The electrical induction machines are part of most production processes (they account for more than 90% of all electrical equipment) due to its simplicity constructive and reliability. Production and economic reasons is necessary to diagnose and detect possible failures that may arise, to avoid unscheduled stops with their consequences. Hence in the last decades, investmenting economic and human resources have been made to research and develop non-invasive techniques able to detect and diagnose incipient failures or malfunctions. Also it is strongly recommended that these techniques can work on line, with the machine running, and indifferent operation regimes.

Traditionally, analysis of signals to detect and diagnose faults in electrical machine has been made in permanent regime. Initially, the measure vibrations of the machine were used. Since several years, current stator measures are also used, based mainly in the frequency analysis by Fourier transform. Recently, new techniques based in advanced signal analysis are being applied to replace the Fourier transform, due to problems, as the effect of dispersion of the signal or commonly known as the effect of "leakage". Also it has begun to use signals captured in the transitional operation, mainly the starting current, for further analysis trought wavelets transform and for diagnosing possible failures. This is a principal work developed for my workgroup.

In this thesis four new methods are developed, based on different techniques and advanced signal detection for diagnosing asymmetries in rotor of electrical machine induction, mainly of the eccentricity also applied to the broken bars and to other types of electromechanical breakdowns. The methods and techniques developed in the thesis are four, two of them for the permanent regime, and the other two for the transient regime.
 
Two of these methods are developed based on the implementation of the Hilbert transform (HT) to one stator current of the machine. The thesis emphasizes the benefits this new technique in permanent regime versus traditional Fourier analysis, such as the elimination of the effect of dispersion, the independence of the power frequency and its validated for any electrical fault in any operation regime (from no load to full load), for example the detection of broken bars at low slip. The method developed in the thesis for the transient regime allows represent the frequency characteristics of faults versus the slip of the machine, expanding the frequencies range used in the diagnosis. In addition to represent versus the slip makes it universal, independent of the load, and valid for the permanent and transient regime.

The other two methods, one for each operating regime, are based on failures patterns of electrical machine, so as to easily locate indicators of failures in them. Thus, for the detection and fault diagnosis indicators are used more than a simple harmonic at a frequency determined as was the case traditionally at risk of inducing errors. Both are based on the extent of the current transitional arrangements in stators or permanent depending on which method is concerned, and after analysis comparing the patterns created using tools such as a mathematical model of the electrical machine.

In addition develops a mathematical model of the machine electric induction, based on an analysis of analytical coupled circuits of electric machinery (MCSA). This model does not have as much precision as a model based on finite element method (FEM) but instead has a lower cost computing. The model is a tool that allows machines to simulate electrical anomalies or failures of a quick and simple. Through this validation techniques can be made as a solution before verify them on an experimental basis. 
The model developed presented as the great new way to calculate the matrix of inductors of the electrical machine to be built. This is calculated by the mathematical technique of convolution that takes place only once stored the matrix of inductors for subsequent use during modelling such electrical machinery. 
It chooses a model circuital compared to models based on finite element model for the circuital is much cheaper and much more computationally efficient because of their temporary savings, in addition to being developed by the author can make changes and improvements when desired, resulting much more adaptable to future work.

This thesis is divided into 7 chapters organized as follows:
 
The first chapter contains a brief introduction, which serves as a starting point of this thesis. 

The second chapter defines the general aim of the thesis and the particular objectives needed to achieve that objective. 

The third chapter provides a historical review at various levels. At the outset are reviewed and made a classification of the failures that occur in the electrical machine. After analyzed the magnitudes that characterize the electrical machines. We review the main techniques and tools used signal, briefly exposing their development. A review of different mathematical models are compared their main features. We review the indicators provided by other authors to diagnose faults. To conclude this chapter realized an experimental critical review of techniques with electrical machines modified to cause failures. 

The fourth chapter is a collection of practical problems that arise when realized the analysis of signals obtained quantities of electrical machines. 

In the fifth chapter is presented the mathematical model used in this thesis. 

In the sixth chapter will exposed four new solutions to detect for diagnosing faults type asymmetries in rotor of the electrical machine. 

To finish in the seventh chapter presents the conclusions to this thesis.