Aircraft engines have peculiar thundering sound that is produced due to high rate of fuel-air combustion. However due to improper combustion caused by non proportional air-fuel ratios, malfunction of fuel feed or any other loose gauging, certain abnormal sounds are produced that indicate fault occurrence. This research is an attempt to automate fault detection in aircraft jet engines by acoustic observations. The experimental analyses were based on sound analysis of Teledyne J69 Jet Engine. The major work revolves around knowledge of noise features and methodology to extract these features for analysis. Frequency spectrums of audio signals from the engines under test were compared to observe the differences generated by fault noise. A total of 140 samples were analyzed for noise feature extraction and a threshold for noise magnitude was empirically found for automated fault noise identification. This system was extensively tested over 700 audio samples from 76 different engines of same type (Teledyne J69) that lead to the system’s accuracy of 98%.
