Franklin Mdemaya Email & Phone Number

I'm a student passionate about data acquisition, processing and machine learning. During my master degree in digital automation engineering, I focused on digital infrastructures and studied subjects like data science, AI and Internet of things (IoT). I developed a huge interest for the mentioned subjects, and later on I've been selected by HPE Group to carry out an internship at their company. Starting from may 2024, the internship focused on detection and diagnosis of anomalies on shaft bearings in electric motors. In particular, I had to develop a robust algorithm that is able to correctly detect abnormal behaviors of the motor, and classify it depending on the type of fault. I mainly focused on the identification and classification of 2 types of faults: bearing looseness and bearing axis misalignment. The main challenge has been to design a unique threshold anomaly detection technique to identify both faults. In order to do that, I acquired skills in data preprocessing techniques like outlier removal, Fourier transforms and wavelet transforms for non stationary signals. I then used two different approaches and compared the results: a convolutional autoencoder (CAE) mixed with contrastive learning and a convolutional autoencoder mixed with kmeans. In both cases, the CAE has been used for dimensionality reduction. Then in the former, a contrastive loss has been applied to closely cluster similar representations (normal together and far away from anomalies); In the latter, a kmeans classifier has been used to cluster together normal data points based on specific metrics I computed from the CAE. I already obtained very good results in both cases, as I'm about to conclude my internship. I used python, tensorflow/keras as main tools to tackle this challenge! I've got extremely passionate on this topic, and I'd like to keep working on this field, willing to learn more and get experience!