Abstract: In modern industrial settings, more and more new products are created in the form of computer images via Computer Aided Design (CAD). On the other hand, CAD files are sometimes created to catalog some well known products whose technical drawings may or may not exist. In either case, the geometric features of the products can be measured with a Coordinate Measuring Machine (CMM). In this paper, we discuss how to judge the geometric quality of a product when its CMM data and CAD file are available. We first introduce a probabilistic model to represent the relationship between the CMM data and the CAD file. Then we define a measure of geometric quality and use it to connect the geometric quality inspection problem to a statistical hypothesis testing problem. The measure we propose is easy to calculate and sensitive in detecting manufacture errors. An application to automobile part inspection is included.
Key words and phrases: Computer assisted design, coordinate measuring machine, hypothesis testing, probabilistic model.