Speaker
Description
Several industries have rapidly adopted Predictive Maintenance (PdM) programmes at the Norwegian continental shelf during the project and utilisation phases. Predictive Maintenance is a cyber-physical system that combines detection techniques, with diagnostics and prognostics algorithms. The capabilities of the applied techniques and algorithms (what is measured, analysed, treated, and how accurate and precise) shall be evaluated. There is still a lack of knowledge of how these techniques and algorithms behave under specific combinations of fault type, fault severity (fault evolution), fixed and variable operating conditions, multi-fault situations, overlapping failure mechanisms, and multi-unit systems. The concept of the probability of detection (PoD) curve can be transferred from the Non-destructive testing (NDT) field of knowledge into the predictive maintenance field and expanded to cover the diagnostic and prognostic algorithms. Therefore, the purpose of this paper is to demonstrate how the probability of diagnostic and prognostic curves can be created to evaluate newly emerging algorithms. Experimental work is performed with a bearing fault simulator test rig at the University of Stavanger, where five diagnostic and prognostic algorithms are tested. The implication of this study is related to maintenance engineers who want to know which algorithm is the most precise, gives the earliest indication of the fault and provides the best predictive horizon in advance.
| Conference Topic Areas | Track7: Smart Operations and Maintenance |
|---|
