Internship Data Science Building a physics-informed ML model

Building a physics-informed ML model for predicting mechanical properties of rolled steel products.

Wat ga je doen?

Tata Steel R&D have developed a through-process physics based model (MeCoMeP) for predicting mechanical properties of steel grades using measurement data obtained throughout the processing of a steel coil. The model is comprised of a data pipeline used to gather the data, combined with standalone C++ executables which perform the model calculations. The results can be used to assist the different production facilities in product development and optimisation, root cause analysis of product and process failures and for supplying customers with positional mechanical property data to enable pre-setup in forming parts for the automotive industry.

The MeCoMeP model is a phenomenological model that tries to describe all relevant metallurgical processes during strip hot rolling and galvanising e.g. grain growth, recrystallisation, phase transformation, precipitation, cementite formation and carbon rejection. Physical equations describe the microstructural evolution taking place during each step of the process and the mechanical behaviour of the resultant microstructure in order to predict the mechanical properties of the steel after processing.

The advantage of this approach is that process steps are identified which are known to affect mechanical properties and are consistent throughout all grades produced over a particular production line. The limitation is that the model does not take into account other process parameters which could affect properties especially when subject to significant variations during operation.

The goal of this research is to combine the physics based model with a more generalised data driven model to improve the predictive ability of existing and future steel products.

• Aim: Improve accuracy of excisting physics and ML models in predicting mechanical properties
• Deliverable: Algorithm for fitting the generalised physics based model prediction to individual steel grades and continuous monitoring of model fit quality:
• Start from base model with 4 steel grade families
• Improving the model fit for individual steel grades
• ​​​​​​​Investigate alternative fitting models for grade specific model predictions
• Compare non-linear modelling techniques against current linear regression model to improve the model performance
• Create model quality monitor to track model fit over time​​​​​​​

Herken jij jezelf hierin?

• You are a MSC student in the field of data science.
• You are familiar with programming languages such as C++, SQL and Python.
• You have an affinity for physical sciences
• You are looking for a graduate internship.

The assignment is intended for an internship of nine months.

Wat bieden wij jou?

• An internship allowance of €450,-.
• A challenging and instructive internship with good supervision.
• Together with you, a clear assignment will be designed in advance, such that it meets the requirements of your college/university.
• You can gain valuable experience within a large organisation and the Netherlands' only steel manufacturing industry.
• You can count on a no-nonsense and result-oriented organisation, in which you can make a concrete contribution to the realisation of improvement objectives.
• We offer the possibility to make use of our library services.
• You can sign up for 'Interns For Steel'. This is a group set up for trainees by trainees. They organise regular tours, lunch together once a week and there is a monthly get together.

Wat is jouw werkomgeving?

The project will be conducted within R&D department of Tata Steel Ijmuiden with great opportunities for networking and being closely involved in the future developments. On a daily base you will working in the R&D innovation centre office, surrounded with other students and colleaguas with different kinds of know-how.

Our ambition

Our ambition is to become a green, clean, and circular steel company that everyone is proud of. We aim to improve how people around the world work, live and move through sustainable steel. To succeed, we need talents like you. Will you join us in shaping the future?