Preface

Reaction engineering and process systems engineering are two distinct yet interconnected fields within chemical engineering that are critical to the design, operation, and manufacturing of various chemical and biochemical products. One of the main research goals these domains share is the development of high-accuracy mathematical models to simulate and predict complex reaction kinetics for process optimisation, control, and scale-up. Traditionally, this has relied heavily on first-principles knowledge to suggest possible kinetic model structures, followed by optimisation algorithms to estimate kinetic parameters and evaluate the model’s identifiability. However, this process can be time-consuming as many chemical and biochemical reactions involve intricate mechanisms that are not fully understood. Furthermore, the highly nonlinear nature of the model structure poses a significant challenge to existing optimisation algorithms, increasing the risk of identifying sub-optimal solutions for kinetic parameters that can undermine the accuracy and reliability of the model.

Over the past decade, the rapid growth of machine learning techniques and the increasing volume of data generated by the process industry have led to a significant shift from physical modelling to more data-driven modelling for reaction engineering applications. The integration of machine learning and first-principles models (known as hybrid or physically informed data-driven modelling) has been particularly successful in identifying (bio)chemical kinetics, optimising complex reactions and scaling up industrial reactors. This modelling approach has been touted as the next-generation norm, with interpretable AI playing a vital role in future reaction engineering research. This book aims to introduce the fundamental principles and practical aspects of different advanced modelling techniques for chemical and biochemical reaction engineering applications, ranging from mechanistic models based on rigorous process systems engineering algorithms to data-driven and hybrid models developed using state-of-the-art machine learning algorithms.

This book is an excellent resource for a range of professionals (e.g. postgraduate students, academics and industrial researchers) working in the fields of chemical and biochemical reaction engineering, engineering chemistry or process systems engineering. It is also ideal for computer scientists and data scientists seeking to understand the specific challenges of applying data intelligence algorithms in the chemical and process industries. To benefit fully from this book, readers should have a basic understanding of reaction engineering and engineering chemistry principles and/or knowledge of fundamental machine learning and optimisation theory.

Overall, this book aims to provide an introduction to applying innovative data intelligence techniques to reaction engineering. It features a wide range of case studies with in-depth discussions highlighting the fundamental theories and practical considerations of different mathematical modelling approaches for chemical and biochemical reaction systems.

Machine learning and hybrid modelling have revolutionised the way chemical manufacturing is approached, particularly in the field of reaction engineering. This book is an interdisciplinary resource organised into three parts, each with a unique focus and scope.

Part I, Model Construction Theory, provides fundamental principles and detailed procedures for developing various mathematical models for kinetics simulation. It covers physical models (Chapter 1), data-driven models (Chapter 2), and hybrid models (Chapter 3), along with key aspects for model structure identification (Chapter 4) and uncertainty analysis (Chapter 5).

Part II, Applications in Reaction Engineering, explores a range of chemical and biochemical reaction applications to illustrate the advantages and limitations of different modelling techniques. It is divided into two sections. Section I demonstrates how state-of-the-art machine learning algorithms can automatically propose kinetic rate expressions (Chapter 6) and learn from chemical structure images to predict key structure–property relations (Chapter 7). It also shows how novel process systems engineering strategies can be applied to systematically reduce the complexity of chemical reaction networks (Chapter 8). Section II focuses on different strategies for hybrid model construction. It delves into the impact of domain knowledge on hybrid model accuracy and uncertainty (Chapter 9), explains how transfer learning can be used to accelerate the development of accurate models for novel reaction systems (Chapter 10), and demonstrates how reinforcement learning can be adopted to simulate complex time-varying and history-dependent kinetics (Chapter 11).

Finally, in Part III, Data Intelligence and Industrial Applications, the book delves into the captivating realm of industrial advancements where data intelligence is harnessed to revolutionise chemical and biochemical reaction systems. This section unveils groundbreaking approaches that integrate multiscale modelling (including kinetic modelling and computational fluid dynamics) with machine learning-driven Bayesian optimisation, showcasing their transformative impact on reactor design and optimisation (Chapter 12). Additionally, Chapter 13 provides invaluable insights into the art of designing effective experiments for investigating reaction mechanisms, leveraging the power of state-of-the-art commercial software. Furthermore, the book offers a comprehensive overview of the current landscape of autonomous synthesis and self-optimising reactors in Chapter 14, shedding light on their cutting-edge capabilities. Lastly, Chapter 15 investigates the realm of batch reactor monitoring and fault detection, exploring the utilisation of machine learning algorithms to enhance efficiency and reliability in this critical domain.

This book presents a comprehensive overview of cutting-edge research in reaction engineering, process systems engineering and machine learning. The production of this book is the result of a collective effort, and we are grateful for the contributions of a talented group of individuals. We would like to extend our sincere thanks to Dr Artur M. Schweidtmann, Dr Bovinille Anye Cho, Dr Keju Jing, Dr Wei Wang, Dr Phil Kay, Dr Maria J. Nieves-Remacha, and Dr Francisco J. Navarro-Brull for their invaluable contributions as corresponding or major contributors for different chapters. Additionally, we would like to express our gratitude to Alexander W. Rogers, Fernando Vega Ramon, Max R. Mowbray, and Sam Kay from the University of Manchester, as well as Miguel Angel de Carvalho Servia, Haiting Wang, and Zhaoyan Zhang from Imperial College London, for their tireless efforts in preparing and proofreading different chapters and providing valuable feedback.

Dongda Zhang and Ehecatl Antonio del Río Chanona