Digital Chemical Engineering最新文献_第3页

Soft-measuring method of iron ore sintering process using transient model 铁矿石烧结过程瞬态模型软测量方法

IF 4.1 Q2 ENGINEERING, CHEMICAL

Digital Chemical Engineering

Pub Date : 2025-09-30 DOI: 10.1016/j.dche.2025.100268

Yoshinari Hashimoto , Satoki Yasuhara , Yuji Iwami

To achieve efficient sintering machine operation in the steel industry, we developed an online soft-measuring method that can visualize the temperature distribution in the sintering process using a two-dimensional (2D) transient model. Although various numerical simulation models of the sintering process have been proposed, the conventional models suffer from estimation errors caused by unmeasurable disturbances, such as the fluctuations in raw material characteristics, when these models are applied for online control in actual plants over a long period. In this study, to reduce the estimation errors, the model parameters were adjusted successively by moving horizon estimation (MHE), considering the effects of the disturbances. The validation results with actual plant data showed that the estimation errors of the burn rising point (BRP) and the exhaust gas compositions were reduced significantly by MHE. In particular, the root mean square error (RMSE) of the BRP estimation was only 1.48 m. In addition, a correlation was confirmed between the estimated high-temperature holding time of the material and the product yield. The developed soft-measuring method is beneficial for process automation to improve product yield.

为了在钢铁行业实现烧结机的高效运行，我们开发了一种在线软测量方法，可以使用二维（2D）瞬态模型可视化烧结过程中的温度分布。虽然已经提出了各种烧结过程的数值模拟模型，但传统模型在长期应用于实际工厂的在线控制时，由于不可测量的干扰（如原料特性的波动）而产生估计误差。为了减小估计误差，在考虑干扰影响的情况下，采用移动地平估计（MHE）对模型参数进行逐次调整。实测数据的验证结果表明，MHE能显著降低燃烧上升点（BRP）和废气成分的估计误差。特别是，BRP估计的均方根误差（RMSE）仅为1.48 m。此外，材料的估计高温保温时间与产品收率之间存在相关性。所开发的软测量方法有利于过程自动化，提高产品成品率。

{"title":"Soft-measuring method of iron ore sintering process using transient model","authors":"Yoshinari Hashimoto , Satoki Yasuhara , Yuji Iwami","doi":"10.1016/j.dche.2025.100268","DOIUrl":"10.1016/j.dche.2025.100268","url":null,"abstract":"<div><div>To achieve efficient sintering machine operation in the steel industry, we developed an online soft-measuring method that can visualize the temperature distribution in the sintering process using a two-dimensional (2D) transient model. Although various numerical simulation models of the sintering process have been proposed, the conventional models suffer from estimation errors caused by unmeasurable disturbances, such as the fluctuations in raw material characteristics, when these models are applied for online control in actual plants over a long period. In this study, to reduce the estimation errors, the model parameters were adjusted successively by moving horizon estimation (MHE), considering the effects of the disturbances. The validation results with actual plant data showed that the estimation errors of the burn rising point (BRP) and the exhaust gas compositions were reduced significantly by MHE. In particular, the root mean square error (RMSE) of the BRP estimation was only 1.48 m. In addition, a correlation was confirmed between the estimated high-temperature holding time of the material and the product yield. The developed soft-measuring method is beneficial for process automation to improve product yield.</div></div>","PeriodicalId":72815,"journal":{"name":"Digital Chemical Engineering","volume":"17 ","pages":"Article 100268"},"PeriodicalIF":4.1,"publicationDate":"2025-09-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145267047","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

CFD and OCT-based optimisation of impeller-induced shear stress on membrane surfaces in a circular test cell 基于CFD和oct的圆形试验池中叶轮引起的膜表面剪切应力优化

IF 4.1 Q2 ENGINEERING, CHEMICAL

Digital Chemical Engineering

Pub Date : 2025-09-25 DOI: 10.1016/j.dche.2025.100267

Masoud Haghshenasfard , Arthur Leon , Robin Starke , Steffi Drescher , Uli Klümper , Thomas Berendonk , Kristin Kerst , André Lerch

