Equipment diagnostics based on comparison of past abnormal behaviors using a big data platform

Carole MAI and Robin CHEVALIER
Submission Type: 
Full Paper
Supporting Agencies (optional): 
phmec_16_021.pdf1.78 MBJune 28, 2016 - 1:00pm

Nuclear monitoring systems provide an increasing amount of real-time measurements to analyse equipment condition. Storage, retrieval, and effective use of information must therefore be made as efficient as possible in order to achieve reliability goals and additional reductions in operating costs. Combining them with other data sources like equipment characteristics, maintenance logs and previous expertise reports will help make better diagnostics while minimizing the time dedicated to an analysis hence leading to better maintenance decisions. This paper presents an overview of the project under study at EDF to develop a case based reasoning platform enabling experts to efficiently retrieve similarities between past events and the current situation. Several technical barriers are encountered such as the comparison of multidimensional time series, of textual information and the extraction of signal features. The problem is here framed as a pattern classification problem where the classes correspond to equipment faults. Similarity criteria have been defined and evaluated against nuclear power plants data for the diagnosis of abnormal patterns on critical equipment. The classification results are compared with service and expertise reports using clustering and classification algorithms. The prospect of this diagnosis is to support the adjustment of maintenance schedule by estimating the remaining useful life of critical equipment.

Publication Year: 
Publication Volume: 
Publication Control Number: 
Page Count: 
Submission Keywords: 
case based reasoning
Pattern classification
Submission Topic Areas: 
Data-driven methods for fault detection, diagnosis, and prognosis
Industrial applications
Submitted by: 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed