October 14, 2021

Recording of the Presentation "RSM Surrogate Model under Monotonic Constraint for Enhanced System Monitoring and Control"

In this project, we want to identify a mathematical method to:

• Create virtual sensors based on machine learning applied to multi-physics systems
• Implement the hereabove virtual sensors in an embedded software
• Constrain the surrogate model to ensure compatibility with closed loop control

The chosen method is the Response Surface Model of GTApprox. The main reasons are:

• The mathematical simplicity (quadratic function)
• Small memory and CPU usage
• Consistency with the physics modelled