This study investigates the distribution of shear stress in a lab-scale membrane bioreactor consisting of a 56 mm-diameter cylindrical test cell, a 0.25 mm-thick polyethersulfone membrane, and a centrally mounted 35 mm rotating impeller. Computational Fluid Dynamics (CFD) simulations were used to examine how impeller speed and geometry affect wall shear stress across the membrane surface. Higher rotational speeds significantly increased shear stress, with the highest levels observed near the impeller rim and a marked decline beyond a radial distance of 0.0175 m due to wall-induced flow dampening. To validate CFD predictions, Optical Coherence Tomography (OCT) was employed for in-situ, real-time biofilm monitoring. OCT results confirmed that low-shear regions—particularly at the membrane periphery—were more prone to rapid and extensive biofilm accumulation, whereas high-shear areas exhibited delayed or reduced fouling. To improve shear distribution and minimize localized fouling, a multi-objective optimization was performed using response surface methodology. This led to an enhanced impeller design that promoted more uniform and effective shear coverage across the membrane. The integration of CFD modeling, experimental validation, and optimization provides a robust framework for the design of membrane systems with improved anti-fouling performance and operational stability.

本研究研究了实验室规模的膜生物反应器中剪切应力的分布，该反应器由直径56 mm的圆柱形试验池、0.25 mm厚的聚醚砜膜和中央安装的35 mm旋转叶轮组成。利用计算流体动力学（CFD）模拟研究了叶轮转速和几何形状对膜表面壁面剪应力的影响。较高的转速显著增加了剪切应力，在叶轮边缘附近观察到的剪切应力最高，在径向距离为0.0175 m时，由于壁面诱导的流动阻尼，剪切应力显著下降。为了验证CFD预测，光学相干断层扫描（OCT）用于现场实时生物膜监测。OCT结果证实，低剪切区域，特别是在膜周围，更容易快速和广泛地积累生物膜，而高剪切区域则表现出延迟或减少的污染。为了改善剪切分布，减少局部污染，采用响应面法进行多目标优化。这导致了叶轮设计的增强，促进了更均匀和有效的剪切覆盖在膜上。CFD建模、实验验证和优化的集成为膜系统的设计提供了一个强大的框架，提高了膜系统的防污性能和运行稳定性。

{"title":"CFD and OCT-based optimisation of impeller-induced shear stress on membrane surfaces in a circular test cell","authors":"Masoud Haghshenasfard , Arthur Leon , Robin Starke , Steffi Drescher , Uli Klümper , Thomas Berendonk , Kristin Kerst , André Lerch","doi":"10.1016/j.dche.2025.100267","DOIUrl":"10.1016/j.dche.2025.100267","url":null,"abstract":"<div><div>This study investigates the distribution of shear stress in a lab-scale membrane bioreactor consisting of a 56 mm-diameter cylindrical test cell, a 0.25 mm-thick polyethersulfone membrane, and a centrally mounted 35 mm rotating impeller. Computational Fluid Dynamics (CFD) simulations were used to examine how impeller speed and geometry affect wall shear stress across the membrane surface. Higher rotational speeds significantly increased shear stress, with the highest levels observed near the impeller rim and a marked decline beyond a radial distance of 0.0175 m due to wall-induced flow dampening. To validate CFD predictions, Optical Coherence Tomography (OCT) was employed for in-situ, real-time biofilm monitoring. OCT results confirmed that low-shear regions—particularly at the membrane periphery—were more prone to rapid and extensive biofilm accumulation, whereas high-shear areas exhibited delayed or reduced fouling. To improve shear distribution and minimize localized fouling, a multi-objective optimization was performed using response surface methodology. This led to an enhanced impeller design that promoted more uniform and effective shear coverage across the membrane. The integration of CFD modeling, experimental validation, and optimization provides a robust framework for the design of membrane systems with improved anti-fouling performance and operational stability.</div></div>","PeriodicalId":72815,"journal":{"name":"Digital Chemical Engineering","volume":"17 ","pages":"Article 100267"},"PeriodicalIF":4.1,"publicationDate":"2025-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145221307","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Integration of sustainability assessment into early-stage carbon capture process design with an explainable AI framework 通过可解释的人工智能框架，将可持续性评估整合到早期碳捕获过程设计中

