Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.387402
Title: Methods for computer aided inspection of geometric tolerances
Author: Carpinetti, Luiz C. R.
ISNI:       0000 0001 3520 7501
Awarding Body: University of Warwick
Current Institution: University of Warwick
Date of Award: 1993
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
This thesis investigates computational methods for assessing tolerance specifications of geometric features in a context of computer aided inspection. It is concerned with checking the sampled features for containment within tolerance zones specified at the design stage, not with explicit shape measurement. The significance of this difference is highlighted when two or more features are to be inspected in combination. The approach adopted is to express the tolerance information as a set of inequality constraints and then to seek efficient methods for determining the feasibility of the set, that is whether all the constraints can be simultaneously satisfied. Roundness inspection is used to introduce all the concepts of the new formulations. By linearisation of the constraints, a standard approximation in roundness measurement, a new algorithm is implemented which provides a “GO-NOGO” result of inspection by checking for feasibility in a highly efficient way. This algorithmic approach is then extended to other inspection situations where naturally linear constraints or valid linearisation occur. Since there are many inspection cases where linearisation is not appropriate, non-linear optimisation techniques are then investigated for their effectiveness in feasibility testing. The inspection of arrays of circular features is used here as a typical test case. Genetic search methods are explored as a possible alternative to formal non-linear programming and guidelines for their efficient use for this problem are proposed. These methods are then compared and contrasted with formal methods, particularly generalised reduced gradient (GRG) and sequential quadratic programming (SQP). The linear algorithm is shown to be the most efficient when it can be used, although all techniques were fast enough for on-line use with modest sized data sets. Currently all the non-linear methods are too expensive for routine use on large data sets. GRG is recommended as having the most favourable combination of good and bad features, but there is some evidence that genetic search might be relatively more efficient for more complex inspection problems.
Supervisor: Not available Sponsor: Conselho Nacional de Desenvolvimento Científico e Tecnológico
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.387402  DOI: Not available
Keywords: QA Mathematics ; TS Manufactures
Share: