# Calvin Tsay: Adjustable Formulations for Process Modeling and Optimization

## Hybrid seminar, on location and available on zoom

Abstract:

Computational optimization tools have found an abundance of applications in engineering processes, supporting decisions in process design, operations, and control. The performance of these decision-making activities strongly hinges on selecting an appropriate mathematical model, or representation of the underlying physical system. A detailed, first-principles mathematical model may capture system behavior accurately, but its size may be computationally prohibitive in certain practical applications. In this talk, I will describe adjustable model approximations that can produce compact optimization formulations with suitable accuracy. Two illustrative approaches will be introduced: (i) representing process dynamics in a lower-dimensional space for scheduling, and (ii) approximating the “convex hull” of neural network surrogate models. In both strategies, the dimensionality of the formulation is adjusted to tune the tradeoff between model accuracy and computational expense.

**Time: **
Thu 2022-04-07 11.00 - 12.00

**Location: **
3721

**Video link: **
Zoom link, meeting id 63658381373

**Language: **
English

**Participating: **
Calvin Tsay, PhD

**Calvin Tsay**, PhD

EPSRC David Clarke Fellow & Imperial College Research Fellow

Department of Computing

Imperial College London