Computers & Chemical Engineering最新文献_第3页

Dynamic simulation and operational optimization of dividing wall batch distillation columns based on rigorous models 基于严格模型的分壁间歇精馏塔动态仿真及操作优化

IF 3.9 2区工程技术 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Chemical Engineering

Pub Date : 2025-12-24 DOI: 10.1016/j.compchemeng.2025.109542

Xing Qian , Jiayi Du , Gonghan Guo , Shengkun Jia , Chao Zhang

Batch distillation, as a widely applied separation technology, demonstrates significant advantages in small-scale, multi-component, and composition-variable separation tasks. Current research primarily focuses on two directions to enhance its overall performance, reduce energy consumption, and shorten separation time: development of novel process configurations and operational optimization. Among these, the dividing wall batch distillation column (DWBDC) remarkably improves separation efficiency of conventional batch distillation through its innovative batch dividing wall process intensification structure. However, as a highly nonlinear dynamic system, DWBDC presents considerable challenges in rigorous simulation and optimization, which is well worth further investigation. Bayesian optimization algorithms have gained extensive application in chemical engineering due to their efficient global search capability in recent years. This study innovatively applies Bayesian optimization to batch distillation processes, establishing an operational optimization framework based on this algorithm and implementing it in the optimal design of various batch distillation configurations. To systematically evaluate DWBDC's superiority, rigorous dynamic simulations were conducted on three configurations: conventional batch distillation column (BDC), middle vessel batch distillation column (MVBDC), and DWBDC, with corresponding concentration control systems developed. The research examined separation performances of these configurations for industrial waste solvents with different feed compositions (light-component-dominant, intermediate-component-dominant, and balanced light/intermediate-component feeds). Bayesian optimization was employed to optimize operational parameters for all three configurations, followed by comparative performance analysis under optimal conditions. Results demonstrate the superior performance of DWBDC in both operating time and economic efficiency, providing crucial theoretical foundations and practical guidance for industrial application of dividing wall batch distillation technology.

间歇精馏作为一种广泛应用的分离技术，在小规模、多组分、多组分的分离任务中具有显著的优势。目前的研究主要集中在两个方向，以提高其整体性能，降低能耗，缩短分离时间：开发新的工艺配置和操作优化。其中，分壁间歇精馏塔（DWBDC）通过其创新的间歇分壁过程强化结构，显著提高了常规间歇精馏的分离效率。然而，作为一个高度非线性的动态系统，DWBDC在严格的仿真和优化方面面临着相当大的挑战，值得进一步研究。近年来，贝叶斯优化算法以其高效的全局搜索能力在化工领域得到了广泛的应用。本研究创新性地将贝叶斯优化应用于间歇精馏过程，建立了基于该算法的操作优化框架，并将其应用于间歇精馏的各种配置优化设计中。为了系统地评价DWBDC的优越性，对常规间歇精馏塔（BDC）、中间容器间歇精馏塔（MVBDC）和DWBDC三种配置进行了严格的动态仿真，并开发了相应的浓度控制系统。研究考察了这些构型对不同原料组成（轻组分为主、中间组分为主、轻/中间组分平衡）的工业废溶剂的分离性能。采用贝叶斯优化方法对三种构型的运行参数进行优化，并在最优条件下进行性能对比分析。结果表明，DWBDC在操作时间和经济效率方面均具有优越的性能，为分壁间歇精馏技术的工业应用提供了重要的理论基础和实践指导。

