Computational Metrology

Coordinate measuring systems (CMSs) assess length-based characteristics of mechanical parts by measuring points on the part surface and analyzing the point data. Data analysis software can contribute significantly to the total measurement error of a CMS. Factors affecting software performance include the choice of analysis method, the quality of the software, and characteristics of the specific measurement task. By computational metrology we mean the study of how data analysis factors affect the measurement uncertainty of a CMS. This paper describes research at NIST on computational metrology. Metrological questions fall into two categories: (1) the proper choice of data analysis objective for a particular application and (2) the performance of the implementation. Our research goal is to develop the basis for a national standard in the U.S. for CMS software performance evaluation. We are implementing a Special Test service to be offered through the NIST Measurement Services Program. The service is based on a black-box model of software, in which the internal structure of the software and the choice of solution methods are assumed to be unknown. The model identifies a number of error sources for data analysis software. We are designing test methods for identifying the various components of the model and how those components relate to measurement uncertainty in inspection applications. INTRODUCTION Data analysis software has become increasingly important in modern dimensional measurement systems. This is particularly true of coordinate measurement systems (CMSs) such as vision systems, theodolites, photogrammetry, and coordinate measuring machines. Software computations to convert raw data to reported results can be a major source of error in a measurement system. Yet there are no standards or 1 This report has been prepared by U.S. Government employees as part of their official duties and therefore is not subject to copyright.