Pattern Recognition as a Quality Control Tool with Applications to Rubber and Plastics

The use of pattern-recognition principles is advanced as an approach to quality assurance in those numerous cases in which quality resides not in a single attribute but in a complex relationship among a number of variables. A key element in the approach is the identification of critical parameters, or features, which determine quality. Once isolated, these features can then be combined into quality indices or “signatures” capable of differentiating between acceptable or unacceptable product, even in cases in which the individual features seem to bear little relation to quality. Pattern recognition is thus seen as an adjunct to advanced testing techniques, which are effective only to the extent that the information derived therefrom can be related to the factors which determine quality. Algorithms for implementing a pattern-recognition approach to quality are discussed and are shown to be non-prohibitive in either complexity or cost.

How surface geometry features affect the of an elastomer as a seal is cited as an example of the benefits which can be realized from the pattern-recognition approach. It is shown that, in general, surface geometry be must characterized in terms appropriate to the function to be served by the surface, and this fact necessitates a multiple-parameter description of the surface rather than the classical single-parameter description in such terms as AA or RMS roughness. Other applications of pattern-recognition techniques to quality control are discussed.