Uncertainties are inherent to the physical world, which has made them a prominent field of research. They manifest themselves in many fields of engineering, one of which is known as tolerance analysis, where engineers attempt to predict system performance affected by the tolerance through analytical models. Ordinarily, discrepancies between analytical and experimental models are overcome by calibration techniques, using experimental data. Acquiring experimental data becomes a monetary expensive endeavor, when performed on assemblies with relative large design parameters. This work presents a methodology to formulate and calibrate analytical models, using Bayesian inference and Gaussian processes. The advantage of this methodology is that it uses a single set of overall performance experimental data, with a more accurate result than conventional methods. Subsidiary to the proposed methodology is a case study, vindicating as well as illustrating its transcendence from the confining extremities delineated by the panoply of current tolerance analysis and calibration techniques.

This content is only available via PDF.
You do not currently have access to this content.