Separation of combustion noise in IC engines under cyclo-non-stationary regime

The separation and ranking of combustion and mechanical noise sources is of prime concern for the noise control of internal combustion (IC) engines. Signal processing techniques have been devised recently that can achieve such a separation using the cyclostationary property of IC engine signals. The object of this paper is to extend this framework to the situation where the engine undergoes a transient speed regime, for instance during a run-up. This raises some new and non-trivial questions. First, the assumption of cyclostationarity has to be relaxed and replaced by the vaguer notion of “cyclo-non-stationarity”. Second – and related to the first point – the practice of cyclic averaging has to be revisited. Third, the design of the separation filter must explicitly incorporate speed dependence. This paper proposes simple but robust solutions to these issues, with a special effort to make them practicable from an industrial point of view. In particular, the cyclic difference operator is introduced in lieu of cyclic averaging, and speed-dependence is captured by use of a flexible basis of B-splines whose knots density is automatically selected from the data. Successful examples of separation are then demonstrated on actual data measured during an engine run-up.

► Separation of noise sources is crucial in automotive engineering. ► Efficient current methods are based on the assumption of cyclostationarity which holds only for stationary speed regimes of the engine. ► An extension is proposed to non-stationary regimes which produce “cyclo-non-stationary” signals. ► A generalisation is proposed to TSA and cyclic Wiener filtering.