Genetic approach for automatic detection of form deviations of geometrical features for effective measurement strategy

Shortening the planning times and reducing manufacturing costs are strategic specifications, to which the entire product developing process must subordinate itself. The consequent use of available metrological processes and components can ensure the long-term stability of production processes. With a worker-oriented integral effort, the process chain from the CAD system to an integrated inspection planning to the feature-based measuring on form tester and coordinate measuring machines can form a quality loop. Automatic production of a complete feature list directly from the CAD system and the uniform inspection planning under specification of feature-oriented measurement strategies creates the premise for standardized methodologies and international comparability of results of measurement. n-comparability of measurement results is often the cause of a lack of trust between customer and supplier, unnecessary discussion and sometimes even litigation. Coordinate measurement offers the operator numerous opportunities for influencing the result of measurement in almost any way. This especially concerns the specifications for performing and evaluating the measurement by considering the form deviation of the real workpiece prospective for the manufacturing process. A measurement strategy is defined which ensures the undershooting of a target uncertainty [Koordinatenmesstechnik, Carl Hanser, München, 1999; Werkergerechte und prozesskettenorientierte Messtechnik, in: Koordinatenmesstechnik, VDI-Verlag, Düsseldorf, 2001, p. 37]. ork presents a method to detect form deviations of standard geometrical features (line, circle, plane, cylinder, cone and sphere) of a manufactured part using genetic algorithm. Deviations of a geometrical element are detected in terms of the combination of several basic deviation types (waviness, random deviation, offset, peak, etc.) with their parameter values. Genetic algorithm arrives at the optimal values of these basic deviation types which reproduce the profile very close to the measured one. This information can be stored in knowledge-based system and the shape of the part can be reproduced to adopt a suitable measurement strategy. An interactive software assistance with graphical view is developed to effectively handle the detection procedure. Effectiveness of the detection software is illustrated with a synthetic and a measured profile of the crank shaft.