Evaluation of computed tomography data from fibre reinforced polymers to determine fibre length distribution

Sub-lm computed tomography (sub-lm-CT) was used to determine the fibre orientation and fibre length distribution in long glass fibre filled polypropylene. For data evaluation two different concepts based on the application of a sequence of different filters were applied. The first concept is based on segmentation by binarisation using a global threshold, followed by a detailed analysis of regions where fibres are touching. The second concept is based on analysis of the original gray value image. For each voxel the local fibre orientation is determined by calculating the Hessian Matrix and analysing its Eigen values. The effectivity of the two data analysis concepts in determining orientation and length was investigated. For this purpose the algorithms were applied to specimens with four different levels of fibre content: 1, 5, 10 and 30% by weight. To quantify the level of error in fibre determination, a minimum and average probability for correct fibre determination were estimated. The results show a strong dependence of the level of error on the fibre content. Whilst the determination of fibre orientation is not significantly affected, determination of fibre length distribution is significantly influenced by fibre content. For samples with fibre content above 5%, concept 1 does not produce correct representations of all fibres. In particular, problems arise if the fibres are touching. Concept 2 delivers much better results and represents most of the fibres correctly even at higher fibre content levels and for touching fibres. This was proven by using artificial CT-data sets generated by CT-simulation and by systematic comparisons. A practical application of the CT-evaluation pipeline is presented for glass fibre reinforced rings produced by injection-moulding and extrusion. For both samples the orientation tensors are calculated and the orientations of the fibres are visualized in three dimensions by colour coding.