Reliability-based selection of manufacturing tolerance of journal bearings by means of adaptive kriging metamodel

Abstract

Uncertainties are present in any engineering system, from the modeling and design up to the manufacturing and performance. The performance of a system based on nominal design parameters may significantly differ from that of the final product. These discrepancias are related to uncertainties associated with the manufacturing process and material properties, which have been traditionally mitigated with strict tolerance specifications and process control parameters, respectively. Traditionally, tolerancing specification is based on pure geometrical analysis trying to minimize undesired effects in the component assembly. However, the objective of this work is to identify the relationship between the manufacturing tolerance of a component and the expected variations on its mechanical performance. To the best of the authors knowledge, such relationship has not been fully addressed in the literature. A framework is proposed where the manufacturing tolerance is described through probability density functions while its effect on performance is addressed via stochastic simulations. The procedure underlies the adoption of a surrogate model under local and global training (based on adaptive Kriging interpolation) to predict the probability to exceed a certain performance. The proposed framework is illustrated and validated studying a tilting pad journal bearing in terms of minimum and maximum credible values for its dynamic coefficients. Results show a significant saving in terms of computational time, making this framework attractive to perform manufacturing tolerances selection.