Pub Date : 2025-02-16DOI: 10.1016/j.compchemeng.2025.109065
Ting Wu , Peilin Zhan , Wei Chen , Miaoqing Lin , Quanyuan Qiu , Yinan Hu , Jiuhang Song , Xiaoqing Lin
Deep eutectic solvents (DESs) are sustainable alternatives to traditional solvents, but their structural complexity makes accurate prediction of melting points and densities challenging. This study utilizes ChemBERTa, a pre-trained Transformer model, to extract high-dimensional embeddings from Simplified Molecular Input Line Entry System (SMILES) strings, effectively capturing complex molecular interactions and subtle structural features. Through feature importance analysis, we identified missing molecular information in the ChemBERTa embeddings and supplemented it with select physicochemical descriptors from RDKit, creating a feature set that enhances both interpretability and predictive accuracy. Optimized ensemble models, including ExtraTreesRegressor (ETR) and XGBRegressor (XGBR), are then applied to this enriched feature set, achieving notable improvements in prediction accuracy for DES melting point and density. Rigorous grid search and ten-fold cross-validation ensure model robustness and generalizability. Experimental results confirm the effectiveness of this approach, underscoring the transformative role of pre-trained deep learning models in chemical informatics and supporting scalable, sustainable DESs design.
{"title":"ChemBERTa embeddings and ensemble learning for prediction of density and melting point of deep eutectic solvents with hybrid features","authors":"Ting Wu , Peilin Zhan , Wei Chen , Miaoqing Lin , Quanyuan Qiu , Yinan Hu , Jiuhang Song , Xiaoqing Lin","doi":"10.1016/j.compchemeng.2025.109065","DOIUrl":"10.1016/j.compchemeng.2025.109065","url":null,"abstract":"<div><div>Deep eutectic solvents (DESs) are sustainable alternatives to traditional solvents, but their structural complexity makes accurate prediction of melting points and densities challenging. This study utilizes ChemBERTa, a pre-trained Transformer model, to extract high-dimensional embeddings from Simplified Molecular Input Line Entry System (SMILES) strings, effectively capturing complex molecular interactions and subtle structural features. Through feature importance analysis, we identified missing molecular information in the ChemBERTa embeddings and supplemented it with select physicochemical descriptors from RDKit, creating a feature set that enhances both interpretability and predictive accuracy. Optimized ensemble models, including ExtraTreesRegressor (ETR) and XGBRegressor (XGBR), are then applied to this enriched feature set, achieving notable improvements in prediction accuracy for DES melting point and density. Rigorous grid search and ten-fold cross-validation ensure model robustness and generalizability. Experimental results confirm the effectiveness of this approach, underscoring the transformative role of pre-trained deep learning models in chemical informatics and supporting scalable, sustainable DESs design.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"196 ","pages":"Article 109065"},"PeriodicalIF":3.9,"publicationDate":"2025-02-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143429981","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-15DOI: 10.1016/j.compchemeng.2025.109042
Felix Birkelbach
For including piecewise linear (PWL) functions in MILP problems, the logarithmic convex combination (Log) formulation has been shown to yield very fast solving times. However, identifying approximations that can be used with Log is a big challenge since the approximation has to be compatible with a J1 triangulation. In this article, an algorithm is proposed that identifies approximations using J1 compatible triangulations. It seeks to satisfy the specified error tolerance with the minimum number of linear pieces, so that the MILP formulation is small. To evaluate the performance of the J1 approach it is applied to two sets of benchmark functions from literature and results are compared to state-of-the-art approaches.
Overall the J1 approach is shown to efficiently approximate functions in up to 3 dimensions. Especially for tight error tolerances, these J1 approximations require fewer auxiliary variables in MILP compared to alternative approaches.
