UTFacultiesEEMCSDisciplines & departmentsPSEducationReinforcement learning for controlling a critical quality variable in a continuous process

Reinforcement learning for controlling a critical quality variable in a continuous process

Intern assignment

By

Pawel Roman, Le Viet Duc

Thesis or intern assignment

Reinforcement learning for controlling a critical quality variable in a continuous process

Wavin department

Discovery

Education of candidate

Data Science / Computer Science / Applied Mathematics / Industrial Engineering (Bachelor)

 

Description of the assignment

Maintaining a critical quality variable close to its target can be challenging due to disturbances and changing operating conditions. Reinforcement Learning (RL) offers a data‑driven approach to learn control strategies, but safety and stability are essential.

In this assignment you will explore a safe RL‑based control concept using a simplified simulation or data‑driven surrogate model. You will define states, actions, and rewards in an engineering‑meaningful way, compare RL performance with baseline control strategies, and incorporate constraints to ensure safe behavior. The result is a proof‑of‑concept control strategy, not a deployed controller.

 

Goal of the assignment

Demonstrate a safe reinforcement‑learning control strategy that keeps a quality variable close to its target.

 

Additional information

Intern type: Bachelor assignment

Duration: 2 months

Workload: Full‑time (minimum 32 hours/week)

Gross reimbursement per month (full‑time): 400 EUR

 

About Wavin

Wavin is an innovative solution provider for the building and infrastructure industry globally. With over 60 years’ of expertise, we aim to solve global challenges in water supply, sanitation, climate-resilient cities and building performance. We are committed to building healthy, sustainable environments. Wavin is headquartered in the Netherlands, and has presence in over 25 countries.

 

Contact:

Le Viet Duc