{"title":"Dynamic simulation and operational optimization of dividing wall batch distillation columns based on rigorous models","authors":"Xing Qian , Jiayi Du , Gonghan Guo , Shengkun Jia , Chao Zhang","doi":"10.1016/j.compchemeng.2025.109542","DOIUrl":"10.1016/j.compchemeng.2025.109542","url":null,"abstract":"<div><div>Batch distillation, as a widely applied separation technology, demonstrates significant advantages in small-scale, multi-component, and composition-variable separation tasks. Current research primarily focuses on two directions to enhance its overall performance, reduce energy consumption, and shorten separation time: development of novel process configurations and operational optimization. Among these, the dividing wall batch distillation column (DWBDC) remarkably improves separation efficiency of conventional batch distillation through its innovative batch dividing wall process intensification structure. However, as a highly nonlinear dynamic system, DWBDC presents considerable challenges in rigorous simulation and optimization, which is well worth further investigation. Bayesian optimization algorithms have gained extensive application in chemical engineering due to their efficient global search capability in recent years. This study innovatively applies Bayesian optimization to batch distillation processes, establishing an operational optimization framework based on this algorithm and implementing it in the optimal design of various batch distillation configurations. To systematically evaluate DWBDC's superiority, rigorous dynamic simulations were conducted on three configurations: conventional batch distillation column (BDC), middle vessel batch distillation column (MVBDC), and DWBDC, with corresponding concentration control systems developed. The research examined separation performances of these configurations for industrial waste solvents with different feed compositions (light-component-dominant, intermediate-component-dominant, and balanced light/intermediate-component feeds). Bayesian optimization was employed to optimize operational parameters for all three configurations, followed by comparative performance analysis under optimal conditions. Results demonstrate the superior performance of DWBDC in both operating time and economic efficiency, providing crucial theoretical foundations and practical guidance for industrial application of dividing wall batch distillation technology.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"206 ","pages":"Article 109542"},"PeriodicalIF":3.9,"publicationDate":"2025-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145880239","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}

引用次数: 0

Kinetic modelling of the CO2 capture and utilisation on NiRu-Ca/Al dual function material via parameter estimation 基于参数估计的nru - ca /Al双功能材料CO2捕集利用动力学模型

IF 3.9 2区工程技术 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Chemical Engineering

Pub Date : 2025-12-22 DOI: 10.1016/j.compchemeng.2025.109537

Meshkat Dolat , Andrew David Wright , Soudabeh Bahrami Gharamaleki , Loukia-Pantzechroula Merkouri , Melis S. Duyar , Michael Short

This study presents a detailed, open-source kinetic modelling computational framework for CO₂ capture and utilisation using a newly formulated dual-function material (DFM) comprising 15 wt% Ni, 1 wt% Ru, and 10 wt% CaO supported on spherical alumina. A finite difference reactor model was developed to simulate the cyclic adsorption, purge, and hydrogenation stages. The model incorporates experimentally-derived rate expressions, accounts for system delay via a second-order response function, and was fitted to time-resolved concentration laboratory data using Bayesian optimisation. The robustness of the estimated parameters was rigorously assessed using Profile Likelihood Analysis (PLA), which confirmed the practical identifiability of the rate-limiting hydrogenation steps while statistically validating the masking effect of system delays on rapid adsorption kinetics. A combined parameter estimation strategy was employed to ensure mass continuity across stages and improve the robustness of purge kinetics. The kinetic parameters extracted reveal that carbonate decomposition, not methanation, is the rate-limiting step during hydrogenation. Temperature-dependent simulations confirm a trade-off between reaction kinetics and CO₂ storage capacity, with methane yield maximised at 300 °C when compared with the other temperature sets. By offering transparent methodology and reproducible code, this work provides a robust platform for researchers and practitioners to study, validate, and optimise DFM systems.

本研究提出了一个详细的、开源的二氧化碳捕获和利用动力学建模计算框架，使用一种新配方的双功能材料（DFM），该材料由15 wt% Ni、1 wt% Ru和10 wt% CaO支撑在球形氧化铝上。建立了一个有限差分反应器模型来模拟循环吸附、吹扫和加氢阶段。该模型结合了实验推导的速率表达式，通过二阶响应函数解释了系统延迟，并使用贝叶斯优化方法拟合了时间分辨浓度实验室数据。使用轮廓似然分析（PLA）对估计参数的鲁棒性进行了严格评估，这证实了限速加氢步骤的实际可识别性，同时统计验证了系统延迟对快速吸附动力学的掩盖效应。采用了一种组合参数估计策略来保证各阶段的质量连续性，提高吹扫动力学的鲁棒性。提取的动力学参数表明，碳酸盐分解，而不是甲烷化，是加氢过程中的限速步骤。温度相关的模拟证实了反应动力学和CO₂储存能力之间的权衡，与其他温度设置相比，甲烷产量在300°C时最大。通过提供透明的方法和可重复的代码，这项工作为研究人员和从业者提供了一个强大的平台来研究、验证和优化DFM系统。

