TY - GEN

T1 - Development of a compact low coherence interferometer based on GPGPU for fast microscopic surface measurement on turbine blades

AU - Li, Yinan

AU - Kästner, Markus

AU - Reithmeier, Eduard

PY - 2015/6/22

Y1 - 2015/6/22

N2 - Vertical scanning interferometry (VSI) techniques are widely used to profile microscopic surface structures of industrial products. This paper introduces a high-precision fast optical measurement system with an optimized small sensor head for the measurement of precision surfaces on a turbine blade or blisks (blade integrated discs). The non-contact measurement system is based on a low coherence interferometer (LCI), which is capable of fast profiling of 3D sample surface with a nanometer resolution and has a larger measurement range compared to conventional microscopes. This results in a large amount of sampled data and a high computational time for the evaluation of the data. For this reason, the used evaluation algorithm in this paper is accelerated by the Compute Unified Device Architecture (CUDA) technology, which allows parallel evaluation of the data stack on independent cores of a General Purpose Graphics Processing Unit (GPGPU). As a result, the GPU-based optimized algorithm is compared with the original CPU-based single-threaded algorithm to show the approximate 60x speedup of computing the Hilbert Transformation, which is used to find the depth position in the correlogram of each pixel of the sampled data. The main advantage of the GPU computing for the evaluation algorithm of the LCI is that it can reduce the time-consuming data evaluation process and further accelerates the whole measurement.

AB - Vertical scanning interferometry (VSI) techniques are widely used to profile microscopic surface structures of industrial products. This paper introduces a high-precision fast optical measurement system with an optimized small sensor head for the measurement of precision surfaces on a turbine blade or blisks (blade integrated discs). The non-contact measurement system is based on a low coherence interferometer (LCI), which is capable of fast profiling of 3D sample surface with a nanometer resolution and has a larger measurement range compared to conventional microscopes. This results in a large amount of sampled data and a high computational time for the evaluation of the data. For this reason, the used evaluation algorithm in this paper is accelerated by the Compute Unified Device Architecture (CUDA) technology, which allows parallel evaluation of the data stack on independent cores of a General Purpose Graphics Processing Unit (GPGPU). As a result, the GPU-based optimized algorithm is compared with the original CPU-based single-threaded algorithm to show the approximate 60x speedup of computing the Hilbert Transformation, which is used to find the depth position in the correlogram of each pixel of the sampled data. The main advantage of the GPU computing for the evaluation algorithm of the LCI is that it can reduce the time-consuming data evaluation process and further accelerates the whole measurement.

KW - CUDA

KW - Interferometer

KW - Low coherence interferometry

KW - Non-contact measurement system

UR - http://www.scopus.com/inward/record.url?scp=84954048112&partnerID=8YFLogxK

U2 - 10.1117/12.2184749

DO - 10.1117/12.2184749

M3 - Conference contribution

AN - SCOPUS:84954048112

T3 - Proceedings of SPIE - The International Society for Optical Engineering

BT - Optical Measurement Systems for Industrial Inspection IX

A2 - Albertazzi G., Armando

A2 - Lehmann, Peter

A2 - Osten, Wolfgang

PB - SPIE

T2 - Optical Measurement Systems for Industrial Inspection IX

Y2 - 22 June 2015 through 25 June 2015

ER -