Time-frequency analysis of single-point engine-block vibration measurements for multiple excitation-event identification

Abstract

The short-term-Fourier-transform (STFT) is used to identify different sources of IC engine-block vibration from single-point acceleration measurements taken with a commercial knock sensor. Interest is focused on using the STFT to distinguish normal combustion from other sources of excitation including valve impact, injector pulses, and abnormal combustion, such as knocking. Positive identification of these other events using a single method can be useful for pre-processing of measured knock-sensor data for neural-network-based reconstruction of cylinder pressure. It can also be useful separately as part of a fast knock detection system. A series of experiments is discussed to create the data to isolate these different events on a 3-cylinder gasoline engine. In each case, the measured data is processed using the STFT to attempt to isolate the occurrence of particular events in the time domain. Four classes of experiments are undertaken: (i) an un-fired (motored) engine, driven by a dynamometer, with spark plugs fitted, and then removed, to isolate valve impact; (ii) a fired engine running under idle conditions, to contrast no-load combustion with no combustion; (iii) a part-loaded engine running normally, and then running with one injector switched-off, and (iv) a fully-loaded engine running normally, and then running with knock-control switched-off. The paper shows that a single Time-frequency analysis method, applied to knock sensor data in the form of an appropriately-tuned STFT, can effectively identify the occurrence of these events in the time domain if responses are adequately separated and strong enough.