{"title":"Piecewise linear approximation using J1 compatible triangulations for efficient MILP representation","authors":"Felix Birkelbach","doi":"10.1016/j.compchemeng.2025.109042","DOIUrl":"10.1016/j.compchemeng.2025.109042","url":null,"abstract":"<div><div>For including piecewise linear (PWL) functions in MILP problems, the logarithmic convex combination (Log) formulation has been shown to yield very fast solving times. However, identifying approximations that can be used with Log is a big challenge since the approximation has to be compatible with a J1 triangulation. In this article, an algorithm is proposed that identifies approximations using J1 compatible triangulations. It seeks to satisfy the specified error tolerance with the minimum number of linear pieces, so that the MILP formulation is small. To evaluate the performance of the J1 approach it is applied to two sets of benchmark functions from literature and results are compared to state-of-the-art approaches.</div><div>Overall the J1 approach is shown to efficiently approximate functions in up to 3 dimensions. Especially for tight error tolerances, these J1 approximations require fewer auxiliary variables in MILP compared to alternative approaches.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"196 ","pages":"Article 109042"},"PeriodicalIF":3.9,"publicationDate":"2025-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143429979","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-15DOI: 10.1016/j.compchemeng.2025.109037
Ashley Dan , Urjit Patil , Abhinav De , Bhavani Nandhini Mummidi Manuraj , Rohit Ramachandran
The development of mathematical models for physical systems often necessitates the use of high-dimensional spaces and fine discretizations to accurately capture complex dynamics. These models, which involve large matrices and extensive mathematical operations, tend to be computationally intensive, leading to slow execution times. In this study, we analyzed various acceleration strategies by comparing the simulation accuracy, computational time, and resource utilization of various vectorization and parallelization methods on both CPUs and GPUs, using a multi-dimensional Population Balance Model simulated in MATLAB and Python. Our findings revealed that GPU-based vectorization provided the highest performance, achieving a 40-fold speedup compared to the serial implementations. Unlike simulations on CPUs, where run time is often limited by processing power, GPUs simulations are limited by the available memory, especially at high resolution. This work highlights the importance of using appropriate resources and code optimization strategies to reduce computational time, for development of an efficient model.
{"title":"CPU and GPU based acceleration of high-dimensional population balance models via the vectorization and parallelization of multivariate aggregation and breakage integral terms","authors":"Ashley Dan , Urjit Patil , Abhinav De , Bhavani Nandhini Mummidi Manuraj , Rohit Ramachandran","doi":"10.1016/j.compchemeng.2025.109037","DOIUrl":"10.1016/j.compchemeng.2025.109037","url":null,"abstract":"<div><div>The development of mathematical models for physical systems often necessitates the use of high-dimensional spaces and fine discretizations to accurately capture complex dynamics. These models, which involve large matrices and extensive mathematical operations, tend to be computationally intensive, leading to slow execution times. In this study, we analyzed various acceleration strategies by comparing the simulation accuracy, computational time, and resource utilization of various vectorization and parallelization methods on both CPUs and GPUs, using a multi-dimensional Population Balance Model simulated in MATLAB and Python. Our findings revealed that GPU-based vectorization provided the highest performance, achieving a 40-fold speedup compared to the serial implementations. Unlike simulations on CPUs, where run time is often limited by processing power, GPUs simulations are limited by the available memory, especially at high resolution. This work highlights the importance of using appropriate resources and code optimization strategies to reduce computational time, for development of an efficient model.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"196 ","pages":"Article 109037"},"PeriodicalIF":3.9,"publicationDate":"2025-02-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143422192","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-14DOI: 10.1016/j.compchemeng.2025.109060
Sumit K. Bishnu , Sabla Y. Alnouri , Dhabia M. Al Mohannadi
There is a need for comprehensive tools that combine data-driven modeling with optimization techniques. In this work, a robust Random Forest Regression (RFR) model was developed to capture the behavior and characteristics of a Sorption Enhanced Steam Methane Reformer (SE-SMR) Reactor system. This model was then integrated into a Simulated Annealing (SA) optimization framework that helped identify the optimal operating conditions for the unit. The combined approach demonstrates the potential of using machine learning models in conjunction with optimization techniques to improve the solving process. The proposed methodology achieved an optimal methane conversion rate of 0.99979, and was successful in effectively identifying the optimal operating conditions that were required for near-complete conversion.
