Induction motors optimal design can involve many variables and objectives, and generally requires to make several trade-offs, especially when including the audible electromagnetic noise criterion beyond the usual performance criteria. Multiobjective optimization techniques based on Pareto optimality are useful to help us finding the most interesting solutions and decide which one(s) to adopt. However, it is not always easy to analyse the Pareto-optimal solutions obtained with such methods, especially when treating more than three objectives, and Pareto fronts may contain more data than we might think. This paper briefly describes an analytical model of the variable-speed squirrel-cage induction machine which computes both its performances and sound power level of electromagnetic origin. The model is then coupled to the Non-dominated Sorting Genetic Algorithm (NSGA-II) in order to perform global optimization with respect to several objectives (e.g. noise level, efficiency and material cost). Finally, an optimization problem is solved and analysed, and some useful visualization tools of the Pareto optimal solutions and their characteristics are presented.