Article
Computer Science, Interdisciplinary Applications
Ren Kai Tan, Chao Qian, Michael Wang, Wenjing Ye
Summary: The study proposes a solution to reduce the training cost of artificial-neural-network (ANN)-based surrogate models by reducing the number of numerical simulations during training data generation. The solution utilizes a Mapping Network to map a coarse field to a fine field, generating fine-scale training data.
STRUCTURAL AND MULTIDISCIPLINARY OPTIMIZATION
(2022)
Article
Construction & Building Technology
Haonan Zhang, Haibo Feng, Kasun Hewage, Mehrdad Arashpour
Summary: This study proposed a data-driven framework that combines machine learning, multi-objective optimization, and multi-criteria decision-making techniques to evaluate and formulate optimal retrofit plans for residential buildings.
Article
Green & Sustainable Science & Technology
C. P. Okonkwo, V. I. E. Ajiwe, M. C. Obiadi, M. O. Okwu, J. I. Ayogu
Summary: Biodiesel produced from the seed oil of Chrysobalanus icaco using periwinkle shell ash catalyst was proven to be a good source of renewable and clean fuel. The calcined catalyst showed better performance in continuous reaction runs. The produced biodiesel met the required fuel properties. Rating: 8/10.
JOURNAL OF CLEANER PRODUCTION
(2023)
Article
Mathematics
Yong Wang, Kunzhao Wang, Gaige Wang
Summary: This paper proposes a neural network algorithm with dropout using elite selection to improve the convergence speed of the algorithm. The introduced dropout strategy enhances the optimization performance of the neural network algorithm, making it a powerful algorithm for solving optimization problems.
Article
Engineering, Electrical & Electronic
Wei Zhang, Feng Feng, Jing Jin, Qi-Jun Zhang
Summary: This letter introduces a new surrogate-based multiphysics optimization technique for microwave devices incorporating artificial neural networks (ANNs) and a trust-region algorithm. By using a parallel data generation technique and recalculating the range of the ANN surrogate model at each optimization iteration, the values of design parameters are effectively updated towards the optimal solution, leading to faster convergence.
IEEE MICROWAVE AND WIRELESS COMPONENTS LETTERS
(2021)
Article
Engineering, Civil
Reza Javanmardi, Behrouz Ahmadi-Nedushan
Summary: This paper presents a combined method using optimized neural networks and optimization algorithms to solve structural optimization problems. It trains an optimized artificial neural network (OANN) as a surrogate model to reduce computations for structural analysis. The main optimization problem is solved using the OANN and a population-based algorithm, and then the problem is further solved using the optimal point obtained and the pattern search (PS) algorithm.
FRONTIERS OF STRUCTURAL AND CIVIL ENGINEERING
(2023)
Article
Engineering, Environmental
Sunghyun Cho, Minsu Kim, Byeongil Lyu, Il Moon
Summary: In this study, operating conditions for a novel method of explosive waste disposal were optimized to minimize NOx emissions, utilizing an artificial neural network to create a surrogate model of CFD output for locating optimal conditions. A 34% reduction in NOx emissions from the reactor was achieved using this surrogate model, showing the potential of fluidized bed reactors in reducing NOx emissions.
CHEMICAL ENGINEERING JOURNAL
(2021)
Article
Computer Science, Interdisciplinary Applications
Mahdad Eghbalian, Mehdi Pouragha, Richard Wan
Summary: In this work, a deep neural network architecture called Elasto-Plastic Neural Network (EPNN) is proposed to efficiently surrogate classical elasto-plastic constitutive relations. The EPNN incorporates physics aspects of classical elasto-plasticity, allowing for more efficient training with less data and better extrapolation capability. The architecture is model and material-independent and can be adapted to a wide range of elasto-plastic materials. The superiority of EPNN is demonstrated through predicting strain-controlled loading paths for sands with different initial densities.
COMPUTERS AND GEOTECHNICS
(2023)
Article
Mechanics
Peng Liao, Wei Song, Peng Du, Hang Zhao
Summary: A new aerodynamic shape optimization framework is proposed, which significantly reduces the calculation cost and improves the optimization efficiency by combining high-fidelity and low-fidelity simulations.
Article
Engineering, Chemical
Srinivas Soumitri Miriyala, Keerthi NagaSree Pujari, Sakshi Naik, Kishalay Mitra
Summary: This paper proposes an alternative model using Artificial Neural Networks to optimize the crystallization process, achieving a significant speed improvement while maintaining accuracy through a neural architecture search strategy for hyperparameter tuning.