{"title":"Stochastic algorithm-based optimization using artificial intelligence/machine learning models for sorption enhanced steam methane reformer reactor","authors":"Sumit K. Bishnu , Sabla Y. Alnouri , Dhabia M. Al Mohannadi","doi":"10.1016/j.compchemeng.2025.109060","DOIUrl":"10.1016/j.compchemeng.2025.109060","url":null,"abstract":"<div><div>There is a need for comprehensive tools that combine data-driven modeling with optimization techniques. In this work, a robust Random Forest Regression (RFR) model was developed to capture the behavior and characteristics of a Sorption Enhanced Steam Methane Reformer (SE-SMR) Reactor system. This model was then integrated into a Simulated Annealing (SA) optimization framework that helped identify the optimal operating conditions for the unit. The combined approach demonstrates the potential of using machine learning models in conjunction with optimization techniques to improve the solving process. The proposed methodology achieved an optimal methane conversion rate of 0.99979, and was successful in effectively identifying the optimal operating conditions that were required for near-complete conversion.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"196 ","pages":"Article 109060"},"PeriodicalIF":3.9,"publicationDate":"2025-02-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143429978","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-11DOI: 10.1016/j.compchemeng.2025.109028
Feiya Lv , Borui Yang , Shujian Yu , Shengwu Zou , Xiaolin Wang , Jinsong Zhao , Chenglin Wen
Reasoning about cause and effect in industrial processes is fundamental to fault diagnosis. However, traditional methods for causal discovery and fault diagnosis are typically developed separately, resulting in complex and fragmented approaches that lack transparency and interpretability. Since the explicit identification of root causes from causal graphs remains an open issue, we propose a unified diagnosis model for chemical processes that integrates causal discovery, fault detection, and root cause diagnosis within a single framework. Granger causality is learned from monitoring time-series data for online predictions. This causal embedding ensures that prediction deviations occur only in variables causally linked to the root cause, effectively mitigating the ’smearing effect’ caused by unrelated variables. The explicit causal graph provides interpretive insights into fault propagation and enhances the traceability of the diagnostic process by enabling the identification of fault evolution paths and root causes. Experimental results on synthetic data, a continuously stirred-tank reactor (CSTR) process, and a real-world continuous catalytic reforming (CCR) process demonstrate that our approach achieves high diagnostic accuracy and low false alarm rates, offering a practical, interpretable, and scalable solution for fault diagnosis in industrial chemical processes.
{"title":"A unified model integrating Granger causality-based causal discovery and fault diagnosis in chemical processes","authors":"Feiya Lv , Borui Yang , Shujian Yu , Shengwu Zou , Xiaolin Wang , Jinsong Zhao , Chenglin Wen","doi":"10.1016/j.compchemeng.2025.109028","DOIUrl":"10.1016/j.compchemeng.2025.109028","url":null,"abstract":"<div><div>Reasoning about cause and effect in industrial processes is fundamental to fault diagnosis. However, traditional methods for causal discovery and fault diagnosis are typically developed separately, resulting in complex and fragmented approaches that lack transparency and interpretability. Since the explicit identification of root causes from causal graphs remains an open issue, we propose a unified diagnosis model for chemical processes that integrates causal discovery, fault detection, and root cause diagnosis within a single framework. Granger causality is learned from monitoring time-series data for online predictions. This causal embedding ensures that prediction deviations occur only in variables causally linked to the root cause, effectively mitigating the ’smearing effect’ caused by unrelated variables. The explicit causal graph provides interpretive insights into fault propagation and enhances the traceability of the diagnostic process by enabling the identification of fault evolution paths and root causes. Experimental results on synthetic data, a continuously stirred-tank reactor (CSTR) process, and a real-world continuous catalytic reforming (CCR) process demonstrate that our approach achieves high diagnostic accuracy and low false alarm rates, offering a practical, interpretable, and scalable solution for fault diagnosis in industrial chemical processes.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"196 ","pages":"Article 109028"},"PeriodicalIF":3.9,"publicationDate":"2025-02-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143429980","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-10DOI: 10.1016/j.compchemeng.2025.109049
Olakunle Ayodeji Omole , Chukwuma C. Ogbaga , Jude A. Okolie , Olugbenga Akande , Richard Kimera , Joseph Lepnaan Dayil
This paper examines machine learning (ML)'s contemporary applications in biofuel production, emphasizing microalgae-based bioenergy systems. The study aims to explore various aspects of ML integration in the biofuel production process, including microalgae detection, classification, growth phase optimization, and dataset quality and quantity considerations. The research methodology is in a detailed literature review of current ML models and their applications in biofuel production. It covers bioenergy systems, microalgae detection, growth phase optimization, dataset quality, ML applications in microalgal biorefineries, and the advantages and disadvantages of ML models over first-principle models. The analysis highlights the challenges and implications of utilizing smaller datasets in biofuel production models and investigates the impact of dataset quality and quantity on ML model performance. Despite sparse datasets, the findings offer insights into leveraging ML techniques for improved efficiency and sustainability in microalgae-based biofuel production systems.