IF 4.1 Q2 ENGINEERING, CHEMICAL

Digital Chemical Engineering

Pub Date : 2025-09-02 DOI: 10.1016/j.dche.2025.100265

Xin Yee Tai , Oliver Fisher , Lei Xing , Jin Xuan

This study introduces a novel framework for reducing environmental impacts by optimising operating conditions using a surrogate modelling approach integrated with Explainable AI (XAI). Two surrogate models were developed: a sequential surrogate model (SSM) with a two-step structure, and a direct surrogate model (DSM) with a single-step architecture. Both were trained on data from a validated physics-based simulation of a monoethanolamine (MEA)-based carbon capture process to predict environmental impacts across human health, ecosystem quality, and resource depletion. SHapley Additive exPlanations (SHAP) were used to enhance transparency by identifying key input variables influencing outcomes. Multi-objective optimisation was conducted using Particle Swarm Optimisation (PSO) and NSGA-II to determine optimal operating conditions. DSM achieved high prediction accuracy (R² up to 0.995) and lower errors, while SSM offered better interpretability and broader exploration of Pareto-optimal solutions. This study also shows that our framework identified optimum parameters that reduced environmental impacts by 76–88 % compared with the experiment optimum. This framework supports sustainable process design by combining interpretability, predictive performance, and computational efficiency.

本研究引入了一个新的框架，通过使用与可解释人工智能（XAI）集成的代理建模方法来优化操作条件，从而减少对环境的影响。开发了两个代理模型：具有两步结构的顺序代理模型（SSM）和具有单步结构的直接代理模型（DSM）。两者都是根据基于单乙醇胺（MEA）的碳捕获过程的经过验证的物理模拟数据进行训练的，以预测人类健康、生态系统质量和资源枯竭方面的环境影响。SHapley加性解释（SHAP）通过识别影响结果的关键输入变量来提高透明度。采用粒子群算法（PSO）和NSGA-II进行多目标优化，确定最优操作条件。DSM具有较高的预测精度（R²高达0.995）和较低的误差，而SSM具有更好的可解释性和更广泛的探索帕累托最优解。该研究还表明，我们的框架确定的最优参数与实验最优参数相比，减少了76 - 88%的环境影响。该框架通过结合可解释性、预测性能和计算效率来支持可持续的过程设计。

{"title":"Integration of sustainability assessment into early-stage carbon capture process design with an explainable AI framework","authors":"Xin Yee Tai , Oliver Fisher , Lei Xing , Jin Xuan","doi":"10.1016/j.dche.2025.100265","DOIUrl":"10.1016/j.dche.2025.100265","url":null,"abstract":"<div><div>This study introduces a novel framework for reducing environmental impacts by optimising operating conditions using a surrogate modelling approach integrated with Explainable AI (XAI). Two surrogate models were developed: a sequential surrogate model (SSM) with a two-step structure, and a direct surrogate model (DSM) with a single-step architecture. Both were trained on data from a validated physics-based simulation of a monoethanolamine (MEA)-based carbon capture process to predict environmental impacts across human health, ecosystem quality, and resource depletion. SHapley Additive exPlanations (SHAP) were used to enhance transparency by identifying key input variables influencing outcomes. Multi-objective optimisation was conducted using Particle Swarm Optimisation (PSO) and NSGA-II to determine optimal operating conditions. DSM achieved high prediction accuracy (R² up to 0.995) and lower errors, while SSM offered better interpretability and broader exploration of Pareto-optimal solutions. This study also shows that our framework identified optimum parameters that reduced environmental impacts by 76–88 % compared with the experiment optimum. This framework supports sustainable process design by combining interpretability, predictive performance, and computational efficiency.</div></div>","PeriodicalId":72815,"journal":{"name":"Digital Chemical Engineering","volume":"17 ","pages":"Article 100265"},"PeriodicalIF":4.1,"publicationDate":"2025-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145007597","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Editorial: Special issue on pioneering digital chemical engineering 社论：关于开创性数字化学工程的特刊

IF 4.1 Q2 ENGINEERING, CHEMICAL

Digital Chemical Engineering

Pub Date : 2025-09-01 DOI: 10.1016/j.dche.2025.100255

Jin Xuan , Jinfeng Liu

引用次数: 0

Fault detection using multiscale recursive principal component analysis for chemical process systems 基于多尺度递归主成分分析的化工过程系统故障检测

IF 4.1 Q2 ENGINEERING, CHEMICAL

Digital Chemical Engineering

Pub Date : 2025-09-01 DOI: 10.1016/j.dche.2025.100264

Oktorifo Gardiola , Abdulhalim Shah Maulud , Muhammad Nawaz , Nabila Farhana Jamaludin

