<p> </p>
<p>A novel approach to the collection of fault related data associated with induction</p>
<p>motor drives is presented. The stator to rotor magnetic flux of an induction motor is</p>
<p>monitored by a number of strategically positioned search coils, each wound around a</p>
<p>single stator pole. The data collected is in the form of time records of the induced</p>
<p>voltage in the coils and is subsequently used to form the data base for a fault detection and diagnosis strategy.</p>
<p>Voltage waveforms obtained from a single coil and from two coils connected in series</p>
<p>are obtained whilst the system is subjected to a range of applied electro-mechanical</p>
<p>faults. The applied faults are applied both to the mechanical load and to the induction</p>
<p>motor itself. A comparison is made of the efficacy of using two search coils</p>
<p>compared to employing a single coil for fault detection.</p>
<p>The fault related data is collected under both steady-state and accelerating running</p>
<p>conditions. Strictly the acceleration period waveforms are non-stationary, however,</p>
<p>since the time dependant frequency changes are relatively slow, the author applied the</p>
<p>FFT technique to both steady-state derived data and the acceleration period derived</p>
<p>data. Processing is carried out on both the time domain and the corresponding frequency domain data.</p>
<p>The non-stationary nature of the acceleration period records is taken into account and the Wigner-Ville technique is employed to establish a time-frequency-distribution. Amplitude-time-frequency 3-D representations are produced, from which the amplitude versus time activity of a typical acceleration period component frequency is</p>
<p>presented.</p>