The analysis and detection of faults in three phase induction machines using finite element techniques.

This thesis reports on the use of commercially available finite element software packages for modelling three phase squirrel cage induction machines with the fault condition of broken rotor bars. Where possible, results obtained from the finite element simulations arc compared directly with results obtained from a laboratory test rig machine having the same fault conditions. The purpose of the work was to enhance present Condition Monitoring techniques. It is accepted that the best way to identify rotor faults on-line (i.e. without affecting machine operation) is to analyse the frequency content of the stator currents feeding the machine. If there are rotor faults present then there will be frequency components of current at (l+2s)fs and (l+2s)fs, where s is the machine slip and fs is the supply frequency. It is normally the lower sideband component, at (l-2s)fs that is used to signify the presence of rotor faults. However, some confusion still exists as to how these components of current are affected by machine parameters, fault severity, and machine design. This work attempts to clarify these points. The development of the models are described in detail, firstly using a fixed mesh finite element package, and later using a time-stepping software package which was designed specifically for modelling rotating electrical machines. The first model, using the fixed mesh package, required the development of an external software program to control the simulation so that rotation of the machine could be modelled and therefore stator current data could be extracted from the simulation. The initial results from this simulation showed significant error in comparison to those obtained from the test rig machine. Several iterative improvements were made to the model, including accounting for the rotor speed ripple which is present with bar breaks, and accounting for the rotor skew. These improved the results obtained, but the type of simulation developed still had considerable limitations, so work progressed to using the time-stepping software. The time-stepping finite element package is relatively simple to use as stator current data is calculated inherently in the program, hence no external software was required to analyse the data. More importantly, the time-stepping solver models the machine with greater accuracy than the fixed mesh solver, and includes an accurate representation of the effects of non-linear magnetic characteristics in the steel regions. The results obtained from using this package show reasonable correlation between simulated and real data. However, there are still limitations to the model accuracy. Instead of attempting to iron out the relatively small variations in results obtained between simulation and test rig, more useful tasks for the model were carried out. Firstly, the model was used to determine the individual effects of machine parameters such as rotor inertia, rotor speed ripple, and magnetic saturation on the fault indicating sidebands of current. The results obtained from this work are conclusive and interesting. Next, the model was used to simulate a machine with progressively increasing numbers of broken bars. The magnitudes of the current components of interest were then plotted against fault severity. As well as giving useful results, these tasks illustrated the power of the simulation in being able to change machine parameters and instantly see the effects of the changes. The work then progressed to simulating a machine with rotor faults under transient conditions. The results from these simulations were used in conjunction with a software tool developed by other researchers for analysing transient currents to predict rotor faults. This work showed that the subject of monitoring a machine under transient conditions holds great potential for detecting faults at an early stage in their development. The finite element model is a great asset for this research in that machine parameters, different machine designs, and fault severity can all be changed easily and quickly. The thesis ends with preliminary findings and ideas on how the finite element models may be used to investigate other types of faults which have received little attention from researchers in the past. This work could be the subject of future research. To summarise, the work reported shows that although it is difficult to obtain perfect correlation in terms of magnitudes of currents at different frequencies between the models and the laboratory test rig, the models are immensely powerful in being able to show the effects of individual machine parameters on machine behaviour, and the general effects of rotor faults on machine behaviour. The model has also proved to be powerful in being able to model the machine under transient conditions with a good degree of accuracy. Results from this work could be used to directly improve present Condition Monitoring techniques.

Author dc:creator

• Paterson, Neil

Advisor dc:contributor.advisor

• J. Watson and D. Dorrell

• Induction motor condition monitoring
• Broken rotor bars
• Finite element modelling
• Stator current analysis
• Rotor fault diagnosis
• Transient machine behaviour

Language dc:language

en

Identifier

oai:rgu-repository.worktribe.com:2807521
https://doi.org/10.48526/rgu-wt-2807521

OAI identifier oai:identifier

oai:rgu-repository.worktribe.com:2807521