Data Processing, Modeling, and Error Analysis in Discrete Part Geometric Inspection

In today’s world of modern metrology, it becomes increasingly common to find new technologies which employ non-traditional measuring mechanisms in attempts to provide new advantages to lever over competing measuring instruments. In addition to such new technologies emerging, parts being measured also tend to grow in complexity, presenting new and unique challenges in their measurement. As a result of such advancements in technology and increasing part complexities, it becomes necessary to develop and optimize data processing, data modeling methods and error analysis methods tailored to a given measurement technique or a given part. This dissertation explores these aspects of discrete part metrology and presents solutions for each case considered. A data processing method is developed to allow measurement of any complex part using a precision rotary table and a set of triangulation-based laser line sensors. Machine learning approaches are also explored for a similar measurement problem. The system described is tailored to measure desired discrete parts (gears in this case) and can be considered a custom coordinate measuring system (CMS). An essential step to pushing such custom machines or new measurement technologies into widespread use, is the development of standardized test procedures that enable users to compare different technologies by their performance against traceable standards. Development of such standards often involve designing performance evaluation test procedures that are sensitive to as many known error sources are possible. Therefore, such an approach is adopted for two measuring technologies, namely X-Ray Computed Tomography (XCT) and Stereo-vision Photogrammetry. In each case, several known error sources are characterized, and recommendations are made on ways to capture them. Further, from an automation perspective, it becomes essential to use a data modeling system that can support descriptions of complex parts and their measurement results, as well as accommodate new measurement technologies and tools. To this end, a promising candidate, the Quality Information Framework (QIF) has been identified.