Article
Computer Science, Interdisciplinary Applications
Hongquan Guo, Jian Zhou, Mohammadreza Koopialipoor, Danial Jahed Armaghani, M. M. Tahir
Summary: This study developed a deep neural network (DNN) model to predict flyrock induced by blasting, which showed a significant increase in prediction accuracy compared to an artificial neural network (ANN) model. The DNN model, optimized using the whale optimization algorithm (WOA), successfully minimized flyrock resulting from blasting and provided a suitable pattern for blasting operations in mines.
ENGINEERING WITH COMPUTERS
(2021)
Article
Automation & Control Systems
Yousaf Ayub, Yusha Hu, Jingzheng Ren, Weifeng Shen, Carman K. M. Lee
Summary: In this research, Aspen Plus simulation and Convolutional Neural Network (CNN) are integrated to estimate the hydrogen content in syngas during poultry litter gasification. The optimized process parameters result in high-quality syngas. Furthermore, CNN exhibits good prediction performance.
ENGINEERING APPLICATIONS OF ARTIFICIAL INTELLIGENCE
(2023)
Review
Environmental Sciences
Jiannan Luo, Xi Ma, Yefei Ji, Xueli Li, Zhuo Song, Wenxi Lu
Summary: This article reviews the application of machine learning-based surrogate models in groundwater contaminant modeling. The article summarizes the state-of-the-art methods, important research challenges, and potential future directions. It is found that machine learning-based surrogate models are widely used in groundwater pollution source identification, remediation design, and uncertainty analysis. The advantages and disadvantages of different methods are analyzed, and method selection recommendations are provided based on the review and experiences. Future research directions include addressing the curse of dimensionality, enhancing transferability, practical applications for real case studies, multi-source data fusion, and real-time monitoring and prediction.
ENVIRONMENTAL RESEARCH
(2023)
Article
Chemistry, Applied
Ga Eun Lee, Ryun Hee Kim, Taehwan Lim, Jaecheol Kim, Suna Kim, Hyoung-Geun Kim, Keum Taek Hwang
Summary: This study analyzed the ellagitannin composition in black raspberry seeds and determined the optimal extraction conditions using artificial neural network and genetic algorithm. The results showed that the maximum total ellagitannin content can be obtained with 63.7% acetone, 4.21 min extraction time, and 43.9 degrees C extraction temperature.
Article
Engineering, Electrical & Electronic
Aysu Belen, Ozlem Tari, Peyman Mahouti, Mehmet A. Belen, Alper Caliskan
Summary: This work studies the design optimization process of a multi-band antenna using artificial neural network (ANN) based surrogate model and meta-heuristic optimizers. The results show that the proposed methodology provides a computationally efficient design optimization process for multi-band antennas.
APPLIED COMPUTATIONAL ELECTROMAGNETICS SOCIETY JOURNAL
(2022)
Review
Chemistry, Multidisciplinary
Daniel S. Wigh, Jonathan M. Goodman, Alexei A. Lapkin
Summary: Interdisciplinary work in chemistry is becoming increasingly important due to advancements in computing, machine learning, and artificial intelligence. Understanding the representation of molecules in a machine-readable format is crucial for computational chemistry. This article introduces different representations of molecules and highlights three significant ones. Researchers often share their work on platforms like GitHub, but discussions on computation time and domain of applicability are often overlooked. The authors propose questions for further consideration to make chemical VAEs more accessible.
WILEY INTERDISCIPLINARY REVIEWS-COMPUTATIONAL MOLECULAR SCIENCE
(2022)
Article
Computer Science, Interdisciplinary Applications
J. Raphael Seidenberg, Ahmad A. Khan, Alexei A. Lapkin
Summary: Conceptual process design involves finding optimal process flowsheets in a large design space. Effective approaches often rely on restricting the search space, which can be done through superstructure optimization or heuristic rules. To enable autonomous process design, it is necessary to formalize knowledge in a machine-readable format. This study proposes incorporating ontological representation of fundamental process knowledge to enhance general-purpose design procedures, while considering problem-specific variability. The framework leverages an ontology to express declarative knowledge and uses a hierarchical reinforcement learning agent to learn procedural knowledge, leading to more efficient and high-quality solutions. The case study on intensified steam methane reforming process demonstrates the benefits of automating domain knowledge in reducing search space and improving computational efficiency and solution quality, highlighting its potential in autonomous process design approaches.
