Master thesis: Artificial intelligence in metallography

The research institute Swerim provides applied research within mining engineering, process metallurgy and materials and manufacturing engineering, mainly for the mining, steel and metals industry. Swerim has 190 co-workers in two locations in Sweden - Luleå and Stockholm.

Background
In recent years artificial intelligence methods have started to find their use in metallographic analysis. Traditional metallographic methods often are time consuming and require field experts to evaluate the results at all steps. Microstructure analysis of the metallic structure are commonly done using light/electron microscopes. Newly developed microscopes allow very fast and automated imaging of material at very high resolution leaving the field experts behind when it comes to the analysis of this number of images.

There has been a developing recent interest in using machine learning for analyzing electron and light microscopy images. For example, machine learning methods have been developed to identify different phases in steel material, quantify particles, or classify inclusions. This becomes important when traditional software that use for example thresholding to annotate specific features fail due to various reasons such as: artifacts, sample preparation differences, poor resolution in some part of the data, the difference in grey level amongst images, etc. The use of automated methods can minimize human time consumed on image analysis and increase the quality of the results significantly.

There are often common challenges for image analysis which regardless of the type of material, can be addressed using similar methods. For example, analysis of particles, precipitates, or non-metallic inclusions in steel or aluminum materials can be analyzed using similar algorithms as porosity analysis in 3D printed materials.

The proposed project aims to build a simple but multipurpose robust analysis protocol that can be used and further developed to automatically remove artefacts analyze particles, precipitates, inclusions, or porosity found in microscopy images for metallic materials. 

The diploma work will be performed at Swerim AB in Kista and will be financed by Swerim´s member research programs in Metallography and Microanalysis (META) and Värmebehandlingscentrum, (VBC) consisting of 12 and 18 member companies, respectively. Representatives from the companies will have an active interest in the work and will contribute with research questions relevant to their activities.
 
Objectives and learning outcome

* Understanding the importance of particle and inclusion classification in steel materials and porosity analysis for example in 3D printed materials (quality control on final structure)
* Review machine learning and deep learning methods used in metallography  
* Developing an analysis tool, or modifications required for using existing software that can separate certain features and output quantitative information: this outcome will be modified based on the successful applicant interest

Required qualifications
Project student in Material science, physics, engineering, or similar fields with a good knowledge in metallic materials. Basic knowledge and some experience in programming is required. Any experience or prior knowledge about machine learning methods is an advantage.  
 
For further questions or more details please contact: Shirin Nouhi, shirin.nouhi@swerim.se