PRINCIPAL COMPONENT ANALYSIS DAN NAIVE BAYES CLASSIFIER UNTUK MENDETEKSI KAVITASI PADA POMPA SENTRIFUGAL

Cavitation is one type of damage that occurs in centrifugal pumps, where the
damage is caused by a decrease in pressure below the saturated vapour pressure.
The failure that occurs causes a reduction in pump performance or can cause
damage to other components. Therefore, a method is needed to detect cavitation
early so that corrective action can be taken immediately.
Pattern recognition is one way to diagnose damage early by using a vibration
signal pattern obtained from the extraction of time-domain statistical parameters.
The parameters used were root mean square, standard deviation, kurtosis,
variance, skewness, peak value, mean, crest factor and shape factor. This study uses
a naïve Bayes classifier and principal component analysis to obtain more accurate
results. The variations of the classified pump conditions are standard pump, level
1 cavitation, level 2 cavitation and level 3 cavitation.
This study resulted in a naïve Bayes accuracy of 98.4% for the first parameter data
set and 99.4% of the second parameter data set. The result of the combination
accuracy of naïve Bayes and principal component analysis is 94.6% of the 4PC
used.