Process monitoring is essential for maintaining operational safety and product quality in chemical industries. Although conventional fault detection techniques are widely used, their static nature often leads to high false alarm rates (FAR) and missed detection rates (MDR) under dynamic conditions. To address these limitations, this study proposes a Multiscale Recursive Principal Component Analysis (MSRPCA)-based fault detection framework that combines multiscale signal decomposition with the adaptive capabilities of Recursive PCA (RPCA). The MSRPCA approach isolates process variations across different frequency bands while continuously updating the Principal Component Analysis (PCA) model using a moving window mechanism. This enables real-time adaptability and enhanced noise resistance. The proposed method is validated using the Tennessee Eastman Process (TEP), a widely used benchmark for chemical process monitoring under a range of fault types, including step, drift, and random variation disturbances. Fault detection performance is quantitatively assessed using FAR and MDR metrics across 20 distinct fault scenarios. The results demonstrate that MSRPCA consistently outperforms traditional techniques, significantly reducing false alarms while improving fault detection accuracy. For instance, in Fault 16, the MDR in the Hotelling’s T² (T²) chart decreased from 70.5 % (PCA) to 10.5 % (MSRPCA), while the FAR in the Squared Prediction Error (SPE) chart dropped from 21.3 % to 0 %. These findings underscore the robustness and effectiveness of MSRPCA for real-time fault detection in complex, time-varying, and noisy industrial environments.

在化工行业中，过程监控对于维护操作安全和产品质量至关重要。传统的故障检测技术虽然被广泛应用，但其静态特性往往导致在动态条件下的高虚警率和漏检率。为了解决这些局限性，本研究提出了一种基于多尺度递归主成分分析（MSRPCA）的故障检测框架，该框架将多尺度信号分解与递归主成分分析（RPCA）的自适应能力相结合。MSRPCA方法在使用移动窗口机制不断更新主成分分析（PCA）模型的同时，隔离了不同频带的过程变化。这实现了实时适应性和增强的抗噪声能力。采用田纳西伊士曼过程（TEP）验证了所提出的方法，TEP是一种广泛使用的化学过程监测基准，用于一系列故障类型，包括阶跃，漂移和随机变化干扰。在20个不同的故障场景中，使用FAR和MDR度量对故障检测性能进行定量评估。结果表明，MSRPCA持续优于传统技术，在提高故障检测精度的同时显著减少了误报。例如，在故障16中，Hotelling 's T2 （T2）图中的MDR从70.5% （PCA）下降到10.5% (MSRPCA)，而平方预测误差（SPE）图中的FAR从21.3%下降到0%。这些发现强调了MSRPCA在复杂、时变和嘈杂的工业环境中实时故障检测的鲁棒性和有效性。

{"title":"Fault detection using multiscale recursive principal component analysis for chemical process systems","authors":"Oktorifo Gardiola , Abdulhalim Shah Maulud , Muhammad Nawaz , Nabila Farhana Jamaludin","doi":"10.1016/j.dche.2025.100264","DOIUrl":"10.1016/j.dche.2025.100264","url":null,"abstract":"<div><div>Process monitoring is essential for maintaining operational safety and product quality in chemical industries. Although conventional fault detection techniques are widely used, their static nature often leads to high false alarm rates (FAR) and missed detection rates (MDR) under dynamic conditions. To address these limitations, this study proposes a Multiscale Recursive Principal Component Analysis (MSRPCA)-based fault detection framework that combines multiscale signal decomposition with the adaptive capabilities of Recursive PCA (RPCA). The MSRPCA approach isolates process variations across different frequency bands while continuously updating the Principal Component Analysis (PCA) model using a moving window mechanism. This enables real-time adaptability and enhanced noise resistance. The proposed method is validated using the Tennessee Eastman Process (TEP), a widely used benchmark for chemical process monitoring under a range of fault types, including step, drift, and random variation disturbances. Fault detection performance is quantitatively assessed using FAR and MDR metrics across 20 distinct fault scenarios. The results demonstrate that MSRPCA consistently outperforms traditional techniques, significantly reducing false alarms while improving fault detection accuracy. For instance, in Fault 16, the MDR in the Hotelling’s <em>T</em><sup><em>2</em></sup> (<em>T</em><sup><em>2</em></sup>) chart decreased from 70.5 % (PCA) to 10.5 % (MSRPCA), while the FAR in the Squared Prediction Error (SPE) chart dropped from 21.3 % to 0 %. These findings underscore the robustness and effectiveness of MSRPCA for real-time fault detection in complex, time-varying, and noisy industrial environments.</div></div>","PeriodicalId":72815,"journal":{"name":"Digital Chemical Engineering","volume":"16 ","pages":"Article 100264"},"PeriodicalIF":4.1,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144920153","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Editorial: Special issue on Emerging Stars in Digital Chemical Engineering 社论：数字化工新星特刊