{"title":"Advancing algal biofuel production through data-driven insights: A comprehensive review of machine learning applications","authors":"Olakunle Ayodeji Omole , Chukwuma C. Ogbaga , Jude A. Okolie , Olugbenga Akande , Richard Kimera , Joseph Lepnaan Dayil","doi":"10.1016/j.compchemeng.2025.109049","DOIUrl":"10.1016/j.compchemeng.2025.109049","url":null,"abstract":"<div><div>This paper examines machine learning (ML)'s contemporary applications in biofuel production, emphasizing microalgae-based bioenergy systems. The study aims to explore various aspects of ML integration in the biofuel production process, including microalgae detection, classification, growth phase optimization, and dataset quality and quantity considerations. The research methodology is in a detailed literature review of current ML models and their applications in biofuel production. It covers bioenergy systems, microalgae detection, growth phase optimization, dataset quality, ML applications in microalgal biorefineries, and the advantages and disadvantages of ML models over first-principle models. The analysis highlights the challenges and implications of utilizing smaller datasets in biofuel production models and investigates the impact of dataset quality and quantity on ML model performance. Despite sparse datasets, the findings offer insights into leveraging ML techniques for improved efficiency and sustainability in microalgae-based biofuel production systems.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"196 ","pages":"Article 109049"},"PeriodicalIF":3.9,"publicationDate":"2025-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143437320","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-10DOI: 10.1016/j.compchemeng.2025.109045
Lea Trescher , Lena-Marie Ränger , David Mogalle , Tobias Seidel , Michael Bortz , Thomas Grützner
Introductory studies on the relationship between design and robustness of optimized Dividing Wall Columns are presented. These columns are optimized multicriterially for different mixtures and numbers of stages. The internal distribution of stages depending on the total number is analyzed. Supporting Material analyzes the stage-dependent vapor demand of the binary sub-systems and shows that the relationships can change fundamentally, especially for low stages. The deviations in process variables considered are defined and screening results are presented. Supporting Material provides an analysis of the nominal and disturbed concentration profiles for deviating internal splits, and shows that at high numbers of stages, the formation of pinch zones leads to significantly higher purity losses. Finally, results for combined deviations are presented. For all result parts, patterns are worked out that depend only on the number of stages, specific mixture properties like characteristics of the Vmin diagram or the individual product.
{"title":"Pareto-optimized Dividing Wall Columns for Ideal Mixtures and Influences of Deviations in Process Variables","authors":"Lea Trescher , Lena-Marie Ränger , David Mogalle , Tobias Seidel , Michael Bortz , Thomas Grützner","doi":"10.1016/j.compchemeng.2025.109045","DOIUrl":"10.1016/j.compchemeng.2025.109045","url":null,"abstract":"<div><div>Introductory studies on the relationship between design and robustness of optimized Dividing Wall Columns are presented. These columns are optimized multicriterially for different mixtures and numbers of stages. The internal distribution of stages depending on the total number is analyzed. Supporting Material analyzes the stage-dependent vapor demand of the binary sub-systems and shows that the relationships can change fundamentally, especially for low stages. The deviations in process variables considered are defined and screening results are presented. Supporting Material provides an analysis of the nominal and disturbed concentration profiles for deviating internal splits, and shows that at high numbers of stages, the formation of pinch zones leads to significantly higher purity losses. Finally, results for combined deviations are presented. For all result parts, patterns are worked out that depend only on the number of stages, specific mixture properties like characteristics of the <em>V<sub>min</sub></em> diagram or the individual product.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"196 ","pages":"Article 109045"},"PeriodicalIF":3.9,"publicationDate":"2025-02-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143402547","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-08DOI: 10.1016/j.compchemeng.2025.109041
Tayfun Eylen, P. Erhan Eren, Altan Koçyiğit
Most manufacturing sector businesses utilize advanced control mechanisms to sustain their ongoing operations. An alarm management system is one of these control mechanisms that works as a safety barrier, and it contains alarm messages indicating abnormal situations to operators. The causes of alarms mainly result in a harmful state of operations that should be eliminated as quickly as possible to minimize possible negative results. However, the size of the system, lack of people directing the system, and process-dependent peak conditions may lead operators to miss some critical alarms. Quality and quantity of products, job safety, and operational costs are some of the features negatively affected by these missing alarms. The proposed work aims to combine a well-established alarm management philosophy with advanced data analytics techniques to optimize decision variables in alarm management processes. This study introduces a novel data-driven optimization method that leverages the Tennessee Eastman Process as a benchmark to validate its effectiveness. The proposed method aims to ensure continuous alarm system health by contributing to the automation of the parameter optimization process in the life cycles of alarm management systems. Key contributions include the development of a method to associate disturbances with alarms, the creation of an alarm simulation platform, and the improvement of alarm parameters through a unique optimization approach. The results show that there is a trade-off between alarm reaction delay, which refers to the time between disturbances and the first relevant alarm and number of alarms and alarm on times. This trade-off can be evaluated in the desired direction by taking into account the priorities of the process.