COMPUTERS & CHEMICAL ENGINEERING
(2023)
Article
Energy & Fuels
Nicholas A. A. Jose, Mikhail Kovalev, Alexei A. A. Lapkin
Summary: This study utilizes an annular microreactor to synthesize defect-rich NiCoLDH nanocrystals, and develops a solvent-based method to produce stable LDH electrodes with high capacitance and improved structural stability. The research provides new approaches for electrode development and explores the mechanisms of electron transport in 2D systems.
Article
Engineering, Environmental
Shambhawi, Jana M. Weber, Alexei A. Lapkin
Summary: Designing a simple and representative reaction network is important for cost-effective evaluation and model solvability. Currently, sensitivity analysis is used to screen reaction species in a comprehensive network, but this approach lacks transferability to other catalyst compositions. In this study, a two-way approach is proposed to address this issue. Firstly, a generalizable model outcome is identified based on mass-flow analysis. Then, a stepwise workflow is developed to construct a partial reaction network that ensures transferability across varying catalyst energetics. This approach is demonstrated for CH4 dry reforming using different catalysts.
CHEMICAL ENGINEERING JOURNAL
(2023)
Review
Chemistry, Multidisciplinary
Connor J. Taylor, Alexander Pomberger, Kobi C. Felton, Rachel Grainger, Magda Barecka, Thomas W. Chamberlain, Richard A. Bourne, Christopher N. Johnson, Alexei A. Lapkin
Summary: From the beginning of a synthetic chemist's training, experiments are conducted based on recipes from textbooks and manuscripts. However, it has been shown that model-based, algorithm-based, and miniaturized high-throughput techniques outperform human chemical intuition in understanding chemical systems and achieving reaction optimization. Many synthetic chemists are not exposed to these techniques, leading to a disproportionate number of scientists unable to utilize these methodologies. This review serves as a reference for inspired scientists, highlighting the basics and cutting-edge of chemical reaction optimization and its relation to process scale-up.
Article
Chemistry, Physical
Daniel S. Wigh, Matthieu Tissot, Patrick Pasau, Jonathan M. Goodman, Alexei A. Lapkin
Summary: Computational reaction prediction is a widely used task in chemistry, and in this work, an algorithm for predicting the rate of protodeboronation of boronic acids is presented. The algorithm is based on a mechanistic model derived from kinetic studies and is validated using cross-validation techniques. The algorithm shows promise in assisting chemists in reactions involving boronic acids.
JOURNAL OF PHYSICAL CHEMISTRY A
(2023)
Article
Biochemical Research Methods
Meng Lu, Charles N. Christensen, Jana M. Weber, Tasuku Konno, Nino F. Laubli, Katharina M. Scherer, Edward Avezov, Pietro Lio, Alexei A. Lapkin, Gabriele S. Kaminski Schierle, Clemens F. Kaminski
Summary: ERnet is a deep learning-based software tool that automatically segments and classifies structures in the endoplasmic reticulum, allowing for the quantification of structural changes and the identification of subtle phenotypic differences. It utilizes state-of-the-art semantic segmentation methods and connectivity graphs to accurately and efficiently quantify the connectivity and integrity of ER networks. ERnet has been validated using data from various imaging methods and can be deployed in a high-throughput and unbiased manner to inform on disease progression and response to therapy.
Article
Chemistry, Applied
Jacek Zakrzewski, Polina Yaseneva, Connor J. Taylor, Matthew J. Gaunt, Alexei A. Lapkin
Summary: Development of scalable processes for C(sp3)-H oxidative carbonylation of alkylamines provides convenient access to the beta-lactam scaffol d. The kinetics study shows that the reaction is CO-limited, and there is an optimal CO concentration for the most effective outcome, which leads to an increase in the turnover number in the optimized process. Two scalable processes, a batch process with low catalyst loading and a continuous process using a copper-tube-flow reactor, were developed. The continuous process achieved good results in oxidative carbonylation of several alkylamines without the need for optimization. This study expands the utility of flow chemistry applications to oxidative carbonylations and scalable metal-catalyzed processes.