{"title":"Kinetic modelling of the CO2 capture and utilisation on NiRu-Ca/Al dual function material via parameter estimation","authors":"Meshkat Dolat , Andrew David Wright , Soudabeh Bahrami Gharamaleki , Loukia-Pantzechroula Merkouri , Melis S. Duyar , Michael Short","doi":"10.1016/j.compchemeng.2025.109537","DOIUrl":"10.1016/j.compchemeng.2025.109537","url":null,"abstract":"<div><div>This study presents a detailed, open-source kinetic modelling computational framework for CO₂ capture and utilisation using a newly formulated dual-function material (DFM) comprising 15 wt% Ni, 1 wt% Ru, and 10 wt% CaO supported on spherical alumina. A finite difference reactor model was developed to simulate the cyclic adsorption, purge, and hydrogenation stages. The model incorporates experimentally-derived rate expressions, accounts for system delay via a second-order response function, and was fitted to time-resolved concentration laboratory data using Bayesian optimisation. The robustness of the estimated parameters was rigorously assessed using Profile Likelihood Analysis (PLA), which confirmed the practical identifiability of the rate-limiting hydrogenation steps while statistically validating the masking effect of system delays on rapid adsorption kinetics. A combined parameter estimation strategy was employed to ensure mass continuity across stages and improve the robustness of purge kinetics. The kinetic parameters extracted reveal that carbonate decomposition, not methanation, is the rate-limiting step during hydrogenation. Temperature-dependent simulations confirm a trade-off between reaction kinetics and CO₂ storage capacity, with methane yield maximised at 300 °C when compared with the other temperature sets. By offering transparent methodology and reproducible code, this work provides a robust platform for researchers and practitioners to study, validate, and optimise DFM systems.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"206 ","pages":"Article 109537"},"PeriodicalIF":3.9,"publicationDate":"2025-12-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145836405","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}

引用次数: 0

On the impact of mechanistic model quality and data availability in hybrid model development 混合模型开发中机制模型质量和数据可用性的影响

IF 3.9 2区工程技术 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Chemical Engineering

Pub Date : 2025-12-20 DOI: 10.1016/j.compchemeng.2025.109536

Margherita Geremia , Tiziana Marella , Elaheh Ardalani , Samira Beyramysoltan , Sayantan Chattoraj , Pierantonio Facco , Massimiliano Barolo , Fabrizio Bezzo

This work presents a methodology for evaluating the effectiveness of hybrid modelling under varying conditions of mechanistic model quality and information available for model training, that is typically expressed as the amount of data available. While hybrid models – which integrate mechanistic and data-driven components – have gained significant attention in process systems engineering, their advantages over purely mechanistic or data-driven alternatives remain inadequately quantified. We address this gap by investigating two critical factors: (i) the impact of mechanistic model fidelity on hybrid model performance, and (ii) the influence of calibration dataset size on prediction accuracy. Our methodology is validated through an in-silico case study of baker's yeast cultivation and a real-world industrial application of ion-exchange chromatography in biopharmaceutical manufacturing. Results demonstrate that hybrid models consistently outperform purely mechanistic and data-driven approaches when the mechanistic component captures fundamental process behaviours, even with structural simplifications. Notably, hybrid models maintain superior predictive capability in extrapolative scenarios; however, when mechanistic knowledge is severely limited and insufficient information is available for compensation, hybridisation benefits diminish substantially. The work provides quantitative guidance for practitioners to determine when hybrid modelling represents a justified investment of modelling resources in process engineering applications.

