Detect and visualize non-uniform yarn orientations on preformed CFRP parts using automatic scanning and image processing

Abstract

This article presents a methodology to measure and visualize the 3D yarn orientation distribution on the surface of preformed carbon fiber reinforced polymer (CFRP) parts. An image acquisition system composing of a camera and a 6-axis robot was established to scan and capture images of a series sampling points on surface of the CFRP part. The acquired images were analyzed by an image processing algorithm based on Canny edge detector, Sobel operator and histogram analysis, to quickly determine yarn orientations at each sampling point. This method was proved to achieve <4.43° and 12.80° of average error in yarn angle and yarn orientation measurement, respectively. The measured yarn orientations and angles were visualized as color-coded maps to display the 3D yarn orientation and angle distribution on the preformed part. 2D interpolation was added to smoothen the color-coded map and accurately approximate yarn orientations and angles outside the sampling points. Compared to the existing yarn orientation/angle measurement methods based on image processing, the presented methods had following advantages: (1) successful detection on both preforms and cured parts with stable accuracy; (2) high measurement speed; (3) accurate images acquisition on severely curved surfaces; (4) economical measurement cost without requirement for customized sensors.