IF 4.1 Q2 ENGINEERING, CHEMICAL

Digital Chemical Engineering

Pub Date : 2025-09-01 DOI: 10.1016/j.dche.2025.100247

Jin Xuan , Jinfeng Liu

引用次数: 0

Root cause identification of fault in hot-rolling process by causal plot 用因果图识别热轧过程故障的根本原因

IF 4.1 Q2 ENGINEERING, CHEMICAL

Digital Chemical Engineering

Pub Date : 2025-08-25 DOI: 10.1016/j.dche.2025.100263

Koichi Fujiwara , Yoshiaki Uchida , Taketsugu Osaka

In the steel manufacturing industry, a hot-rolling process produces a thick steel plate from a slab as a batch operation; however, off-spec steel plates are sometimes produced when abnormalities occur during rolling operations. To improve the product yield, it is necessary to appropriately ascertain the root cause of a fault. Because the physicochemical behaviors of the slab during hot-rolling are complicated and yet to be fully understood, we adopted a data-driven approach to identify the cause of the fault in the hot-rolling process. We previously proposed a data-driven fault diagnosis method, referred to as a causal plot, that considers the causal relationships between process variables and monitoring indexes for process monitoring. In the causal plot, monitoring indexes were calculated using existing process monitoring methods, and the causal relationships between the process variables and the calculated monitoring indices were estimated. A linear non-Gaussian acyclic model (LiNGAM) can be adopted for causal inferences between the process variables and calculated monitoring indexes. In this study, we propose a new fault diagnosis method for a batch process, referred to as a b-causal plot, utilizing the causal plot and dynamic time warping (DTW). We analyzed real operation data when defective coils were produced in the hot-rolling process with the proposed b-causal plot and confirmed that the identified root cause was consistent with process engineers’ knowledge, which is typically a low-importance variable that operators do not constantly monitor in daily operation. Because the root cause identification of faults is crucial for maintaining product quality and efficiency in batch processes, the proposed b-causal plot contributes to improving productivity across industries, as demonstrated in this work.

在钢铁制造行业中，热轧工艺将板坯批量生产成厚钢板；然而，当轧制过程中出现异常时，有时会产生不符合规格的钢板。为了提高产品成品率，有必要适当地确定故障的根本原因。由于板坯在热轧过程中的物理化学行为复杂且尚未完全了解，我们采用数据驱动的方法来识别热轧过程中的故障原因。我们之前提出了一种数据驱动的故障诊断方法，称为因果图，该方法考虑过程变量与监控指标之间的因果关系，进行过程监控。在因果图中，利用现有的过程监测方法计算监测指标，并估计过程变量与计算出的监测指标之间的因果关系。采用线性非高斯无循环模型（LiNGAM）对过程变量与计算出的监测指标进行因果推理。在这项研究中，我们提出了一种新的故障诊断方法，称为b-因果图，利用因果图和动态时间规整（DTW）。我们利用提出的b因果图分析了热轧过程中产生缺陷线圈的实际操作数据，并确认确定的根本原因与工艺工程师的知识一致，这通常是操作员在日常操作中不经常监控的低重要性变量。由于故障的根本原因识别对于维持批处理过程中的产品质量和效率至关重要，因此所提出的b-因果图有助于提高各行业的生产率，正如本工作所示。

