A framework for electronic nose based condition monitoring and diagnosis of automobile engine faults.

Abstract

A framework for condition monitoring approach that uses the sense of smell was investigated to diagnose the faults of plug-not-firing, loss of compression and carburettor faults from the exhaust fumes of gasoline fuelled automobile engine. An electronic nose based condition monitoring hardware and software was developed using the framework to obtain smell prints that correspond to normal operating conditions and various induced abnormal operating conditions. Fuzzy C-means and K means clustering were used as exploratory data visualization tools to ascertain if the obtained smell prints from the developed system could characterize the faults considered. The results of exploratory cluster analysis showed that the obtained smell print could typify the faults considered.