Filtrowanie danych obciążenie-deformacja w pomiarach zamykania się pęknięcia

Rozważono metodykę szumu w sygnałach siły i odkształcenia rejestrowanych w pomiarach zamykania się pęknięcia metodami podatności. Przeprowadzona metodą Monte Carlo walidacja trzech różnych technik numerycznego filtrowania sygnałów wykazała,że funkcje sklejane trzeciego stopnia umożliwiają optymalne wygładaznie danych pomiarowych w zaproponowanym przez Autorów algorytmie do oceny zamykania się pęknięcia.

---

The type of compliance analysis employed in an algorithm for crack closure evaluation developed by the present authors is conditioned by an appropriate smoothess of the measured load-strain data. In the present paper, three candidate techniques for digitalsmoothing the recorded load and strain signals are considered, namely the moving average method, the Butterworth low-pass filter and the cubic smoothing spline method. Tuning the above-mentioned techniques has been performed based on a visual examination of loadoffset strain loops obtained from the raw and smoothed load-strain data and from studyingthe behaviour of the corresponding residuals. To judge the effectiveness of the smoothing procedures, the present algorithm has been employed to obtain a Monte Carlo simulation of the entire closure measurement process. The results, evaluated in terms of the bias and scatter, indicated that of the three procedures for filtering noisy signals studied, the spline technique is most recommendable to smooth the load-strain data. At the same time, these results demonstrate the significance of minimizing the amount of measurement noise in situ, as no filtering technique can fully compensate for a deterimental effect of a severe noise on the accuracy and reliability of crack closure estimates using the present algorithm.