{"title":"Data-driven alarm parameter optimization","authors":"Tayfun Eylen, P. Erhan Eren, Altan Koçyiğit","doi":"10.1016/j.compchemeng.2025.109041","DOIUrl":"10.1016/j.compchemeng.2025.109041","url":null,"abstract":"<div><div>Most manufacturing sector businesses utilize advanced control mechanisms to sustain their ongoing operations. An alarm management system is one of these control mechanisms that works as a safety barrier, and it contains alarm messages indicating abnormal situations to operators. The causes of alarms mainly result in a harmful state of operations that should be eliminated as quickly as possible to minimize possible negative results. However, the size of the system, lack of people directing the system, and process-dependent peak conditions may lead operators to miss some critical alarms. Quality and quantity of products, job safety, and operational costs are some of the features negatively affected by these missing alarms. The proposed work aims to combine a well-established alarm management philosophy with advanced data analytics techniques to optimize decision variables in alarm management processes. This study introduces a novel data-driven optimization method that leverages the Tennessee Eastman Process as a benchmark to validate its effectiveness. The proposed method aims to ensure continuous alarm system health by contributing to the automation of the parameter optimization process in the life cycles of alarm management systems. Key contributions include the development of a method to associate disturbances with alarms, the creation of an alarm simulation platform, and the improvement of alarm parameters through a unique optimization approach. The results show that there is a trade-off between alarm reaction delay, which refers to the time between disturbances and the first relevant alarm and number of alarms and alarm on times. This trade-off can be evaluated in the desired direction by taking into account the priorities of the process.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"196 ","pages":"Article 109041"},"PeriodicalIF":3.9,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143422194","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-08DOI: 10.1016/j.compchemeng.2025.109039
Lavinia Marina Paola Ghilardi , Sakshi Naik , Emanuele Martelli , Francesco Casella , Lorenz T. Biegler
This study presents an economic Nonlinear Model Predictive Control for optimizing gas pipeline operation. The operation of gas networks is governed by dynamic gas transport equations, compressor performance characteristics, and control valve modeling. Given the daily fluctuations in demand, these systems often do not operate under steady-state conditions. To address this, we propose a controller formulation designed for cyclic steady-state systems, incorporating stabilizing and terminal constraints to ensure asymptotic stability. The application of this approach to real-world, complex branched pipelines involves dealing with non-smoothness and switching conditions, which we tackle through smoothing and complementarity reformulations. The effectiveness of our method is demonstrated in a test network as well as the nationwide Italian gas network, showcasing its practicality for large-scale applications.
