Electric Power Components and Systems, 34:259–269, 2006 Copyright © Taylor & Francis Group, LLC ISSN: 1532-5008 print/1532-5016 online DOI: 10.1080/15325000500240862 A Thermal Model for a TEFC Induction Motor—Development and Sensitivity Analysis N. BENAMROUCHE M. BOUHERAOUA S. HADDAD Faculté de Génie Electrique et Informatique Université de Tizi Ouzou, Algeria A great deal has been written on the subject of motor heating, much of which are testimony to the complexity of the problem. In this article, the authors consider a thermal model based on the lumped parameter method to give both steady-state and transient temperatures in different parts of a 2.2 KW TEFC induction motor. The work then is extended to study the sensitivity of this model to heat transfer coefficients and to the distribution of iron losses. The stray load losses and the frame-ambient film coefficient were found to be the key parameters that influence most the predicted temperatures. Keywords induction motors, heating, thermal modeling 1. Introduction Due to their low cost, simplicity, and high reliability induction motors by far are the most common motors used by industry. The maximum rating of a machine for a given frame size is governed predominantly by temperature rise, which, when reaching its limits, will produce some undesirable effects including loss of dialectical property of the insulating material, mechanical distortion, and fatigue, to name a few. To ensure a satisfactory lifespan for the machine, temperature rise, therefore, must be limited to safe values. Several thermal models have been reported in literature. In general, they can be classified into two groups. The first model is based on those that use numerical techniques such as finite difference or finite element methods [1, 2, 3]. These methods give accurate results for the conduction mode, however, the modeling of convection and radiation modes is difficult; moreover, it may be difficult to account for any changes in the machine design parameters. The second group of models adopted and further developed by many authors are referred to as lumped parameter models, and are based on dividing the machine into several elements, each of which is identified as a node having a thermal capacitance and a heat source, and then interconnected through thermal conductances forming an equivalent thermal network. Using this approach, approximate thermal networks have been used to represent the whole machine [4, 5]. Others introduced very drastic assumptions to alleviate the complexity of modeling the actual bulk conduction and convection heat Manuscript received in final form on 6 June 2005. Address correspondence to Nacereddine Benamrouche, Aomar Gare, Wilaya de Bouira, 10270, Algeria. E-mail: benamrouchen@yahoo.com 259