Detecting faults in a propulsion system in time can make great savings in maintenance and in avoiding succession errors. The object of this thesis is to design a fault detection system for three typical errors occurring with a squirrel cage induction motor. The errors in question are phase-shifted motors, motors with broken shrinkage fits and motors with a locked rotor. Appropriate methods were initially evaluated theoretically in open loop. Simulations extended the theory to include closed loop. A real-time simulation environment was then used to verify the method of choice. Finally the algorithm was verified in an experimental set-up. The resulting algorithm of the study is based on the estimation error of the stator current. Lab verifications show that with slight modifications the algorithm should work for all three faults in question.