Condition Monitoring of a Reciprocating Compressor Using Wavelet Transformation and Support Vector Machines

Shawn Falzone and Jason R. Kolodziej
Submission Type: 
Full Paper
AttachmentSizeTimestamp
phmc_17_005.pdf734.98 KBSeptember 7, 2017 - 5:14pm

Condition monitoring techniques were applied to a reciprocating compressor in order to determine if faults were present in a system. Through the use of vibration based sensors, fault monitoring of the crank-side discharge valve springs was accomplished. Data was collected through a range of injected fault conditions and analyzed through the use of discrete wavelet transformations. The wavelet coefficients produced were transformed into a six-dimensional feature space though the use of first and second order statistics. By using a support vector machine classifier, the nominal and faulted condition data was used to train a fault monitoring classifier. This classifier was verified through the use of additional test data, and resulted in classification rates of 90% and above. This result is based on the trial of a multitude of different wavelets and support vector kernels in order to achieve the optimal performance for the dataset.

Publication Year: 
2017
Publication Volume: 
8
Publication Control Number: 
005
Page Count: 
7
Submission Keywords: 
condition based maintenance (CBM)
data driven methods
applications: industrial
Submission Topic Areas: 
CBM and informed logistics
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