearing vibration dataset. The accurately lateral displacement trajectories from reaction forces over range constant rotating speeds support stiffnesses. mean absolute error (MAE) LSTM-based 0.0063 mm test in complete condition space. performance shown significantly MAE 0.0442 mm, if dataset limited speed range.