A Convolutional Neural Network to Locate Unbalance in Turbomachinery Supported by AMBs

Due to the complex structure of turbomachinery systems, the process of fault detection assumes paramount importance, in particular rotor unbalance faults are particularly risky and common. This research paper introduces an innovative and straightforward approach to locate rotor unbalance faults for turbomachinery supported by Active magnetic bearings (AMB) exploiting the AMB sensors and utilizing Deep Learning techniques (1D Convolutional Neural Networks). The main goal of this study is to develop a fault dictionary, built using fault signatures derived from position sensor signals, and a classifier specialized in locating the unbalance faults in turbomachinery supported by AMBs, that generally occur in the turbomachine impellers. These are the most prevalent unbalance faults that affect turbomachinery systems and that commonly impact the performance of AMB systems. The effectiveness of this approach is demonstrated through a case study involving an expander-compressor supported by two active magnetic bearings in the oil and gas field. Five distinct fault classes are considered, and the neural network fault classifier achieves an impressive accuracy rate of 98% on the test dataset.