A Study on the Theoretical Background of
Autoregressive Based Damage Detection Scheme
Using Transfer Function
Kwang Youn Park
This thesis presents a background of autoregressive based on a damage detection scheme, suggesting a transfer function. Adopting a transfer function of autoregressive model, it is possible to separate a frequency band by filter pass band; one part of a frequency band is shown as a predictable condition and the other part is distorted abnormally.
It is including the criterions to get proper coefficients, sampling rate and autoregressive order with the basis of reinforced theory as well. The changes of sampling rate and autoregressive order affect the alteration of transfer function in an autoregressive model. It is possible to get better transfer function to detect damage by controlling of sampling rate and autoregressive order.
Additionally, it is mentioning advantages of a non-casual filter for damage detection. Non-causal filter, utilizing past data and future data at once, is relatively easier to establish filter design. The reason is that transfer function of non-causal filer has a wide and flat filter pass band.
Transfer function, frequency domain analysis, non-causal filter, autoregressive model, detection of abrupt change, damage feature