{"title":"Root cause identification of fault in hot-rolling process by causal plot","authors":"Koichi Fujiwara , Yoshiaki Uchida , Taketsugu Osaka","doi":"10.1016/j.dche.2025.100263","DOIUrl":"10.1016/j.dche.2025.100263","url":null,"abstract":"<div><div>In the steel manufacturing industry, a hot-rolling process produces a thick steel plate from a slab as a batch operation; however, off-spec steel plates are sometimes produced when abnormalities occur during rolling operations. To improve the product yield, it is necessary to appropriately ascertain the root cause of a fault. Because the physicochemical behaviors of the slab during hot-rolling are complicated and yet to be fully understood, we adopted a data-driven approach to identify the cause of the fault in the hot-rolling process. We previously proposed a data-driven fault diagnosis method, referred to as a causal plot, that considers the causal relationships between process variables and monitoring indexes for process monitoring. In the causal plot, monitoring indexes were calculated using existing process monitoring methods, and the causal relationships between the process variables and the calculated monitoring indices were estimated. A linear non-Gaussian acyclic model (LiNGAM) can be adopted for causal inferences between the process variables and calculated monitoring indexes. In this study, we propose a new fault diagnosis method for a batch process, referred to as a b-causal plot, utilizing the causal plot and dynamic time warping (DTW). We analyzed real operation data when defective coils were produced in the hot-rolling process with the proposed b-causal plot and confirmed that the identified root cause was consistent with process engineers’ knowledge, which is typically a low-importance variable that operators do not constantly monitor in daily operation. Because the root cause identification of faults is crucial for maintaining product quality and efficiency in batch processes, the proposed b-causal plot contributes to improving productivity across industries, as demonstrated in this work.</div></div>","PeriodicalId":72815,"journal":{"name":"Digital Chemical Engineering","volume":"17 ","pages":"Article 100263"},"PeriodicalIF":4.1,"publicationDate":"2025-08-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145107110","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Multi-criteria decision support for flexible dividing wall distillation columns 柔性分壁精馏塔多准则决策支持

IF 4.1 Q2 ENGINEERING, CHEMICAL

Digital Chemical Engineering

Pub Date : 2025-08-19 DOI: 10.1016/j.dche.2025.100258

David Mogalle , Patrick Otto Ludl , Tobias Seidel , Lea Trescher , Thomas Grützner , Michael Bortz

When designing a dividing wall column, some decisions regarding the layout of the column cannot be altered once the unit is built, whereas decisions regarding the column’s operation can, to some extent, be adjusted later. During the design phase, both layout and operation can be optimized to achieve an optimal column performance. However, such solutions are tailored to pre-specified process demands. If these demands change later, the physical layout can become suboptimal. Hence, we are interested in design decisions that keep the losses in performance minimal, leading to a column design that is flexible across different demands.

In this paper, we present a new methodology to measure flexibility. The approach is based on multi-criteria optimization, where Pareto fronts with fixed design variables and optimized operating variables are compared to an ideal Pareto front that optimizes both the layout and the operation simultaneously. The difference between two such fronts, representing the losses in performance of the fixed layout for a wide range of demands, is measured by a novel flexibility indicator. We apply our methodology to designing a dividing wall column separating an example mixture. A fast computation of the corresponding Pareto fronts is achieved by solving the arising optimization problems using a reduction method based on stage-to-stage calculations.

当设计分隔墙柱时，有关柱的布局的一些决定一旦建成就不能改变，而有关柱的操作的决定可以在一定程度上稍后调整。在设计阶段，布局和操作都可以优化，以实现最佳的列性能。然而，这样的解决方案是针对预先指定的过程需求量身定制的。如果这些需求后来发生了变化，那么物理布局可能会变得不理想。因此，我们感兴趣的是使性能损失最小化的设计决策，从而使列设计能够灵活地满足不同的需求。在本文中，我们提出了一种新的方法来衡量灵活性。该方法基于多标准优化，将具有固定设计变量和优化操作变量的帕累托前沿与同时优化布局和操作的理想帕累托前沿进行比较。两个这样的前沿之间的差异，代表了固定布局在广泛需求范围内的性能损失，是通过一个新的灵活性指标来衡量的。我们应用我们的方法来设计分离样品混合物的分壁柱。通过使用基于逐级计算的约简方法解决出现的优化问题，实现了相应Pareto前沿的快速计算。

