Identification aveugle paramétrique basée sur la cyclostationnarité à l’ordre deux de la fonction de transfert à partir des mesures vibratoires

This paper addresses the blind identification of the impulse response function from only the response measurement. The method consists in modeling the measured vibration signal as the output of a AutoRegressif Moving Average ARMA filter. An identification method, based on the parameter extraction of ARMA filter from the cyclic spectral correlation, is proposed in order to identify the transfer function of a spindle-tool-workpiece system of a high speed milling process. It is shown that the vibration signal coming from the milling process is cyclostationnary. The performance of this algorithm is demonstrated on a simulation signal and an experimental application has been validated on a test bench of a high speed milling using two accelerometers. The ability of this method to extract the impulse response from only the vibration measurement is demonstrated. This allows us to identify "online" the system dynamic without the use of classical modal methods which require that the machine be off.