Detection of Faulty Rolling Element Bearings in Inverter-Fed Induction Machines via Examination of the Stator Current

Since its introduction towards the end of the 19th century, the induction machine (IM) has become an important drive type in the industrial sector, particularly due to its robustness and comparatively straightforward construction. Moreover, as a result of the continuous progress in the field of power electronics over the past decades, this machine has gained increasing significance and has therefore almost entirely replaced the previously preferred DC-machine in variable-speed drives, especially in the railway-sector and passenger transport. This trend has also become apparent in recent years in the field of electromobility, where a combination of permanent magnet synchronous motors and the more cost-efficient IM is now being frequently employed. Although the aforementioned IM is being considered as robust, up to 40% of its failures are attributed to defective rolling element bearings in the rotor. In order to prevent a decrease of its efficiency throughout the entire intended service life and to prevent a complete breakdown, special attention has to be paid to the condition monitoring of rolling element bearings, with the aim of reliably detecting an emerging damage at an early stage. In the scope of this dissertation, a method is investigated, which aims to detect bearing defects in grid- and inverter-fed IMs through the analysis of the stator current. One major advantage of this method, when being compared to other well-established condition monitoring techniques, which incorporate acceleration sensors, is its robustness against vibrations, which are not caused by the machine¿s bearings themselves. Hence, avoiding unwanted misinterpretations when the acceleration sensor picks up vibration, that is caused by the monitored machine¿s surroundings. Other drawbacks of using traditional vibration sensors are their rather high expenses and especially electrical machines lack the possibility of providing flat mounting surfaces in close proximity of the bearings¿ housing. To begin, the system comprising the machine (electrical part) and the rolling element bearing (mechanical part) gets considered in its entirety, whereby the first part of this work is dedicated to the design and simulation of a novel bearing model with five translational degrees of freedom and various defect shapes and locations. The so derived findings serve as a basis for a deeper understanding of how vibrations occurring in a defective bearing yield an additional angle-dependent frictional torque. The second part of this work addresses the question of how a periodic load torque (e.g. from a defective bearing) that gets applied to the shaft of an IM affects the stator current waveform. The derived mathematical model of the relationship between the stator current and the applied load torque, as well the experimental validation indicate, that the stator current waveform resembles the characteristics of a phase-modulated signal, in the presence of a periodic load torque. As a consequence, the amplitude spectrum of the raw stator current signal shows the occurrence of additional peaks. The insights obtained in the scope of the second part regarding the phase-modulated nature of the stator current in the presence of an (additional) periodic load torque serve as the basis for the development of a suitable algorithm, which will be addressed in the third part. With the help of this algorithm, the automatic detection of defective rolling element bearings such as single and multiple fractures, false brinelling on the outer and/or inner ring, pitting on both raceways, moisture-induced corrosion, damage to the rolling elements themselves, etc.- gets accomplished. The subsequent evaluation of the algorithm is then being carried out using different types of rolling element bearings, which exhibit the aforementioned defects. These bearings are installed in four different mac