On the Selection of Sensor Locations for the Fictitious FRF based Fault Detection Method

Abstract

Recently a machinery fault detection technique based on the fictitious frequency response function (FRF) was developed especially when a measurement sensor cannot be attached to an ideal point due to geometric or environmental restrictions. The fictitious FRF based fault detection method has the great flexibility in regards to choice of sensor locations such that they can be selected almost arbitrarily. However, the method requires two sensors and assumes low correlation between two simultaneously measured signals within the interested frequency range. This assumption of low correlation is vague in practical applications. In this paper, the fault diagnostic performance of the method is investigated in regards to the correlation coefficient between two measured signals. A practical guideline for the selection of the two sensor locations is then proposed.