Machine Learning and Dynamic Optimization for Engineers

Machine Learning and Dynamic Optimization is a 3 day short course on the theory and applications of numerical methods for solution of time-varying systems with a focus on machine learning and system optimization. It includes hands-on tutorials in data science, classification, regression, predictive control, and optimization.

Dates

Location

Jan 13-15, 2020

Seoul, South Korea (47 participants)

Mar 24-26, 2020

Manama, Bahrain with University of Bahrain

May 11-15, 2020

Salt Lake City, Utah, USA (5 day) with APCO, Inc

May 20-22, 2020

Idaho Falls, Idaho, USA

July 14-16, 2020

Houston, Texas, USA

Jan 4-8, 2021

Seoul, South Korea (73 participants)

Dec 13-16, 2021

Seoul, South Korea (16 participants)

Concepts taught in this course include machine learning, regression, classification, mathematical modeling, nonlinear programming, and advanced control methods such as model predictive control.

Day 1

Topic

Activity

9:00 AM

Overview of Course, Optimization, and GEKKO

Gekko Introduction and Machine Learning

9:30 AM

TCLab Overview

Begin Python with TCLab

10:30 AM

Break

10:45 AM

Digital Twin with Physics-based Simulation

Lab A - SISO Model or Lab B - MIMO Model

12:00 PM

Lunch Break

1:00 PM

Machine Learning Classification, Deep Learning, and LSTM Networks

TCLab Classification

2:00 PM

Data Regression for SISO/MIMO Identification

Lab C - Parameter Estimation

3:00 PM

Break

3:30 PM

Moving Horizon Estimation with Objectives/Tuning

Lab D - MHE or Lab E - Hybrid Model Estimation

4:30 PM

TCLab Incubator Project

5:30 PM

Day 1 Review

Day 1 Assessment Activity

6:00 PM

Conclude Day 1


Day 2

Topic

Activity

9:00 AM

Dynamic Control Introduction

Velocity Control

9:30 AM

Dynamic Optimization Benchmarks

Integral Objective and Economic Objective

10:30 AM

Break

10:45 AM

Crane Pendulum or Flight Control

Lab F - Linear Model Predictive Control

12:00 PM

Lunch Break

1:00 PM

Nonlinear MPC, Control Objectives/Tuning, and Orthogonal Collocation

Lab G -Nonlinear Model Predictive Control

2:00 PM

Mixed Integer Optimization

Mixed-Integer TCLab

3:00 PM

Break

3:30 PM

Multi-Objective Optimization

Lab H - Adaptive Model Predictive Control

4:30 PM

Group Projects

Project Overview

5:30 PM

Day 2 Review

Day 2 Assessment Activity

6:00 PM

Conclude Day 2


Day 3

Topic

Activity

9:00 AM

Group Project Proposals

Project Proposals

9:30 AM

Physics-based Modeling Review

Stage 1 - Develop Digital Twin Model

10:30 AM

Break

10:45 AM

Machine Learning and Time-Series Regression Review

Stage 2 - Machine learning or time-series models

12:00 PM

Lunch Break

1:00 PM

Parameter Regression Review

Stage 3 - Parameter Regression

2:00 PM

Moving Horizon Estimation Review

Stage 4 - Adaptive Model Update (MHE)

3:00 PM

Break

3:30 PM

Model Predictive Control Review

Stage 5 - Model Predictive Control

4:30 PM

Group Project Presentation Preparation

5:30 PM

Group Project Presentations (3 min each)

6:00 PM

Conclude Day 3 and Course

Certificates of Completion

Each participant has a Temperature Control Lab for hands-on exercises. Exercises are conducted in-class with additional supplementary material that can be completed after the class concludes. The objective of the 3 day short-course is to give enough background information so that researchers and practitioners can extend the methods to applications related to their field of study or industrial process.