1 Vibration Class at GIST, Korea 5 Fig. 1.5 Vibration excitation methods (a) Vibration exciter with a stinger (b) Impact hammer frequency domain corresponding to each signal are described. Second case is to use an impact hammer as shown in Fig. 1.5b. In this case, the characteristics of excited force in the frequency domain are explained. Also, effects of the header and tip of the hammer corresponding to the frequency range of interest are discussed. Lastly, the advantages and disadvantages of both methods are compared. The sensors for detecting a response and excitation force applied to the structure are explained. There are two kinds of sensors to measure vibration signal, which are contact and non-contact. The contact sensors are an accelerometer and a strain gauge. Because they are attached to the structure, there is a mass added effect. The non-contact sensors are a position sensitive detector (PSD) to measure displacement and a LSV to measure velocity. Because LSV can rapidly measure the vibration of several positions, it makes EMA easier. LSV is a device that was developed in this laboratory; a detailed description is attached in the Appendix. Digital signal processing theory for obtaining FRF from measured data is introduced. Because EMA contains the sampling process which is converting an analog signal into a digital signal, Nyquist sampling theorem as shown in Eq. (1.10) should be described. fs 2f0 (1.10) The aliasing phenomenon in which high-frequency component is detected in the low-frequency component occurs when the sampling rate is not enough. Hence, antialiasing filter is used to prevent the aliasing phenomenon. Fourier transform is explained because frequency conversion is needed to obtain FRF from the measured signal. Also, the leakage in which the frequency component power leaks to adjacent frequency component is explained. The window function which can reduce the leakage is shown and, its principle is explained. In processing techniques of experimental measurement signal, correlation function and spectral density function can obtain the correlation between the two signals in time domain and frequency domain. The linear relationship between the input and output signals (FRF) can be expressed in the correlation function and the spectral density function. And the coherence function which may indicate the degree of noise mixed in the signal via the spectral density function is introduced. Therefore, a coherence function can show whether the characteristic values measured in the experiment are being measured correctly. The peak picking and circle fitting which extracts the modal parameters from experimentally measured FRF are explained. And it is possible to observe the mode shapes using predicted values at some points, i.e. by the mode analysis method introduced.
RkJQdWJsaXNoZXIy MTMzNzEzMQ==