Research

Research Interests:

Artificial intelligence is creating new paradigms and opportunities in the design of advanced process control systems for chemical processes. Designing control systems and optimization tools that utilize machine learning techniques with well-characterized accuracy is a new frontier in process systems engineering that will impact the next generation of industrial control systems. 

Our group’s research interests are in the general area of process systems engineering and artificial intelligence. Our objective is to integrate artificial intelligence and chemical engineering to

  • develop cutting-edge solutions to important industrial considerations, including production profitability, process operational safety and cybersecurity, and
  • accelerate the discovery and design of novel catalysts, materials, etc.

Major themes of our current research are:

 

Machine Learning in Process Control and Optimization

Traditionally, model predictive control (MPC), a constrained optimization-based control problem formulation that is the gold standard employed in advanced control of chemical processes,  is formulated with linear data-based empirical models and is used to compute control actions to maintain optimal process operation while accounting for process and control actuator constraints. However, chemical processes are inherently nonlinear and often require nonlinear models in order to be controlled efficiently.

Machine learning (ML) tools like recurrent neural networks and ensemble learning provide an efficient way to build nonlinear dynamic models from data t

hat can be used in the model predictive control system, thereby improving control system performance. In the proposed research, we will develop both 1) theoretical study of ML-based control approaches, and 2) their applications to real industrial chemical processes. Various novel ML modeling methods will be investigated, including a) physics-informed ML, 2) online ML, and 3) reduced-order ML for large-scale chemical processes, etc.

 

Machine Learning in Process Operational Safety and Cybersecurity

Safety is critical in the chemical process industries due to the severe consequences for both lives and property when safety is not maintained. One novel perspective on process safety that has been advocated in several recent works is a systems view of process safety in which accidents are seen as the result of the process state migrating to an unsafe operating region from where an accident may quickly follow. In the context of the proposed research, we will investigate the use of ML tools both for nonlinear process modeling and for the characterization of the unsafe regions and the computation of the Barrier functions using feedforward neural network function approximators.

Cybersecurity has become increasingly important in chemical p

rocess industries in recent years as cyber-attacks that have grown in sophistication and frequency have become another leading cause of process safety incidents. Unlike fault diagnosis, the intelligence of cyber-attacks and their accessibility to control system information has recently motivated researchers to develop cyber-attack detection and resilient operation control strategies to address directly cybersecurity concerns. In this direction, we will develop an integrated detection and control system for process cybersecurity, in which several types of intelligent cyber-attacks, ML detection methods and resilient control strategies will be studied.

 

Computational Modeling and Simulation of Chemical Processes

Recent developments in computing power and technology are providing chemical engineers with new tools for understanding and predicting process behavior. An important part of the research in our group involves leveraging this technology to improve process models and control system design.  Specifically, we will apply the ML-based MPC designs to large-scale processes in Aspen Plus Dynamics using industrial-data informed parameters and accounting for process disturbances and noise.

(Figure: the isometric view of an industrial-scale, top-fired, co-current reformer)

We will further investigate the implementation of machine-learning-based MPC within a high-fidelity computational fluid dynamics (CFD) model (created in ANSYS Fluent CFD software). To improve computational efficiency, ML methods will be used to derive a set of reduced-order models to predict the nonlinear distributed dynamics of the process model. In this study, we will also develop user-friendly software via seamless integration of ASPEN Plus Dynamics, Data Management Tools and of the Machine-Learning MPC schemes.

 

 

Digital Twin For Cyber-Physical Production Systems (CPPS)

In this project, we focus on developing data-driven methods for the eco-efficient utilisation of energy resources, which include renewable energy sources and a Battery Energy Storage System (BESS), to support production systems. The uncertain nature of renewable energy sources makes it challenging to utilise them for stable and resilient grid operation. To address this challenge, data-driven, AI and machine learning methods were employed in this project.

A testbed site was implemented with an industrial partner’s manufacturing factory, where it was retrofitted with IIoT-enabled sensors to create a CPPS. Additionally, a digital twin will be developed for the CPPS for simulation, integration, testing, monitoring, and maintenance.

 

 

 

 

Machine Learning in Accelerating Chemical Discovery With HTE Techniques

The integration of ML with high-throughput experimentation (HTE) techniques forms an indispensable cornerstone in accelerating chemical discovery and development. ML-augmented HTE enables contemporaneous screening, synthesis, and navigation through the enormous potential chemical search space. Specifically, the screening and optimization stages of the HTE can be accelerated through ML by extracting and elucidating the intricate relationships between synthesis variables and measurable material performance. ML-reinforced HTE has demonstrated promising results in various domains (e.g., the discovery of new catalysts and molecules). In this project, we will use ML tools to develop to learn the relationship between different species concentrations/physical parameters and selectivities/yields.

 

 

Our current research projects are:

1. Physics-based machine learning (ML) algorithms for modeling of chemical processes using a priori process knowledge

2. Machine learning modeling of large-scale chemical processes simulated in Aspen Plus / CFD simulation software

3. ML-based optimization theory and application to process design & synthesis, process operations including scheduling, supply chain and real-time optimization, and process control

4. Lyapunov-based model predictive control algorithms with statistical machine learning theory

5. ML in accelerating photocatalysis and electrochemical synthesis with HTE data

 

 

 

Skip to toolbar