Abstract:Machine learning recently has been used to identify the governing equations for dynamics in physical systems. The promising results from applications on systems such as fluid dynamics and chemical kinetics inspire further investigation of these methods on complex engineered systems. Dynamics of these systems play a crucial role in design and operations. Hence, it would be advantageous to learn about the mechanisms that may be driving the complex dynamics of systems. In this work, our research question was aimed at addressing this open question about applicability and usefulness of novel machine learning approach in identifying the governing dynamical equations for engineered systems. We focused on distillation column which is an ubiquitous unit operation in chemical engineering and demonstrates complex dynamics i.e. it's dynamics is a combination of heuristics and fundamental physical laws. We tested the method of Sparse Identification of Non-Linear Dynamics (SINDy) because of it's ability to produce white-box models with terms that can be used for physical interpretation of dynamics. Time series data for dynamics was generated from simulation of distillation column using ASPEN Dynamics. One promising result was reduction of number of equations for dynamic simulation from 1000s in ASPEN to only 13 - one for each state variable. Prediction accuracy was high on the test data from system within the perturbation range, however outside perturbation range equations did not perform well. In terms of physical law extraction, some terms were interpretable as related to Fick's law of diffusion (with concentration terms) and Henry's law (with ratio of concentration and pressure terms). While some terms were interpretable, we conclude that more research is needed on combining engineering systems with machine learning approach to improve understanding of governing laws for unknown dynamics.

What problem does this paper attempt to address?

### Problems the paper attempts to solve This paper aims to explore the feasibility and effectiveness of machine - learning methods in identifying control equations in the dynamics of complex engineering systems. Specifically, the researchers are concerned with: 1. **Background and motivation**: - At the micro - scale, control laws such as heat transfer, diffusion, and pressure change are already very clear and have been studied for decades. - However, for complex engineering systems, these control laws have not been fully established. The dynamics of these systems are crucial for design and operation, so it is very beneficial to understand the underlying mechanisms. 2. **Research objectives**: - Identify the control equations of complex engineering systems by using machine - learning methods, especially the Sparse Identification of Nonlinear Dynamics (SINDy) method. - Focus on the distillation column, a typical unit operation in the chemical process, because the dynamics of the distillation column are very complex, combining heuristic rules and basic physical laws. 3. **Research methods**: - Use ASPEN Dynamics software to generate dynamic simulation data of the distillation column. - Apply the SINDy method to extract control equations from time - series data. - Test the dynamics of the system by externally perturbing the input stream and generate time - series data. 4. **Expected results**: - Hope to reduce the number of equations in dynamic simulation through machine - learning methods, from thousands of equations to only 13 (one equation for each state variable). - Evaluate the prediction accuracy of the extracted equations on the test data, especially their performance within the perturbation range. - Try to explain whether some terms in the extracted equations are related to known physical laws (such as Fick's law of diffusion and Henry's law). 5. **Research significance**: - If successful, this will provide new tools and methods for the design and operation of complex engineering systems. - Help to understand the dynamic mechanisms of complex systems for which control laws have not been fully established. ### Main findings - **Reduction in the number of equations**: Through the SINDy method, the number of equations in dynamic simulation was successfully reduced from thousands to 13. - **Prediction accuracy**: On the test data within the perturbation range, the prediction accuracy is relatively high; but on the data beyond the perturbation range, the equations perform poorly. - **Explanation of physical laws**: Some of the extracted equation terms can be explained as related to Fick's law of diffusion and Henry's law, but no definite conclusion has been reached. - **Future research directions**: Further research is needed on how to combine engineering systems and machine - learning methods to better understand the unknown dynamic mechanisms. ### Conclusion The research results show that machine - learning methods have certain potential in identifying the control equations of complex engineering systems, but still need further improvement. In particular, more research is needed to combine engineering knowledge and data - driven methods to improve the understanding of the dynamics of complex systems.

Can Machine Learning Identify Governing Laws For Dynamics in Complex Engineered Systems ? : A Study in Chemical Engineering

Automatedly Distilling Canonical Equations from Random State Data

Reduced order dynamical models for complex dynamics in manufacturing and natural systems using machine learning

Discovering governing equations from data: Sparse identification of nonlinear dynamical systems

Application of machine learning in chemical engineering: outlook and perspectives

Machine Learning for Molecular Thermodynamics

Machine Learning in Chemical Engineering: Strengths, Weaknesses, Opportunities, and Threats

An integrated approach for machine-learning-based system identification of dynamical systems under control: application towards the model predictive control of a highly nonlinear reactor system

Chapter 14. Machine Learning Applications in Chemical Engineering

Machine learning based system identification of a realistic heat integrated distillation column using particle swarm optimization

Machine Learning for Fluid Mechanics

Deep Generative Modeling for Identification of Noisy, Non-Stationary Dynamical Systems

Discovering governing equations from data by sparse identification of nonlinear dynamical systems

Development of a smart multiphase system for disperse flows using machine learning

Machine Learning for Predicting Thermodynamic Properties of Pure Fluids and Their Mixtures

Machine Learning Phase Space Quantum Dynamics Approaches

Scientific Deep Machine Learning Concepts for the Prediction of Concentration Profiles and Chemical Reaction Kinetics: Consideration of Reaction Conditions

Physics-informed identification of PDEs with LASSO regression, examples of groundwater-related equations

Automatically identifying ordinary differential equations from data

Deductive Machine Learning Challenges and Opportunities in Chemical Applications

An Analytical Predictive Control Law for A Class of Nonlinear Processes