Efficient tolerance analysis by using surrogate models

Producing low cost products with high quality has become one of the most important goals of the production in today's competitive world. Decreasing production costs without increasing quality loss is possible through working in a multi disciplinary manner. As long as the manufacturing tolerances are controlled, the production costs can efficiently be reduced. Continuing growth of computing power and speed increases the complexity of assembly models and tolerance details, thereby the tolerance analysis becomes difficult. In the last decades, inexpensive yet reliable models or simulation techniques have become available in place of overwhelming traditional computer analysis codes/techniques. Surrogate models, which are widely used in many research fields, are usually used in place of overwhelming and computationally expensive traditional methods, due to their low computational cost and reliable solutions. In this study, the main objective is to show that surrogate models are suitable for tolerance analysis. Traditional torelance analysis methods such as WC, RSS and MCS along with response surface model and Kriging surrogate model based tolerance analysis methods are applied on a two dimensional nonlinear example problem and three dimensional example problem which is more complicated than the two-dimensional problem. It is observed that the surrogate based tolerance analysis methods are more efficient than the traditional tolerance analysis methods. It is also observed that Kriging based tolerance analysis is more accurate than response surface based tolerance analysis.