Improving template-based CT data evaluation by integrating CMM reference data into a CAD model-based high fidelity triangle mesh

Abstract

One of the key characteristics of dimensional X-ray computed tomography (CT) is the possibility of complete capturing of all interfaces of the measurement object with its environment (atmosphere and clamping contact points) and itself (in case of certain multi-material conditions and pores). This property distinguishes CT from optical measurement systems (e.g. structured-light scanning), which in general lack the capability to capture the clamping surfaces without additional expenses into additional positioning and referencing measures. Therefore, CT data is in general well suited for data template-based1 measurement data evaluation (in this context also known as nominal-actual comparison) because the whole surface of the measurement object is captured it is straightforeward to identify critical surface regions. In practice (e.g. as described in [2]), a nominal-actual comparison is usually conducted by aligning the measured geometry with the nominal geometry (CAD model) of the measurement object, followed by the calculation of the local distance field. Following this approach, it is not possible to separate measurement errors introduced by the measuring system and production deviations introduced by the workpiece. Therefore, the approach is less suited for a thorough measurment system analysis. Tactile probing is well established in dimensional metrology [3] and tactile coordinate measuring systems (CMS) perform highly accurate measurements on various workpieces [4, 5]. Furthermore, the simulation of the measurement evaluates the task specific measurment uncertainty [6]. Nevertheless, only a limited number of measurment points is captured by a tactile CMS and some features might not be accessible. This contribution demonstrates a method to integrate point cloud CMM (reference measurement) data into the CAD model of a workpiece in order to create a hybrid CAD/CMM model represented by a high fidelity triangle mesh, which can be used for further evaluation [7– 9] of the local measurement uncertainty characteristics of the measurement task as well as for standard nominal-actual comparisons. Additionally, as a necessary pre-processing step, a method to transform a CAD model into a triangle mesh with user-defined sampling point density is presented, such that each triangle edge point lies exactly on the CAD model surface. This represents the prerequisite for correct nominal-actual comparisons with triangle meshes used as nominal geometry.