The paper describes a model-based statistical methodology to design adaptive inspection plans for the geometric control of mechanical parts with Coordinate Measuring Machines (CMM). The inspection is adaptive because the design and measurement phases are not separate in time, as they usually are. Rather, they are carried out in a combined way: first designing the next measurement location, then measuring at that location and so on. This strategy is most informative as it allows for the exploitation of all of the currently available measurements. The next measurement point is selected by using predictions and prediction uncertainty of geometric deviations provided by non-parametric statistical models, known as kriging models. Based on stationary Gaussian stochastic processes, their merit is the ability to vary flexibly at each added point. The methodology is demonstrated in an illustrative case study, then its performance is compared to that of two statistical non adaptive plans, and two deterministic adaptive plans proposed in the literature. In each comparison kriging-based plans have proved to be superior in terms of the accuracy of the predicted geometric error and the inspection cost. The method is sufficiently general to enable technology transfer to different metrological sectors. (c) 2012 Elsevier Inc. All rights reserved.
Adaptive inspection in coordinate metrology based on kriging models
ROMANO, DANIELE
2013-01-01
Abstract
The paper describes a model-based statistical methodology to design adaptive inspection plans for the geometric control of mechanical parts with Coordinate Measuring Machines (CMM). The inspection is adaptive because the design and measurement phases are not separate in time, as they usually are. Rather, they are carried out in a combined way: first designing the next measurement location, then measuring at that location and so on. This strategy is most informative as it allows for the exploitation of all of the currently available measurements. The next measurement point is selected by using predictions and prediction uncertainty of geometric deviations provided by non-parametric statistical models, known as kriging models. Based on stationary Gaussian stochastic processes, their merit is the ability to vary flexibly at each added point. The methodology is demonstrated in an illustrative case study, then its performance is compared to that of two statistical non adaptive plans, and two deterministic adaptive plans proposed in the literature. In each comparison kriging-based plans have proved to be superior in terms of the accuracy of the predicted geometric error and the inspection cost. The method is sufficiently general to enable technology transfer to different metrological sectors. (c) 2012 Elsevier Inc. All rights reserved.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
