Valve Event Detection Using Knock Sensor Signals

To make electrical or electro-hydraulic gas exchange valve actuators more economically feasible for the production of cam-less engines, it is desirable to achieve valve position information from existing engine sensors rather than to develop and install new sensors for each gas exchange valve. Time and frequency domain characteristics of acoustic (knock) sensor signals were analyzed using both cross-correlation and Fourier analyses to determine characteristics that indicate valve position. The time-domain signatures for valve events were found to be inconsistent from cycle to cycle and thus difficult to identify using cross-correlation techniques. Frequency content of the engine vibration as it varied with crankshaft angle was analyzed to find valve position dependent patterns. Analysis revealed a wideband signal that correlates to intake and exhaust valve closing events. Tests conducted using several different knock sensors indicate that reliable closing event information is contained within the resonant frequency of the particular sensor utilized. Several filtering schemes were implemented to maximize the coherence of the valve closing event signal. A narrow band-pass filter centered at the sensor's natural frequency provides a good means of valve closure detection using a reasonable number of engine cycles.