| KEYWORD |
Quality Control of Metal Powders for Additive Manufacturing Using Models Based on Convolutional Neural Networks
Reference persons MARCO ACTIS GRANDE
External reference persons Marta Ceroni
Matteo Giardino
Thesis type APPLIED RESEARCH
Description The thesis consists of the development of a convolutional neural networks-based model for the detection and quantification on contaminants in metal powders for additive manufacturing, employing absorption measurements in the visible and near-infrared.
The first part of the thesis work will focus on building a database of optical and magnetic properties of different metals (refractive indices and magnetic permeability) from data available in the literature. A custom code will be developed to extract images contained in PDF files of scientific articles, which will then be segmented to extract the desired data.
The optical data thus extracted will be provided to a Python code (already written) that will allow for estimating the optical response of different batches of metal powders.
To detect and quantify any possible contaminants in the powders batched produced in the gas atomization plant at the Alessandria campus of the Politecnico di Torino, the student will develop a regression model based on convolutional neural networks. The model will be trained on simulated data and subsequently validated on experimental data obtained from powders with known amounts of contaminants added.
This thesis work is highly innovative and has clear applications in an industrially relevant field such as metallic additive manufacturing.
Required skills Candidates are required to have experience in developing deep learning models based on neural networks using Python and one library of their choice among PyTorch, Keras, and TensorFlow. The ability to create graphical interfaces using libraries such as Tkinter or Qt is also particularly appreciated.
No prior knowledge in metallurgy and solid-state physics is required to undertake this thesis project. The thesis supervisor will provide the necessary material to acquire the minimal knowledge required for a better understanding of the work to be carried out.
Deadline 06/06/2025
PROPONI LA TUA CANDIDATURA