{"title":"Multi-criteria decision support for flexible dividing wall distillation columns","authors":"David Mogalle , Patrick Otto Ludl , Tobias Seidel , Lea Trescher , Thomas Grützner , Michael Bortz","doi":"10.1016/j.dche.2025.100258","DOIUrl":"10.1016/j.dche.2025.100258","url":null,"abstract":"<div><div>When designing a dividing wall column, some decisions regarding the layout of the column cannot be altered once the unit is built, whereas decisions regarding the column’s operation can, to some extent, be adjusted later. During the design phase, both layout and operation can be optimized to achieve an optimal column performance. However, such solutions are tailored to pre-specified process demands. If these demands change later, the physical layout can become suboptimal. Hence, we are interested in design decisions that keep the losses in performance minimal, leading to a column design that is flexible across different demands.</div><div>In this paper, we present a new methodology to measure flexibility. The approach is based on multi-criteria optimization, where Pareto fronts with fixed design variables and optimized operating variables are compared to an ideal Pareto front that optimizes both the layout and the operation simultaneously. The difference between two such fronts, representing the losses in performance of the fixed layout for a wide range of demands, is measured by a novel flexibility indicator. We apply our methodology to designing a dividing wall column separating an example mixture. A fast computation of the corresponding Pareto fronts is achieved by solving the arising optimization problems using a reduction method based on stage-to-stage calculations.</div></div>","PeriodicalId":72815,"journal":{"name":"Digital Chemical Engineering","volume":"16 ","pages":"Article 100258"},"PeriodicalIF":4.1,"publicationDate":"2025-08-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144907779","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Neural network implementation of model predictive control with stability guarantees 神经网络实现具有稳定性保证的模型预测控制

IF 4.1 Q2 ENGINEERING, CHEMICAL

Digital Chemical Engineering

Pub Date : 2025-08-18 DOI: 10.1016/j.dche.2025.100262

Arthur Khodaverdian , Dhruv Gohil , Panagiotis D. Christofides

This work explores the use of supervised learning on data generated by a model predictive controller (MPC) to train a neural network (NN). The goal is to create an approximate control policy that can replace the MPC, offering reduced computational complexity while maintaining stability guarantees. Through the use of Lyapunov-based stability constraints, an MPC can be designed to guarantee stability. Once designed, this MPC can be used to generate a dataset of various state-space points and their resulting immediate optimal control actions. With the MPC dataset representing an optimal control policy, an NN is trained to function as a direct substitute for the MPC. The resulting approximate control policy can then be applied in real-time to the process, with stability guarantees being enforced through post-inference validation. If, for a given set of sensor readings, the NN yields control actions that violate the Lyapunov stability constraints used in the MPC, the control action is discarded and replaced with stabilizing control from a fallback stabilizing controller. This control architecture is applied to a benchmark chemical reactor model. Using this model, a comprehensive study of the stability, performance, robustness, and computational burden of the approach is carried out.

这项工作探索了对模型预测控制器（MPC）生成的数据使用监督学习来训练神经网络（NN）。目标是创建一个可以取代MPC的近似控制策略，在保持稳定性保证的同时降低计算复杂度。通过使用基于lyapunov的稳定性约束，可以设计出保证稳定性的MPC。一旦设计好，这个MPC可以用来生成各种状态空间点的数据集，以及它们产生的即时最优控制行为。使用MPC数据集表示最优控制策略，训练神经网络作为MPC的直接替代品。由此产生的近似控制策略可以实时应用于过程，并通过推理后验证强制执行稳定性保证。如果对于给定的一组传感器读数，神经网络产生的控制动作违反了MPC中使用的李雅普诺夫稳定性约束，则控制动作被丢弃，并由一个回降稳定控制器的稳定控制取代。将该控制体系结构应用于一个基准化学反应器模型。利用该模型，对该方法的稳定性、性能、鲁棒性和计算量进行了全面的研究。

{"title":"Neural network implementation of model predictive control with stability guarantees","authors":"Arthur Khodaverdian , Dhruv Gohil , Panagiotis D. Christofides","doi":"10.1016/j.dche.2025.100262","DOIUrl":"10.1016/j.dche.2025.100262","url":null,"abstract":"<div><div>This work explores the use of supervised learning on data generated by a model predictive controller (MPC) to train a neural network (NN). The goal is to create an approximate control policy that can replace the MPC, offering reduced computational complexity while maintaining stability guarantees. Through the use of Lyapunov-based stability constraints, an MPC can be designed to guarantee stability. Once designed, this MPC can be used to generate a dataset of various state-space points and their resulting immediate optimal control actions. With the MPC dataset representing an optimal control policy, an NN is trained to function as a direct substitute for the MPC. The resulting approximate control policy can then be applied in real-time to the process, with stability guarantees being enforced through post-inference validation. If, for a given set of sensor readings, the NN yields control actions that violate the Lyapunov stability constraints used in the MPC, the control action is discarded and replaced with stabilizing control from a fallback stabilizing controller. This control architecture is applied to a benchmark chemical reactor model. Using this model, a comprehensive study of the stability, performance, robustness, and computational burden of the approach is carried out.</div></div>","PeriodicalId":72815,"journal":{"name":"Digital Chemical Engineering","volume":"16 ","pages":"Article 100262"},"PeriodicalIF":4.1,"publicationDate":"2025-08-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144865569","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