{"title":"Economic Nonlinear Model Predictive Control for cyclic gas pipeline operation","authors":"Lavinia Marina Paola Ghilardi , Sakshi Naik , Emanuele Martelli , Francesco Casella , Lorenz T. Biegler","doi":"10.1016/j.compchemeng.2025.109039","DOIUrl":"10.1016/j.compchemeng.2025.109039","url":null,"abstract":"<div><div>This study presents an economic Nonlinear Model Predictive Control for optimizing gas pipeline operation. The operation of gas networks is governed by dynamic gas transport equations, compressor performance characteristics, and control valve modeling. Given the daily fluctuations in demand, these systems often do not operate under steady-state conditions. To address this, we propose a controller formulation designed for cyclic steady-state systems, incorporating stabilizing and terminal constraints to ensure asymptotic stability. The application of this approach to real-world, complex branched pipelines involves dealing with non-smoothness and switching conditions, which we tackle through smoothing and complementarity reformulations. The effectiveness of our method is demonstrated in a test network as well as the nationwide Italian gas network, showcasing its practicality for large-scale applications.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"196 ","pages":"Article 109039"},"PeriodicalIF":3.9,"publicationDate":"2025-02-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143422049","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-02-06DOI: 10.1016/j.compchemeng.2025.109031
Boneshwar V K , Deepesh Agarwal , Bala Natarajan , Babji Srinivasan
Single-cell RNA sequencing (scRNA-seq) has transformed biological research, enabling detailed analysis of disease pathways, cellular differentiation, and immune responses at a cellular level. However, the noisy and sparse nature of scRNA-seq datasets often impedes accurate downstream analyses. Cell clustering and gene imputation serve as foundational tasks in harnessing scRNA-seq data for complex biological insights. While various graph-based methods have been developed to enhance imputation and clustering accuracy, traditional transductive models require entire graphs during training, limiting computational efficiency on large biological networks. This study introduces a novel inductive framework that efficiently learns relationships among graph nodes by utilizing subgraphs rather than full neighbor sets for node embedding generation, significantly reducing computational demands while maintaining robust performance. The proposed model achieves up to 60% improvement in Silhouette score, 14.9% in Adjusted Rand Index, 48% in runtime, and 4.5% in L Median error over baseline models, validating the effectiveness of inductive graph learning. Evaluated on diverse scRNA-seq datasets—GSE75748 (progenitor cell types derived from human embryonic stem cells (hESCs)), GSE131928 (adult and pediatric IDH-wildtype glioblastomas (GBM)), and Goolam et al (blastomeres from early-stage Mus musculus (mouse) embryos collected at the 2-cell, 4-cell, 8-cell, 16-cell, and 32-cell stages of preimplantation development).—this framework demonstrates scalability and adaptability, offering a reliable approach for future applications in trajectory inference and gene pathway analysis.
{"title":"Inductive graph neural network framework for imputation of single-cell RNA sequencing data","authors":"Boneshwar V K , Deepesh Agarwal , Bala Natarajan , Babji Srinivasan","doi":"10.1016/j.compchemeng.2025.109031","DOIUrl":"10.1016/j.compchemeng.2025.109031","url":null,"abstract":"<div><div>Single-cell RNA sequencing (scRNA-seq) has transformed biological research, enabling detailed analysis of disease pathways, cellular differentiation, and immune responses at a cellular level. However, the noisy and sparse nature of scRNA-seq datasets often impedes accurate downstream analyses. Cell clustering and gene imputation serve as foundational tasks in harnessing scRNA-seq data for complex biological insights. While various graph-based methods have been developed to enhance imputation and clustering accuracy, traditional transductive models require entire graphs during training, limiting computational efficiency on large biological networks. This study introduces a novel inductive framework that efficiently learns relationships among graph nodes by utilizing subgraphs rather than full neighbor sets for node embedding generation, significantly reducing computational demands while maintaining robust performance. The proposed model achieves up to 60% improvement in Silhouette score, 14.9% in Adjusted Rand Index, 48% in runtime, and 4.5% in L<span><math><msub><mrow></mrow><mrow><mn>1</mn></mrow></msub></math></span> Median error over baseline models, validating the effectiveness of inductive graph learning. Evaluated on diverse scRNA-seq datasets—GSE75748 (progenitor cell types derived from human embryonic stem cells (hESCs)), GSE131928 (adult and pediatric IDH-wildtype glioblastomas (GBM)), and Goolam et al (blastomeres from early-stage Mus musculus (mouse) embryos collected at the 2-cell, 4-cell, 8-cell, 16-cell, and 32-cell stages of preimplantation development).—this framework demonstrates scalability and adaptability, offering a reliable approach for future applications in trajectory inference and gene pathway analysis.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"195 ","pages":"Article 109031"},"PeriodicalIF":3.9,"publicationDate":"2025-02-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143379076","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号