Machine Learning in Engineering Calculations (FSI-QAI)

The aim of the subject is to familiarize students with the basic approaches of the application of machine learning, specifically Reinforcement Learning in engineering calculations.

The graduate of the course will gain basic knowledge of algorithms and data structures, useful for effective implementation of Reinforcement Learning algorithms through the Python programming language.

The graduate will also gain practical experience in tasks in the field of connecting machine learning and selected engineering calculations, which can serve as inspiration for further development in this area.

Basic knowledge of physical and engineering principles, as well as basic knowledge of the Python programming language.

The course outlines possible ways of applying machine learning in the context of engineering calculations. Students will get to know the basic principles of machine learning and artificial intelligence, with an emphasis on Reinforcement learning. The course includes an introduction to the Python programming language and its use for implementing suitable libraries for linking dynamic computational models and artificial intelligence tools.

The course-unit credit award is conditional on active participation in the exercises, where the activity within the sub-tasks is continuously checked.

Attendance in exercises is compulsory, participation is checked by the teacher. The form of replacement of missed lessons is solved individually with the lecturer or with the guarantor.

1. Division of machine learning/artificial intelligence


2. Reinforcement learning (RL): definition, introduction to basic concepts


3. Introduction to the Python programming language


4. Supervised learning in the Python environment (learning from data)


5. Example of an RL task in the Python environment (Gymnasium)


6. Model-based and model-free RL


7. FMU: Creation, usage


8. Construction of a computational model (Adams/Chrono/NVIDIA Modulus/Mujoco)


9. Definition of inputs and outputs for RL: observations, actions


10. Selection/creation of an agent and its policy


11. Definition of reward function


12. Training: definition of parameters influencing the process


13. Using a trained agent in a simulation

SANGHI, Nimish. Deep Reinforcement Learning with Python. 1st ed. Apress Berkeley, CA, c2021. ISBN 978-1-4842-6809-4.