Master thesis: Optimization of insert-tray matching using machine learning
Sandvik Coromant in Gimo, Sweden offers a thesis project position within the area of data science.
Sandvik Coromant in Gimo is a world leader in the manufacture of cemented carbide tools for turning, milling and drilling in metallic materials. With around 1500 employees, it is also Uppsala County's largest private employer.
The production in Gimo is divided into two factories, one for the manufacture of cemented carbide inserts and one for tool holders. Both factories are world leaders in their respective field and an example of this is that the tool factory was recognized as one of the world's 16 'Lighthouse' manufacturers by the World Economic Forum in 2019.
Our biggest customers are the metal, automotive and aerospace industries, and we aim to meet the market's increased demands for new products with precision and durability.
This thesis project is an opportunity to apply theoretical approaches to a practical setting, where statistical tools such as machine learning can be used to solve optimization issues in the manufacturing of cemented carbide inserts.
Background and scope of the project
Throughout the carbide insert factory, the products must be placed and replaced on different tray systems to accommodate the individual demands of the different operations and machines. There are many variables considered for the selection of proper and functioning trays for the different sections.
The data used to determine which tray is used for an order is largely based on experimental or non-existent data, contributing to sub-optimal or improvised selections of trays. What is required to ascertain the best approach going forward is collecting relevant tray and insert data from databases and determine a workflow or reference for optimal matching depending on parameters.
The initial scope of the study is restricted to standard articles and a single section of production. There are good opportunities for expansion beyond the initial scope.
The central questions we want to be answered are 1 and 2, with 3 and 4 in the optional, extended scope:
- What insert variables are determining parameters for tray matching?
- Can we make good predictions of which trays should hold a given insert at a specific operation, if the model is trained/based on existing data for standard articles?
- Can we make good predictions of which trays should hold a given insert at multiple operations, if the model is trained/based on existing data for standard/special articles?
- Can we make better predictions than the existing data on articles already assigned a tray?
The study includes the collection of insert/tray data and to determine the key parameters for an optimized matching. This includes:
- Designing and planning the study
- Collecting the relevant information
- Performing statistical analysis or comparable machine learning approach on collected data
- Interpret the results
- Give suggestion of the recommended course of action
The study is suitable to be performed by an M.Sc. student in Engineering Physics/IT. Students from alternate disciplines may also be considered if relevant Statistics/Machine Learning/Data Science courses/experience has been acquired.
We actively work to create a workplace that is characterized by diversity and inclusion.
The thesis project consists of 30 credits and lasts 20 weeks, starting first quarter of 2021. The position is placed in Gimo, Uppsala.
For more information about this thesis project, please contact:
Alvin Ljung, phone: +46 70 616 21 13
Lars Rönne, phone: +46 173 848 72
Please send your application to:
Katarina Thyrestam, email: firstname.lastname@example.org
The selection is done continuously. Please send us your application as soon as possible, but no later than 29 November 2020.