Integration of hierarchical clustering to improve the prognostic results of fuzzy similarity

Jie Liu and Enrico Zio
Submission Type: 
Full Paper
AttachmentSizeTimestamp
phmec_16_060.pdf1.77 MBJune 29, 2016 - 3:20am

Fuzzy similarity has been widely used for prognostics. Normally, a library of failure scenarios is available and a number of them most similar to the observed scenario are selected to generate the Remaining Useful Life (RUL) of the observed scenario, using a distance weighted-sum approach. By clustering the library of failure scenarios, those most similar to the observed scenario can be selected considering the strength of membership to the different clusters. To this aim, in this paper, hierarchical classification is integrated into the fuzzy similarity approach. First, hierarchical classification is built by Support Vector Machine (SVM), considering different failure modes. Then, for the observed scenario, fuzzy similarity is applied to select the most similar failure scenarios from different classes, considering the membership of the observed scenario to the different clusters. The selected scenarios are aggregated to generate the RUL along the observed scenario. A real case study of a system composed of a pneumatic valve and a centrifugal pump in a nuclear power plant is considered to verify the RUL prediction power of the proposed framework.

Publication Year: 
2016
Publication Volume: 
7
Publication Control Number: 
060
Page Count: 
8
Submission Topic Areas: 
Data-driven methods for fault detection, diagnosis, and prognosis
Submitted by: 
  
 
 
 

follow us

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