Abstract

An electronic nose-based condition monitoring of three automobile engines was conducted to obtain smell prints that correspond to normal operating conditions and various induced abnormal operating conditions. Fuzzy Cmeans clustering was used to ascertain the extent to which the smell prints can characterize faulty engine conditions. Silhouette diagrams and silhouette width figures were used to validate the clusters. Results obtained indicate that the smell prints do in general characterize the faults as most clusters have silhouette width greater than 0.5. In particular the results showed that the following automobile engine faults; plug-notfiring faults and loss of compression faults are diagnosable from the automobile exhaust fumes.