Course syllabus
Course-PM
IMS135 IMS135 Machine learning and data-driven modelling in mechanics lp1 HT23 (7.5 hp)
Course is offered by the department of Industrial and Materials Science (IMS) in cooperation with Mechanics and Maritime Sciences (M2).
Contact details
- examiner: Magnus Ekh (magnus.ekh@chalmers.se)
- lecturer: Magnus Ekh (magnus.ekh@chalmers.se)
- teachers: Jim Brouzoulis (jim.brouzoulis@chalmers.se) Lars Davidson (lars.davidson@chalmers.se), Peter Folkow (peter.folkow@chalmers.se), Fredrik Larsson (fredrik.larsson@chalmers.se)
- supervisors: Caroline Ansin (caroline.ansin@chalmers.se), Shuai Li (shuai.li@chalmers.se)
Course purpose
The course includes an introduction, evaluations and applications of machine learning and data-driven modelling for problems in mechanics, solid mechanics and fluid dynamics.
Course contents
Data-assisted modelling: singular value decomposition, principal component analysis, correlation; Machine learning concepts: optimization techniques, regularization, neural networks, training a network, activation functions, automatic differentiation, surrogate models, physics-informed neural networks, data-driven inference, data-driven identification, deep energy method, reduced-order modelling, Support Vector Regression (SVR), Support vector machines (SVM), Python programming, PyTorch.
Schedule
Course literature
E-books:
Deep learning in computatational mechanics, Stefan Kollmannsberger
Data-driven science and engineering : machine learning, dynamical systems, and control, Steven L. Brunton, J. Nathan Kutz.
Course design
Lectures and supervision. Supervised projects where machine learning and data driven modelling are applied to material mechanics, fluid dynamics and dynamics. The choice of project is decided by background and interest of the student.
Lectures are complemented by selected videos on: databookuw.com
Week 1-4 (and repetition on week 8) contains lectures and supervision (own problem solving). The contents of this part will be the base for the exam. The exam will be conducted in computer rooms where the computers have Python (anaconda) and Matlab installed. But no connection to internet. The exam takes place 2023-10-28 from 14:00 to 18:00 on campus Johanneberg (and re-exam is on 2024-01-03). The exam is graded failed, 3, 4, 5 and gives 4.5 credits.
Week 5-7 contains project work, and supervision. Presentations of the projects will be given in week 8 (2023-10-17, 13:15-15:00). It is mandatory to present your project at this time and also to participate during the presentations of the other groups. Project work is only graded as approved or not approved. The project part of the course gives: 3 credits. If you would fail the project or cannot participate then you need to present on another occasion and write a short report about your project (decide details with your responsible supervisor). Responsible supervisors for the projects are: Jim Brouzoulis, Lars Davidson, Peter Folkow and Fredrik Larsson.
Course summary:
| Date | Details | Due |
|---|---|---|