Extraction of Transient Signals From Nonperiodic Dynamic Responses

In practice, vibration signals generated by machines that are considered both linear and excited periodically are often slightly nonperiodic. A small variability in the periodic response may result from small variations of speed in the driving motors, in small variation in the system parameters, or nonlinearities causing mild chaotic motion. This paper deals with the problem of extracting transient components from almost periodic responses. One of the difficulties involved in the problem results from the presence of a broad range of frequencies in a transient response. The main difficulty, however, comes from the fact that the steady-state response is not purely periodic and the transient is totally submerged in it. Two major approaches of extracting the transient component have been investigated: algebraic subtraction and adaptive notch filtering. Both methods have been tested numerically before they were applied to the signals obtained from experiments. The results show that it is possible to extract transients from a dynamic response even if the period of the response varies.