Sensing for inspection involves making measurements of an object or process and comparing the results to a desired specification. There are unavoidable errors involved in the processes generating the phenomena being measured. In manufacturing, costs are often inversely related to the precision with which this variability must be controlled. As a result, minimally sufficient tolerances are often associated with specifications. Process engineering in part involves the choice of a manufacturing plan which can meet these specifications with the least expense. Inspection operations insure that the specifications have in fact been achieved.
The importance of quantifying tolerance in the specification, design, manufacturing and inspection process is obvious. Unfortunately, adequate representations of tolerance do not exist which permit dialog between these various aspects of the manufacturing process. This lack is particularly acute in systems which tightly integrate all of the aspects of prototyping. Tolerance specifications, whatever their form, must have the same meaning for all steps in the process. Maximum acceptable deviations must be optimized across all steps so that productivity is not adversely impacted by requirements for unneeded precision; information about the tolerance required and the manufacturing processes also make it possible to focus sensing on areas of higher systematic or statistical concern. We propose to use the tolerance specification in conjunction with knowledge of the manufacturing process plans to determine more optimized sensing strategies.
While published standards such as ANSI-Y14.5M [3] exist for describing manufacturing tolerances, they are not yet sufficiently complete and unambiguous to precisely specify the information needed for automated inspection. Furthermore, these standards provide no help in determining how to optimize the inspection process so as to increase the confidence that out of tolerance features are discovered with a reasonable expenditure of resources. In this paper, we give preliminary examples of an approach to this problem based on a detailed understanding of the manufacturing processes involved in the production of machined parts.