Loading [a11y]/accessibility-menu.js
Zero-Shot Learning for Intelligent Fault Detection | IEEE Conference Publication | IEEE Xplore

Zero-Shot Learning for Intelligent Fault Detection


Abstract:

Signal-based fault detection, as an essential technology in many engineering and industrial applications, has received extensive attention. Nevertheless, in a real-world ...Show More

Abstract:

Signal-based fault detection, as an essential technology in many engineering and industrial applications, has received extensive attention. Nevertheless, in a real-world application setting, collecting samples for some specific types of faults may be time-consuming and damaging, implying that samples of unseen faults may be unavailable in the training step. Keeping this in mind, we present a new semantic space-based zero-shot learning (SSB-ZSL) method for fault detection. The implementation of the method comprises three phases: feature extraction, human-defined semantic space, and a feature embedding model to address this issue. To evaluate the proposed method, we carry out experiments on two real datasets: a vibration-based rolling element bearing dataset from Case Western Reserve University, and an acoustic-based air compressor fault dataset. The results are promising and significantly better than that given by the compared state-of-the-art methods, showing that the proposed method can effectively detect unseen types of faults in the absence of their samples.
Date of Conference: 01-03 September 2022
Date Added to IEEE Xplore: 10 October 2022
ISBN Information:
Conference Location: Bristol, United Kingdom

References

References is not available for this document.