This paper proposes a new global optimization approach for electric ma
chine design problems. The algorithm uses an error function between th
e actual objective function, at a given computational step, and its li
mit value to search for the global minimum. The error function provide
s a good indication of how far or close the objective function approac
hes its ultimate solution. This method is considered to be self adapti
ve in a sense that the higher the error is, the larger the step size o
f the variables will be. This feature helps the search to escape local
solutions and to reach eventually the global solution. The proposed m
ethod was tested on various typical multimodal functions and particula
rly applied to induction motor design optimization problems. This algo
rithm is very simple and easy to be implemented, yet powerful and effe
ctive as indicated by the test results.