Real-time Diagnosis Of Physical Failures Using Causation-based AI

##plugins.themes.bootstrap3.article.main##

##plugins.themes.bootstrap3.article.sidebar##

Published Jun 29, 2021
Navid Zaman Evan Apostolou Yan Li Patrick Conroy

Abstract

Every system in existence is prone to failure and analysis and early detection of said failures (for Predictive Maintenance) is becoming a crucial aspect of modern systems design. Most catastrophic issues start from the smallest parts of a component within the system (physical failures) and minute changes to certain sensor readings from this level may indicate that an incipient failure will occur. Much of this information and system knowledge is often captured during typical RAM activities but is not repurposed for diagnostics. Recent endeavours have been made to utilize AI-based number crunchers to analyse such measurements to notify that an anomaly is observed, however the nature of such correlation-based methods have a limit both on the information on such events and the reliability of the analysis due to spurious correlation. In this paper, we present a novel strategy to 'catch' failures before they happen using a combination of both correlation and causation, i.e. a causation-based AI and demonstrate its advantage over ‘classic’ correlation based methodology.

How to Cite

Zaman, N., Apostolou, E. ., Li, Y., & Conroy, P. (2021). Real-time Diagnosis Of Physical Failures Using Causation-based AI. PHM Society European Conference, 6(1), 7. https://doi.org/10.36001/phme.2021.v6i1.2831
Abstract 305 | PDF Downloads 300

##plugins.themes.bootstrap3.article.details##

Keywords

Diagnostics, AI, Causation, Predictive Maintenance, RAMS, Physical Failure, Data-driven

Section
Technical Papers