Ultrasonic imaging of a turbine blade model using a 360° synthetic-aperture-focusing-technique and reverberation suppression Thomas Scharrer * , Andreas Koch §* , Stefan J. Rupitsch * , Alexander Sutor * , Helmut Ermert §* , Reinhard Lerch * * Chair of Sensor Technology, Friedrich-Alexander-University Erlangen-Nuremberg § High Frequency Engineering Research Group, Ruhr-University Bochum Email: thomas.scharrer@lse.eei.uni-erlangen.de Abstract—In this contribution, we present a method to comple- ment missing X-ray computed tomography (CT) data of irregular shaped metallic specimen, like turbine blades, by ultrasonic testing in immersion mode. Therefore, an adapted 360° synthetic- aperture-focusing-technique (SAFT) is used to obtain particular information of the inner pattern of the specimen. These structural information cannot be imaged by X-ray CT due to limited pene- tration in specific directions and consequently absent projections in the dataset. However, arising artifacts caused by refraction and signal reverberation impede the detection of the desired structural information by ultrasonic testing of those irregular metallic specimens. To cope with refraction effects caused by the immense discrepancies in speed of sound (SOS) at the interface of the couplant and the specimen, the approach is based on virtual source elements. Furthermore, the covering of near- surface structures by the first reflection signals is treated by a subspace based filtering approach and signal reverberation artifacts are suppressed by predictive deconvolution. I. I NTRODUCTION Today, X-ray 3D cone-beam computed tomography (CT) is a common method in non-destructive evaluation of turbine blades [1]. However, materials used in such investigations are intensely scattering and consequently, artifacts such as cupping, streaks and a reduction in contrast arises in the reconstructions [2]. For large scale samples, limited pene- tration is observed additionally. This leads to missing spe- cific projections and therefore to blurred edges in the non- transmissive directions (Fig. 1). An approach to eliminate such artifacts is the supplementation of the incomplete dataset by complementary information gained with other testing methods (e.g., ultrasonic testing, optical 3D surface scans) [3]. The 3D scan enables a correction of blurred contour edges while ultrasonic testing copes with inner structures. In this contribution, an ultrasonic evaluation method is presented to provide such complementary data of irregular shaped metallic specimens with the a priori knowledge of the specimen’s contour. To account for these curved surfaces of the specimens, it is reasonable to use ultrasonic immersion mode testing. However, due to the substantial discrepancies in the speed of sound (SOS) of the couplant with respect to the specimen’s material, the detection of desired structural 0 0.2 0.4 0.6 0.8 1 10 mm normalised houndsfield unit blurred edges Fig. 1: X-ray CT reconstruction of the examined specimen. Due to limited penetration, artifacts arises as blurred edges of inner structures and the contour. information is impeded by refraction and reverberation ef- fects. Therefore, an adapted 360° synthetic-aperture-focusing- technique (SAFT) based on a virtual point source is applied to cope with surface refraction [4]–[6]. But as a result of a full angle SAFT, echoes of the desired structures are rever- berated from the specimen’s surface and mask target echoes of illumination angles in the contrary illumination half-plane. So it is extremely difficult to identify the inner structures of the specimen and separate them from ghosts. Accordingly, an adequate suppression or at least a sufficient mitigation of multiples is important. Most of the existing approaches deal with an ideal acoustic wave propagation model in layered media [7]. In this paper, predictive deconvolution (PD) is used for dereverberation purposes. It is a common method for the removal of multiples in seismic exploration [8]–[10] and has also been applied to ultrasonic evaluation [11]. Additionally, a subspace based orthogonal projection is utilized to mitigate the signal echoes of the contour of the specimen, which mask near- surface structural information. This method is predicated on a principal component analysis (PCA) and is used in ultrasonic testing mostly for clutter suppression caused by reverberant layers [12]. The paper is composed as follows. Sec. II describes a specimen representing a model of a turbine blade. The applied measure- ment principle based on the adapted SAFT and successive signal data processing by means of subspace analysis and projection as well as subsequent predictive deconvolution are 150 978-1-4673-5686-2/13/$31.00 ©2013 IEEE 2013 Joint UFFC, EFTF and PFM Symposium 10.1109/ULTSYM.2013.0039