Models in electromagnetism are more and more accurate. In some applications, the gap between the experience and the model comes from the deviation on input data of the model which are not perfectly known. The stochastic approach can be used to quantify the effect of these input data uncertainties on the outputs of the model. In this article, the application of such approach in computational electromagnetics is presented. The four steps development of the model, characterization and modeling of the input data variability, uncertainty quantification, postprocessing (sensitivity analysis) are described and illustrated by an example of electrical machine with uncertain dimensions
