Malfunction Detection in Multi-cylinder Engines Using Wavelet Packet Dictionary

In this paper, wavelets as signal processing tools are used to analyze the acceleration data acquired at the cylinder head for the detection and characterization of combustion malfunctions in multi-cylinder industrial engines. The objectives were to collect data on 1) normal operations, and 2) operations with a deactivated cylinder to simulate a faulty condition. Wavelet packet and local discriminatory basis algorithm are used to select wavelets that can recognize different conditions. It is shown that the wavelet packet provides a useful data analysis structure for extracting features that are capable of detecting the combustion malfunction of one cylinder in a 12-cylinder engine. Feature extraction is followed by a classification that uses a neural network for the fault identification phase.