Employment of an auto regressive moving average model for on-board knock detection of a spark ignited engine

Knock control is still considered one of the main issues of spark ignition (SI) engines. In this apper, a proper mathematical rpocedure for knock detection is presented. More precisely, an Auto Regressive Moving Average (ARMA) algorithm is implemented on vibrational signals acquired by an accelerometer placed on the cylinder block of a SI engine. In order to demonstrate the effectiveness of the methodology, the same analysis is carried out by using the traditional IMPO (Integral of Modulus of Pressure Oscillations) based method, relying on the frequency domain processing of the in-cylinder pressure data. Results demonstrate the high sensitivity and reliability of the porposed mathematical technique in accurately identifying knock events by using the engine block vibrational signals. The comparison with a standard processing of the knock sensor output definitely proves the potential of the model for developing more sensitive on-board control strtegies for knock diagnostics and prevention.