Application of Crow-AMSAA to Predict Failure of Centrifugal Pumps with Increasing Failure Rates

The purpose of this research is to focus on the failure model analysis on the
centrifugal pumps data using Crow-AMSAA. The data were analyzed using two
trend test methods which were Mann test and Laplace test. The data that satisfied
those tests were subjected to Crow-AMSAA failure model analysis. This project
also analyzed and evaluated the differences in the result acquired with the CrowAMSAA
analysis and the actual data. The accuracy between the two parameters
determined the accuracy of Crow-AMSAA analysis. Each of the pump selected, for
which the criteria of selection were pumps with more than 5 failure occurrences, the
result of each analysis procedure were reported. The graphs that were plotted in both
the trend test and the Crow-AMSAA analysis were presented in this report. From the
results of the analysis, the accuracy of the failure prediction of centrifugal pumps
that were made using Crow-AMSAA as the prediction model were accurate with an
error range of 16% to 19% percent.