AutoRL framework for bioprocess control: Optimizing reward function, architecture, and hyperparameters 生物过程控制的AutoRL框架：优化奖励函数、结构和超参数

IF 4.1 Q2 ENGINEERING, CHEMICAL

Digital Chemical Engineering

Pub Date : 2025-08-14 DOI: 10.1016/j.dche.2025.100261

D.A. Goulart , R.D. Pereira , F.V. Silva

This study proposes a structured AutoRL framework for the development of deep reinforcement learning (DRL) controllers in chemical process systems. Focusing on the control of a 3

\times

3 MIMO yeast fermentation bioreactor, the methodology jointly optimizes three key internal components of the DRL agent: the reward function, the neural network architecture, and the hyperparameters of the algorithm. A parameterizable logistic reward formulation is introduced to encode control objectives, such as steady-state accuracy, minimalization of actuation effort, and control smoothness, into a flexible and tunable structure. A dual loop optimization strategy combines grid search and Bayesian optimization to systematically explore and refine the agent’s design space. The resulting controller achieved average steady-state errors of 0.009 °C for reactor temperature and 0.19 g/L for ethanol concentration, while maintaining smooth and stable behavior under diverse operational scenarios. By formalizing reward design and integrating it with hyperparameter and architecture optimization, this work delivers a AutoRL methodology for DRL-based control, reducing reliance on expert heuristics and enhancing reproducibility in complex bioprocess applications.

本研究提出了一个结构化的AutoRL框架，用于开发化学过程系统中的深度强化学习（DRL）控制器。该方法以3x3 MIMO酵母发酵生物反应器的控制为重点，对DRL agent的三个关键内部组件：奖励函数、神经网络架构和算法的超参数进行了联合优化。引入了一个参数化的逻辑奖励公式，将控制目标（如稳态精度、驱动努力最小化和控制平滑度）编码为一个灵活可调的结构。采用网格搜索和贝叶斯优化相结合的双环优化策略，系统地探索和细化智能体的设计空间。该控制器在反应器温度和乙醇浓度的平均稳态误差分别为0.009°C和0.19 g/L，同时在各种操作场景下保持平稳稳定的行为。通过将奖励设计形式化并将其与超参数和架构优化相结合，本研究为基于drl的控制提供了一种AutoRL方法，减少了对专家启发式的依赖，并提高了复杂生物过程应用的可重复性。

{"title":"AutoRL framework for bioprocess control: Optimizing reward function, architecture, and hyperparameters","authors":"D.A. Goulart , R.D. Pereira , F.V. Silva","doi":"10.1016/j.dche.2025.100261","DOIUrl":"10.1016/j.dche.2025.100261","url":null,"abstract":"<div><div>This study proposes a structured AutoRL framework for the development of deep reinforcement learning (DRL) controllers in chemical process systems. Focusing on the control of a 3<span><math><mo>×</mo></math></span> 3 MIMO yeast fermentation bioreactor, the methodology jointly optimizes three key internal components of the DRL agent: the reward function, the neural network architecture, and the hyperparameters of the algorithm. A parameterizable logistic reward formulation is introduced to encode control objectives, such as steady-state accuracy, minimalization of actuation effort, and control smoothness, into a flexible and tunable structure. A dual loop optimization strategy combines grid search and Bayesian optimization to systematically explore and refine the agent’s design space. The resulting controller achieved average steady-state errors of 0.009 °C for reactor temperature and 0.19 g/L for ethanol concentration, while maintaining smooth and stable behavior under diverse operational scenarios. By formalizing reward design and integrating it with hyperparameter and architecture optimization, this work delivers a AutoRL methodology for DRL-based control, reducing reliance on expert heuristics and enhancing reproducibility in complex bioprocess applications.</div></div>","PeriodicalId":72815,"journal":{"name":"Digital Chemical Engineering","volume":"16 ","pages":"Article 100261"},"PeriodicalIF":4.1,"publicationDate":"2025-08-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144895879","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0