My Bachelor's thesis:
Triboelectric Nanogenerator-based Material Classification with Machine Learning Assistance

Download (pdf)

In the summer semester 2023, I wrote my bachelor's thesis in physics at the Chair of Functional Materials at TUM. The aim of the thesis was to distinguish different materials/objects with triboelectric nanogenerators using machine learning.

Triboelectric nanogenerators were invented in 2012 and can convert mechanical energy into electrical energy. They can be very simple and inexpensive to produce and thus have great potential in a variety of applications.

First, I fabricated a simple Triboelectric Nanogenerator from a PDMS film and copper tape. I then repeatedly pressed the nanogenerator against various materials to build a data set. Using this data, I trained a neural network which was then able to identify which material the nanogenerator was pressed against, with acceptable accuracy. This means such a nanogenerator could, for example, be suitable for use on the fingers of a robot and would enable the robot to "feel" what it touches.

Since the measuring devices to which the nanogenerators are usually connected are very large and expensive, they are not suitable for use in mobile applications. That's why I tried to build a cheap and mobile measuring device for measuring the nanogenerator voltage. It uses a microcontroller (Arduino-Nano) and an amplification stage to record the signal and transmit it to a computer via Bluetooth. However, the device was very susceptible to electromagnetic interference (EMI), which is why it was not used for the experiments.

The thesis was graded with a 1.0.