Engine diagnosis based on vibration analysis using different fuel blends

Abstract

Fault diagnosis of an internal combustion engine is proposed herein by means of vibration analysis; in order to show the reliability of it, this paper presents a comparative study of normal and faulty scenarios. An engine test bench was used to acquire the vibration signals. For this study, the fault considered on the bench was misfire, which was induced by removing the spark plug wire of a cylinder. Fast Fourier Transform was used to obtain the frequency domain of the signal as a preliminary step to the subsequent identification process based on statistical characteristics extraction. In order to validate previous works on misfire with pure gasoline, measurements included tests performed with ethanol-gasoline fuel blends, namely E30, E20 and commercially available E8 at three different speeds. A simpler classification process was obtained with the extraction of several statistical characteristics from several frequency bands, based on the excited frequency components. The presence of three peaks (at 0.75, 1.25, and 1.5 of the combustion frequency) in the vibration signal of the engine block in the transversal direction for the induced misfire condition, provided differentiation between normal and faulty conditions with all tested fuel blends. According to results, changes in the fuel mix seem to have little impact on the performance and behavior of the engine vibration signals.