Executive Summary : | Electrical machines, especially induction machines,are widely applied in industries and form essential parts of industrial systems.They are found in various applications in marine propulsions,aerospace actuators, process industries and traction system in electric vehicles/hybrid electric vehicles.Despite their rugged construction, they are subjected to fault due to aging, severe operating conditions, and harsh environments. Diagnosis and severity evaluation of fault have become an important topic to help increase the availability and reliability of electrical machines and the associated systems. Downtimes and associated economic loss due to failures is high, and so the condition monitoring systems are necessary . However, the current research has the limitation due to the effects of ambiguous factors affecting early diagnosis. In this project, effort will be to combine two complementary schemes (Model-based and data-driven) to form a hybrid approach for early fault detection and isolation.
Model-based technique is one of the approaches to address the challenge of early fault detection under the influence of the ambiguous factors. Its strength lies in the ability to incorporate the electrical fault parameters in the model and hence the analysis of fault effects. It is based on the multi-physics nature of electrical machines, and uses different methods based on multiple coupled circuits, dq based and sequence component models.One approach will not be able to take into account these effects and also the supply artifacts.Therefore, a multiple model approach, which takes into account the non-linearity of the system and supply artifacts, will be developed.
Another complementary area of research is based on data-driven approach for diagnosis of electrical machines, which do not directly take into account the machine parameters, but it depends on the data available at the terminals of the machine for diagnosis and identification. Accurate representation of the condition is critical in this method. Recently proposed DTCWTs (Dual Tree Complex Wavelet Transform) would be employed, which are nearly shift insensitive and method based on them are capable of detecting early faults, and represent the condition accurately, compared to traditional discrete wavelet methods.
In the proposed novel approach, the complementing schemes have to be combined into one single hybrid scheme, in which, one can take the advantage of each to form better and more reliable system. In this present project, other than voltage imbalances, certain ambiguous conditions such as power quality events (PQ events like sag, swell, interharmonics, notch, spike etc..), which has not been investigated before, in presence of faulty machine conditions, will be taken into account. This is key for improving the robustness of the proposed hybrid scheme.
The validation of the model developed and the approach would be done using a hardware-in-loop system, encompassing a prototype test bed. |