这项工作提出了一种方法，用于评估混合建模在不同条件下的有效性，这些条件下的机械模型质量和模型训练可用的信息，通常表示为可用的数据量。虽然混合模型-集成了机械和数据驱动的组件-在过程系统工程中得到了极大的关注，但它们相对于纯机械或数据驱动的替代方案的优势仍然没有充分量化。我们通过研究两个关键因素来解决这一差距：(i)机制模型保真度对混合模型性能的影响，以及（ii）校准数据集大小对预测精度的影响。我们的方法通过烘焙酵母培养的计算机案例研究和离子交换色谱在生物制药制造中的实际工业应用得到验证。结果表明，当机械组件捕获基本过程行为时，混合模型始终优于纯机械和数据驱动的方法，即使结构简化了。值得注意的是，混合模型在外推情景中保持了优越的预测能力；然而，当机械知识严重有限，并且可供补偿的信息不足时，杂交的好处就会大大减少。这项工作为实践者提供了定量指导，以确定混合建模何时代表过程工程应用中建模资源的合理投资。

{"title":"On the impact of mechanistic model quality and data availability in hybrid model development","authors":"Margherita Geremia , Tiziana Marella , Elaheh Ardalani , Samira Beyramysoltan , Sayantan Chattoraj , Pierantonio Facco , Massimiliano Barolo , Fabrizio Bezzo","doi":"10.1016/j.compchemeng.2025.109536","DOIUrl":"10.1016/j.compchemeng.2025.109536","url":null,"abstract":"<div><div>This work presents a methodology for evaluating the effectiveness of hybrid modelling under varying conditions of mechanistic model quality and information available for model training, that is typically expressed as the amount of data available. While hybrid models – which integrate mechanistic and data-driven components – have gained significant attention in process systems engineering, their advantages over purely mechanistic or data-driven alternatives remain inadequately quantified. We address this gap by investigating two critical factors: (<em>i</em>) the impact of mechanistic model fidelity on hybrid model performance, and (<em>ii</em>) the influence of calibration dataset size on prediction accuracy. Our methodology is validated through an <em>in-silico</em> case study of baker's yeast cultivation and a real-world industrial application of ion-exchange chromatography in biopharmaceutical manufacturing. Results demonstrate that hybrid models consistently outperform purely mechanistic and data-driven approaches when the mechanistic component captures fundamental process behaviours, even with structural simplifications. Notably, hybrid models maintain superior predictive capability in extrapolative scenarios; however, when mechanistic knowledge is severely limited and insufficient information is available for compensation, hybridisation benefits diminish substantially. The work provides quantitative guidance for practitioners to determine when hybrid modelling represents a justified investment of modelling resources in process engineering applications.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"206 ","pages":"Article 109536"},"PeriodicalIF":3.9,"publicationDate":"2025-12-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145836404","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}

引用次数: 0

Uncertainty-aware design optimization of hybrid membrane–cryogenic nitrogen rejection units (NRU): decoupling energy and product quality 复合式膜-低温脱氮装置的不确定性优化设计：解耦能量与产品质量

IF 3.9 2区工程技术 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Chemical Engineering

Pub Date : 2025-12-19 DOI: 10.1016/j.compchemeng.2025.109533

Norfamila Che Mat , Edelin Emirra Empir , Kasih Syazwina Qistina Syaheezam , Muhammad Abdul Qyyum

