Deconvolution methods for ultrasonic NDT
Florian Bossmann,G. Plonka,Thomas,Peter,O. Nemitz,T. Schmitte
2011-01-01
Abstract:Ultrasonic testing is the method for volume nondestructive testing for industrial materials like steel. It is long established and widely used. However, as the demands on parts integrity increase in various industrial applications the demand on measurement sensitivity is lifted correspondingly. E.g. the energy sector needs more and more high pressure tubing, which in turn leads to higher demands in terms of integrity. In order to increase sensitivity of any measurement system it is necessary to remove unwanted noise from the signals. This would lead to better signal to noise ratio and thus allows higher amplification of signals or lower assessment thresholds. Although there are many denoising techniques available (e.g. wavelet denoising, see [1,10], anisotropic diffusion [14], variational methods [11] or hybrid techniques [5]), the specific problems and the high noise level require adapted denoising and detection methods. In this work we present two deconvolution methods that are both adapted to the applied special ultrasonic pulses. These two methods can be applied in a preprocessing step in order to solve the inverse problem of flaw detection as well as for pure denoising purposes or for efficient storing of the measured ultrasonic data. The first method is a special matching pursuit (MP) algorithm in order to deconvolve the mixed data (signal and noise), and thus to remove the unwanted noise. The second method is based on the approximate Prony method (APM). Both methods employ some prior knowledge about the measured ultrasonic signal. We assume here that the impulse response of the ultrasonic transducer can be estimated or computed in advance. If the impulse response can only be estimated, we offer an iteration method for adjusting the parameters in the used Gaussian echo model. The MP algorithm is F. Boßmann · G. Plonka · T. Peter Institute for Numerical and Applied Mathematics, University of Göttingen, Lotzestraße 16-18, 37083 Göttingen, Germany E-mail: {f.bossmann,plonka,t.peter}@math.uni-goettingen.de O. Nemitz · T. Schmitte Salzgitter Mannesmann Forschung GmbH, Ehinger Straße 200, 47259 Duisburg, Germany E-mail: {o.nemitz,t.schmitte}@du.szmf.de 2 Florian Boßmann et al. used to derive a sparse representation of the measured data by a deconvolution and subtraction scheme. An orthogonal variant of the algorithm (OMP) is presented as well. The APM technique also relies on the assumption that the desired signals are sparse linear combinations of (reflections of) the transmitted pulse. Several test results show that the methods work well even for high noise levels. Further, an outlook for possible applications of these deconvolution methods is given.