Determination of sample size using power analysis and optimum bin size of histogram features
Vibration signals are used in fault diagnosis of rotary machines as a source of information. Lots of work have been reported on identification of faults in roller bearing by using many techniques. Of late, application of machine learning approach in fault diagnosis is gaining momentum. Machine learning approach consists of chain of activities like, data acquisition, feature extraction, feature selection and feature classification. While histogram features are used, there are still a few questions to be answered such as how many histogram bins are to be used to extract features and how many samples to be used to train the classifier. This paper provides a mathematical study to choose the bin size and the minimum sample size to train the classifier using power analysis with statistical stability. A typical bearing fault diagnosis problem is taken as a case for illustration and the results are compared with that of entropy based algorithm (J48) for determining minimum sample size and bin size.
Volume (Year): 3 (2011)
Issue (Month): 1 ()
|Contact details of provider:|| Web page: http://www.inderscience.com/browse/index.php?journalID=282|
When requesting a correction, please mention this item's handle: RePEc:ids:injdan:v:3:y:2011:i:1:p:21-41. See general information about how to correct material in RePEc.
For technical questions regarding this item, or to correct its authors, title, abstract, bibliographic or download information, contact: (Graham Langley)
If references are entirely missing, you can add them using this form.