Ongoing and completed student projects
Here we present ongoing and completed projects within Sentio.
Classification of brake discs
Master's project with LTH
Supervisors: Maria Sandsten, Mathematical Statistics, and Oleksandr Gutnichenko, Production and Materials Engineering.
Author: Emil Svalfors
Description: Accurate classification of time-varying and non-stationary time-series signals is a central problem in many scientific and engineering disciplines, including bioacoustics, seismology and climate science. The aim of this study is to investigate the possibility of connecting of time-frequency representations (TFRs) with machine learning techniques to improve signal classification. The idea behind the study was to investigate if analysis of sound can be used to differentiate between brake discs in good, versus bad, working order.
Title and link to thesis: Optimizing Time-Frequency Representations for Time-Series Signal Classification Using Neural Networks
Acoustic leak detection signal analysis
Master's project with Alfa Laval
Supervisors: Axel Knutsson, Alfa Laval & Maria Sandsten, Mathematical Statistics.
Author: Oscar Stackenland
Description: Alfa Laval produces millions of heat exchangers every year and among those some are subject to faults and leakage. To find and classify these faults, a huge amount of time has to be expended by technicians and materials experts. The goal with this master thesis project is to explore if this process can be done more efficiently by looking at sound recordings of water-filled heat exchangers which give rise to air-bubbles with clear popping sounds once it reaches the water surface.
Title and link to thesis: Acoustic Leak Classification in Heat Exchangers by Time-Frequency Analysis and Machine Learning
Wireless connectivity solution for industrial sensing applications
Advanced Course project in Electrical and Information Technology (EITN35)
Supervisors: Baktash Behmanesh, Electrical Information Technology & Adam Burke, Solid State Physics.
MSc students: Armon James & Tingyi Fan
Description: This project focuses on developing a custom printed circuit board (PCB) capable of acquiring data from sensors embedded in devices such as metal cutting tools. Communication is based on the Bluetooth Low Energy (BLE) standard. The proposed solution supports a mesh network architecture, enabling the management of numerous sensors while providing wide-area coverage. Development begins with off-the-shelf components, followed by the fabrication of custom PCBs tailored to fit the tool in later phases.
Project ongoing.