Masum Billah

-Constructed a measurement platform for condition monitoring of induction machines with industrial drives. -Extracted time and frequency domain features using FFT, STFT, Wavelet, and EMD from stator current and vibration signals. -Developed high-performance ML classifiers using Python and Scikit-learn, achieving 100% accuracy for condition monitoring of induction machines. -Implemented a Conditional Generative Adversarial Network (CGAN) using Python and PyTorch to generate synthetic data for condition monitoring. -Developed a novel synthetic feature augmentation using error estimation and interpolation techniques, improving XGBoost classifier accuracy from 78% to 96%. -Developed a novel data augmentation method using probabilistic regression models and ensemble learning techniques, improving ML classifier accuracy (KNN: 70% → 98%, SVM: 64% → 91%, DT: 60% → 99%). Developed a 1D CNN classifier using Python and TensorFlow, achieving 100% accuracy for condition monitoring of induction machines. -Designed a surrogate model-based calibration of the FE model of induction machines using MATLAB. -Proficient in Microsoft Office (Word, PowerPoint, Excel), MS Visio, and LaTeX for documentation tasks. -Supervised Master’s thesis students in AI/ML research.

- Developed an ultra-fast iron losses computation technique from a static finite element field solution in FCSMEK - This method (0.5s) is 70 times faster than time-stepping method (35s) - Total iron loss difference between developed method and time-stepping method is 8.94%