Orynbek, Kuanysh2022-06-302022-06-302022-04Orynbek, K. (2022). MOTOR/GENERATOR FAULT PROGNOSIS USING VIBRATION SIGNATURE AND FORECASTING TECHNIQUES (Unpublished master's thesis). Nazarbayev University, Nur-Sultan, Kazakhstanhttp://nur.nu.edu.kz/handle/123456789/6354Over the last few years, the industrial dependency to operate induction motors and generators has been significantly increased. In this instance, it is necessary to monitor the state of induction (asynchronous) machines, as the motor/generators will face with overloading, under-voltage, overvoltage, or even catastrophic failures over the course of their operation. To address this important concern, a fault forecasting architecture using machine learning techniques is studied and developed over the motor/generator vibration signature in this thesis. Initially, mathematical modeling is provided to find the normal and abnormal parameters, and also explore the vibrational frequencies along with the analytical analysis for motor/generators. The Föppl/Jeffcot’s rotor modeling system is employed for rotor vibration modeling; however, the transformer core and winding vibration model was considered as a basic theory for the stator core and winding of the motor/generator in an analytical approach. To emulate a faulty condition over an induction motor in this thesis, an experimental setup is designed and developed. The voltage excitation condition for induction motor along with single phasing are considered to be the fault types and examined practically in the laboratory to conduct experiments and collect vibrational data. Afterwards, 1D Convolutional Neural Network (CNN) model is constructed for accurately detecting faults. The MSE of voltage excitation prediction was obtained as 0.000426, whereas the highest fault detection accuracy of single phasing reveled to be 99.58%.enAttribution-NonCommercial-ShareAlike 3.0 United StatesType of access: Open AccessResearch Subject Categories::TECHNOLOGYrotor vibration modelingartificial neural networksANNconvolutional neural networksCNNMaster's thesis