This thesis has two objectives: the first one is the introduction and validation of an analysis method of stator currents for the diagnosis of induction machines connected to the grid and operating in the transient state. The procedure is based on the use of notch filters in combination with the Wigner-Ville distribution, with emphasis on their application for detection of incipient faults.

The second objective is to replicate as faithfully as possible the natural process of a bar breakage in the rotor of an induction motor. For this task, a fatigue test was conceived to cause such failure, a test bed was assembled, and the necessary sensors and a data collection system were mounted for performing the trial in an automated manner.

In order to verify the novelty of the method and substantiate the theoretical environment in which it is developed, a state of the art is presented, which includes both cage induction motors and certain experimental models, their faults and the methods of signal analysis, especially those suitable for the study of transient operation.

Next, the proposed diagnosis technique based on the Wigner-Ville distribution is presented in stages: its development from the Discrete Wavelet Transform, its evolution through the application of IIR (Infinite Impulse Response) filters, and its improvement by FIR (Finite Impulse Response) filters. This technique provides a seamless time-frequency diagram able to reflect a greater amount of information compared with other procedures, which facilitates the identification of a defect as the evolution in two dimensions of the components of failure during the transient of the machine, in both low frequency, traditionally used for this purpose, and high frequency, introduced here as a novelty, is followed. A quantization procedure based on the calculation of the instantaneous frequency is also presented.

The validation of the proposed method is performed in two stages, firstly simulated stator currents of three types of induction motors, in different states and operating conditions are obtained. Four simulation methods are used to produce these currents, two analytical ones, devised by the research group and that have led to a development that is also presented in this thesis, and two others based on finite elements.

This validation is extended subsequently by the analysis of stator currents from real squirrel cage motors, as the main defects of this type of machine are studied. The experimental part continues with an analysis of the results from the bar breakage fatigue test, which manages to reproduce in the most realistic way the evolution of this fault. Various diagnostic methods based on different techniques of signal analysis are used in the study of the corresponding stator currents, in order to verify its ability to follow the progression of the defect. This trial has been extended to the design and construction of an inverted motor, that is, a winding rotor motor with squirrel cage mounted on the stator, and the data collection system necessary for monitoring it, which provides additional information regarding the process of bar breakage.

The main conclusions of this thesis are:

- The validation of a diagnosis method based on the Wigner-Ville distribution that studies the transient of induction motors connected directly to the grid and can detect all kinds of faults from this type of equipment.

- The validation of a method for quantifying such defects based on the calculation of the instantaneous frequency.

- The reproduction in the most realistic manner of a rotor bar breakage, and the compilation of a database of 82,265 operation cycles during this fatigue test.