High-resolution estimation and on-line monitoring of natural frequency of aircraft structure specimen under random vibration

The growth of structural damage reduces the stiffness which results in decrease in the natural frequency of a specimen (coupon) under random vibration. However, the identification of the natural frequency is not a trivial problem because the coupons are shaken in a random fashion in which many closely-spaced frequencies are involved. The signal may have other redundant frequencies involved which are close to the natural frequency The simplest way to identify the frequencies of multiple sinusoids is the Fourier transform spectral analysis. However, when the frequencies of interest are closely spaced, the Fourier transform spectrum analysis fails to resolve these frequencies. In this paper these facts are shown for a real vibration signal. High resolution frequency estimation approach, such as the forward-backward linear prediction (FBLP) method, is adopted here to identify all the prominent frequencies in a certain bandwidth. From the p^th order FBLP model, we estimate p positive frequencies. Among these p frequencies, the one that will decrease as the structure degenerates, is the natural frequency and is monitored in an on-line mode. An estimation of natural frequency and its prediction is made regularly (every one second) based on all available estimated natural frequencies once the structure enters the fatigue zone. The system is useful for on-line health monitoring of any structure under random vibration. We have applied this system to monitor several aluminum coupons of aircraft structure. The tests were performed at the Wright Research Lab and the software monitoring system was developed at Tennessee State University