Technical Portfolio
Every project here is public, and every claim links to the code. Where a result is weak or a project is early-stage, it says so.
Fault Diagnosis in PMSM Motors — CNN on Wavelet Scalograms # github.com/mulhamfetna/pmsm-fault-diagnosis-cnn-scalogram · Python · MIT
Detecting inter-turn stator faults in permanent-magnet synchronous motors by converting raw sensor signals into continuous-wavelet scalograms and classifying them with a CNN.
Trained on the real KAIST motor dataset (DOI 10.17632/rgn5brrgrn.5) — not synthetic data.
Results, reported honestly:
Sensor channel Balanced accuracy Vibration @ 25.6 kHz 1.00 Current @ 100 kHz 0.69 The vibration result is excellent. The current-channel result is not, and the repository says so. The dataset contains only four healthy recordings, so the perfect vibration score may not generalize — that caveat is in the README, not buried in a footnote.
38 unit tests. This is the project I would want to be judged on: real data, a strong result, a weak result, and no hiding of either.
Merged Open-Source Contributions — OpenCV & OpenDR # Three pull requests accepted into major open-source projects. Two landed in OpenCV core, one of the most widely used computer-vision libraries in the world.