Noise Separation Experiment for Flexible Structures Using Variable Mass

Abstract

This paper presents a technique for separating unknown measurement noise from vibration data. The technique is based on the spectral subtraction using variable structural parameters of adaptive structures. The difference between two periodograms obtained from measured vibration data is calculated, where each periodogram is corresponding to the different value of the variable structural parameter, and resulting difference is normalized with its maximum value. Then, it is multiplied to each periodogram to reduce the unknown measurement noise included in vibration data, and thus present technique needs no prior knowledge of properties of noise. Simple numerical examples using a spring-mass system with a variable parameter are demonstrated. Also, experimental results are presented to show the feasibility of the proposed technique using the appendage mass attached to a flexible cantilevered beam structure. The lower modal frequencies of the test beam are shifted by the variation of the appendage mass, and the present technique is implemented for separating measurement noise including random and periodic noise. The results are also compared with that obtained by a conventional averaging technique such as Welch’s method, and consequently it is indicated that the present technique is suitable for extracting the system’s dominant mode in the presence of periodic noise or colored noise, which is difficult to be reduced by the averaging method.