STATISTICAL ENGINE KNOCK DETECTION

Abstract A knock detection circuit which is based on the signal of the accelerometer installed on the engine block of spark ignition automotive engine has a band-pass filter with a certain frequency as a parameter to be calibrated. New statistical method for determination of the frequency which is the most suitable for the knock detection in real-time applications is proposed. The knock detectability is associated with a statistical hypothesis for testing the equality of two mean values, i.e., mean values of the amplitudes for knocking and non-knocking cycles, and signal-to-noise ratio is associated with the value of the statistic. The frequency with the largest signal-to-noise ratio is chosen for implementation in the engine knock detection circuit.