The condition monitoring of an electric motor is critical to its stable and reliable operation. The consequence of a fault is the variation in output performance and changes in the input electrical signature. One of the most common methods of fault detection is by analyzing the input current signature but this often requires expensive sensors. The other symptom of a faulty machine is when it runs "hot" due to an increase in losses generated. Temperature changes could be a useful indicator of the underlying fault. In this paper, a method of condition monitoring using temperature sensors embedded in an electric motor is presented. The components of electromagnetic losses generated are determined indirectly from temperature measurement in real time for a continuous drive cycle simulating normal and faulty operations. This is achieved through the inversion of the thermal model of the machine coupled with the application of a recursive parameter estimation method. Using temperature measurement for condition monitoring is advantageous since temperature sensors are low cost and already embedded in the machine for thermal protection. However, application of the method to a faulty motor and evaluation of its effectiveness in fault detection will be reported in a future work.