An active learning hybrid reliability method for positioning accuracy of industrial robots

Abstract

Popsitioning accuracy is an important index for evaluating the capacity of industrial robots. As a mechanism with multi-degree of freedom, the uncertainties of industrial robots are diverse and analyzing the positioning accuracy reliability is time consuming. To improve computation efficiency, a new active learning method based on Kriging model is proposed for hybrid reliability analysis of positioning accuracy with random and interval variables. In this study, the updated samples were selected through U learning function in the vicinity of limit-state function. A new stopping criterion based on expected risk function was exploited to judge whether the accuracy of Kriging model is enough. Two numerical examples and one engineering example were provided to verify the efficiency and accuracy of the proposed method. The results indicate that the proposed method is accurate and efficient.