ORGANIC PROCESS RESEARCH & DEVELOPMENT
(2023)
Article
Chemistry, Multidisciplinary
Connor J. Taylor, Kobi C. Felton, Daniel Wigh, Mohammed I. Jeraal, Rachel Grainger, Gianni Chessari, Christopher N. Johnson, Alexei A. Lapkin
Summary: This study explores the use of multitask Bayesian optimization (MTBO) to accelerate the optimization of new reactions by leveraging reaction data collected from historical optimization campaigns. The methodology was successfully applied in medicinal chemistry applications, demonstrating its effectiveness in accelerating reaction optimization.
ACS CENTRAL SCIENCE
(2023)
Article
Chemistry, Physical
Magda H. Barecka, Pritika D. S. Dameni, Marsha Zakir Muhamad, Joel W. Ager, Alexei A. Lapkin
Summary: Electrosynthesis of ethanol from carbon dioxide (CO2) is a promising route for sustainable fuel and chemical manufacturing. However, the bottleneck of ethanol separation has hindered the scaling of this process. This study presents vacuum membrane distillation as an efficient method to concentrate dilute ethanol streams produced by CO2 electrolysis (CO2R), achieving high ethanol concentrations in pure water. The work also considers thermodynamic properties and proposes strategies for precise estimation of energy inputs to separation processes, supporting the optimization of complex systems for industrial use.
ACS ENERGY LETTERS
(2023)
Article
Chemistry, Multidisciplinary
Adarsh Arun, Zhen Guo, Simon Sung, Alexei A. A. Lapkin
Summary: This study presents an automated impurity prediction workflow based on data mining chemical reaction databases. The workflow accurately evaluates potential chemical reactions between functional groups in the user-supplied query species and extracts reaction templates from analogous reactions to suggest impurities and transformations of interest. Three case studies were conducted, demonstrating the effectiveness of the workflow in suggesting impurities. The interpretable nature of this work makes it a valuable benchmark for more advanced algorithms or models.
Article
Chemistry, Multidisciplinary
Chonghuan Zhang, Alexei A. Lapkin
Summary: Computer-assisted synthesis planning (CASP) accelerates the development of complex functional molecule synthesis routes. This study presents a method that combines conventional organic synthesis and synthetic biological reaction datasets to guide synthesis planning. A hybrid dataset was created by combining organic reactions from the Reaxys & REG; database and metabolic reactions from the KEGG database. Reinforcement learning was used to assemble synthetic pathways from multiple building blocks to a target molecule. The benefits of the hybrid synthetic chemical plus synthetic biological reaction decision space in reaction route optimization were discussed by evaluating and comparing the near-optimal synthetic routes planned from the three reaction pools.
REACTION CHEMISTRY & ENGINEERING
(2023)
Article
Chemistry, Multidisciplinary
Simon D. Rihm, Mikhail K. Kovalev, Alexei A. Lapkin, Joel W. Ager, Markus Kraft
Summary: Utilising CO2 to synthesise chemicals through electrocatalysis has potential for energy storage and decarbonisation. Copper-based electrodes enable high conversion rates, but our incomplete understanding of reaction paths hampers catalyst design. Here, we identify ten new minor products of CO2 reduction and propose two distinct reaction paths based on selectivity trends. This study contributes to the comprehension of electrocatalytic CO2 reduction mechanisms and calls for further exploration of minor products and reaction conditions.
ENERGY & ENVIRONMENTAL SCIENCE
(2023)
Article
Chemistry, Multidisciplinary
Jan G. Rittig, Kobi C. Felton, Alexei A. Lapkin, Alexander Mitsos
Summary: We propose Gibbs-Duhem-informed neural networks for predicting binary activity coefficients at varying compositions. Unlike recent hybrid machine learning approaches, our method does not rely on embedding a specific thermodynamic model and shows improved thermodynamic consistency and generalization capabilities.
Article
Chemistry, Multidisciplinary
Dogancan Karan, Guoying Chen, Nicholas Jose, Jiaru Bai, Paul Mcdaid, Alexei Lapkin
Summary: In this study, a machine learning workflow coupled with a flow chemistry platform was used to optimize the reaction conditions of a lithium-halogen exchange reaction. The results showed that the algorithm successfully identified the optimal conditions in different optimization campaigns and provided insights into the operating regime of the system for different mixing intensifications. Compared to traditional methods, the machine learning workflow proved to be robust and data efficient, providing rich information for the study.
REACTION CHEMISTRY & ENGINEERING
(2023)
