CT parameter studies for porous metal samples. Sören R. Lindemann Daimler AG Werk Untertürkheim

Transcription

1 CT parameter studies for porous metal samples Sören R. Lindemann Daimler AG Werk Untertürkheim

2 Where do we stand and what are we looking for? small material samples (high absorption coefficient, low porosity) expectation: porosity changes with increasing distance from sample surface goal: derive quantitative characteristics from X-ray CT volumes to overcome a solely qualitative diagnosis and to generate completely new knowledge obvious characteristics: volume porosity, pore size distribution (both depending on distance to surface) image processing of sectional images from X-ray CT data similar approaches came up in the past years due to increasing availability of high-resolution CT problem: available solutions either very specific or not satisfying (black box solutions) central question: what can be used to evaluate the quality of the CT measurement and how volatile are these due to CT parameter variations? optional: do X-ray CT values match those from e.g. polished sections, synchrotron-based or neutron CT? 2 Sören R. Lindemann Daimler AG

3 cavity portion [%] norm. intensity [%] How are we approaching the problem? step #1: specify a standard sample size and apply the same CT parameters to all threshold step #2: align sectional images parallel to sample surface and calculate gray values gray value distribution of CT volume shows two characteristic and clearly separated maxima step #3: find a binarization approach surface step #4: count pixels in binary image cavity portion and not porosity, because evaluation is based on gray-scale values average minimum distance from sample surface 3 Sören R. Lindemann Daimler AG

4 Which parameters influence the image processing chain? digital CT volume filtering binary image feature extraction segmentation classification X-ray source acceleration voltage tube current beam hardening metal sheets detector acquisition time sensitivity acquisition parameters number of projections (average/skip) virtual volume size (export) voxel outlier part high- and low-pass filters resolution and distance of sectional images color depth choice of threshold universal (for all measurements) global (for all sectional images of one measurement) local (for one sectional image) regional (for one part of one sectional image) adaptive (using information from adjacent pixels) 4 Sören R. Lindemann Daimler AG

5 norm. intensity [%] norm. intensity [%] What are criteria to evaluate the results? material contrast: distance of characteristic gray values FWHM of these maxima criterion for sharpness of characteristic maxima (arbitrary) threshold material peak air peak FWHM threshold sensitivity (plus/minus gray value) criterion for the binarization error still in progress: criterion for edge sharpness or detail resolution in sectional images 5 Sören R. Lindemann Daimler AG

6 norm. intensity [%] mat. contr. What is the effect of varying acceleration voltage? 175kV 160kV 145kV 130kV gray value distribution slightly influenced spreading of gray values higher with increasing voltage voltage [kv] material contrast does not stop increasing 6 Sören R. Lindemann Daimler AG

7 mat. contr. norm. intensity [%] How does varying tube current influence the results? little influence on position of characteristic maxima 120µA 100µA 80µA 60µA current [µa] material contrast increases until 100µA (probably) inner unsharpness visible at currents >100µA Sören R. Lindemann Daimler AG

8 norm. intensity [%] mat. contr. Which impact has beam hardening through copper sheets? 0,0 mm 0,1 mm 0,2 mm 0,5 mm 1,0 mm 1,5 mm determining influence on spreading of histogram decreasing absolute intensity vs. smaller energy spectrum 0,0 0,5 1,0 1,5 thickness of copper [mm] material contrast becomes maximal at 1,0mm copper 8 Sören R. Lindemann Daimler AG

9 What can we learn from this? answer: deriving quantitative characteristics from stacks of X-ray CT sectional images with image processing is possible but challenging no universal recipe and a lot of influences need to be kept in mind challenge here: nobody knows how high the porosity really is and we do not want to check it on an atomistic scale (despite from being able to do so) our approach: apply well-founded parameter study to identify and analyze possible influences repeatability from 10 measurements of same sample (6 months, filament changes, maintenance): minimum=1,04% (σ min =0,08%) and average=2,72% (σ avg =0,10%) material contrast and threshold sensitivity are helpful criteria to evaluate the quality of CT measurements over the shown parameter study another criterion for edge sharpness/detail resolution would help to quantify the quality of sectional images 9 Sören R. Lindemann Daimler AG