Electrical Motor Current Signal Analysis using Modified Bispectrum for Fault Diagnosis of Reciprocating Compressors

The induction motor is the most common driver in industry and has been previously proposed as a means of inferring
the condition of an entire equipment train, predominantly through the measurement and processing of power supply
parameters. This has obvious advantages in terms of being non-intrusive or remote, less costly to apply and improved
safety. However, it is difficult to perform diagnosis for varying load machines because the spectral signal are the same.
This paper presents the use of the induction motor current to identify and quantify a number of common faults on a
two-stage reciprocating compressor which produces varying load to the driving motor. Bispectrum representation of
current allows the inclusion of phase information and the elimination of Gaussian noise. It will provide more details of
the machine conditions. Moreover, a modified bispectrum based the amplitude modulation feature of the current signal
produces more accurate presentation than the conventional bispectrum. Based on this new bispectrum, a more
effective diagnostic feature - averaged bispectral peak is developed for fault classification. In conjunction with the
kurtosis from the current the bispectrum feature gives rise to reliable fault classification results. The low feature values
can differentiate the belt looseness from other fault cases and valve leakage and inter-cooler leakage can be separated
easily using two linear classifiers. This work provides a novel approach to the analysis stator current data for the
diagnosis of motor drive faults.