Detection of Defects in Materials Characterised by Large Grain Backscattering Using K-Distribution Model
P. Wu, T. Stepinski, Uppsala University, Uppsala, Sweden
ABSTRACT
A number of statistical models are known which can be used for relating backscattered noise to the microstructure of a solid material. The best-known in NDE are an independent scattering model, predicting rms of the backscattered noise in weakly-scattered materials, and a K-distribution model capable of predicting probability density function (pdf) of the backscattered noise. In this paper we explore the K-distribution model and apply it both to estimation of the grain noise and to flaw detection in the inspected material, with emphasis on the detection. The K-distribution model employs K-distribution to approximate non-Gaussian pdf of the observed RF ultrasonic noise signal. The parameter of K-distribution which characterises the number of scatterers in material is denoted here by Alpha. It is known that the model is valid for prediction of statistical properties of scattering from materials with both densely and sparsely distributed scatterers provided that no single strong scatterers exist. When single strong scatteres (e.g. flaws, voids, etc.) exist in the inspected material, the model may fail. We propose using this feature of the K-distribution model for the detection of flaws in grainy materials. To verify the method experimentally we inspected a copper sample by means of an ultrasonic array system and stored C-scans acquired for various depths. The copper sample had a coarse structure and a number of flat bottom holes. The Alpha parameter in the K-distribution was estimated in three different ways, by calculating the normalised 1st, 4th and 6th central moments from the measured data, respectively. The results have shown that when single strong scatterers were present in the region of interest (ROI), the estimated Alpha parameter resulted in a K-distribution which did not agree with the pdf estimated directly in the ROI, while perfect agreement was observed otherwise. The effects of acoustic diffraction and attenuation on Alpha were investigated by changing the focusing law of the array and adding distance-attenuation-compensation (DAC). The results demonstrate that the attenuation does affect Alpha while the diffraction has a little effect on it.
Publication Source: First International Conference on NDE in Relation to Structural Integrity for Nuclear and Pressurised Components , 20 - 22 October 1998, Amsterdam, Netherlands. Held by the Joint Research Centre of the European Commission. Publisher:Woodhead Publishing Limited
NDTnet