Hybrid membrane–cryogenic nitrogen rejection units (NRUs) are increasingly proposed for upgrading sub-quality natural gas (NG), yet the mechanistic basis for their performance benefits remains insufficiently understood. This work develops a surrogate-assisted multi-objective optimization framework that integrates perturbation-expansion membrane modelling with Super Learner ensemble surrogates and analytical Bayesian uncertainty quantification to enable computationally efficient hybrid system design under data-sparse conditions. The optimization explores a 14-dimensional design space, yielding Pareto-optimal trade-offs with CH₄ recovery spanning 0.80–0.97 and specific power consumption (SPC) of 0.32–0.52 kWh/kg CH₄. Permutation-based sensitivity analysis identifies functional decoupling between system domains: membrane parameters predominantly govern SPC, while cryogenic column variables control CH₄ recovery, with minimal cross-influence. These results challenge the common assumption that membrane pre-concentration directly reduces energy demand. Instead, optimal hybrid configurations achieve >93% recovery at 0.49–0.51 kWh/kg CH₄—comparable to standalone cryogenic performance—demonstrating that hybrid value arises primarily from operational flexibility, enabling independent manipulation of product quality and energy consumption. Analytical Bayesian inference achieves 24–30% reduction in predictive uncertainty with cross-validation consistency <0.06. The framework produces performance–confidence trade-off maps across the design space, supporting systematic selection of operating strategies based on specified confidence thresholds. By reducing optimization time from 23 to 5 h while quantifying prediction reliability, the proposed approach offers an alternative to empirical tuning practices and provides clearer visibility into performance–flexibility interactions for hybrid NRU design.

混合膜-低温脱氮装置（nru）越来越多地被提出用于改善亚质量天然气（NG），但其性能优势的机制基础仍未得到充分了解。这项工作开发了一个代理辅助的多目标优化框架，该框架将微扰扩展膜建模与超级学习者集成代理和分析贝叶斯不确定性量化相结合，以实现数据稀疏条件下计算效率高的混合系统设计。优化探索了一个14维的设计空间，得到了帕累托最优权衡，硫酸铵回收率为0.80-0.97，比功耗（SPC）为0.32-0.52 kWh/kg氯化铵。基于置换的敏感性分析确定了系统域之间的功能解耦：膜参数主要控制SPC，而低温柱变量控制CH₄回收率，交叉影响最小。这些结果挑战了膜预浓缩直接降低能量需求的普遍假设。相反，最佳混合配置在0.49-0.51 kWh/kg CH₄下可实现>；93%的回收率，与单独的低温性能相当，这表明混合配置的价值主要来自操作灵活性，能够独立操纵产品质量和能耗。分析贝叶斯推理预测不确定性降低24-30%，交叉验证一致性<；0.06。该框架生成跨设计空间的性能-置信度权衡图，支持基于指定置信度阈值的操作策略的系统选择。通过将优化时间从23小时减少到5小时，同时量化预测可靠性，该方法为经验调优实践提供了替代方案，并为混合NRU设计提供了更清晰的性能-灵活性交互可见性。

{"title":"Uncertainty-aware design optimization of hybrid membrane–cryogenic nitrogen rejection units (NRU): decoupling energy and product quality","authors":"Norfamila Che Mat , Edelin Emirra Empir , Kasih Syazwina Qistina Syaheezam , Muhammad Abdul Qyyum","doi":"10.1016/j.compchemeng.2025.109533","DOIUrl":"10.1016/j.compchemeng.2025.109533","url":null,"abstract":"<div><div>Hybrid membrane–cryogenic nitrogen rejection units (NRUs) are increasingly proposed for upgrading sub-quality natural gas (NG), yet the mechanistic basis for their performance benefits remains insufficiently understood. This work develops a surrogate-assisted multi-objective optimization framework that integrates perturbation-expansion membrane modelling with Super Learner ensemble surrogates and analytical Bayesian uncertainty quantification to enable computationally efficient hybrid system design under data-sparse conditions. The optimization explores a 14-dimensional design space, yielding Pareto-optimal trade-offs with CH₄ recovery spanning 0.80–0.97 and specific power consumption (SPC) of 0.32–0.52 kWh/kg CH₄. Permutation-based sensitivity analysis identifies functional decoupling between system domains: membrane parameters predominantly govern SPC, while cryogenic column variables control CH₄ recovery, with minimal cross-influence. These results challenge the common assumption that membrane pre-concentration directly reduces energy demand. Instead, optimal hybrid configurations achieve >93% recovery at 0.49–0.51 kWh/kg CH₄—comparable to standalone cryogenic performance—demonstrating that hybrid value arises primarily from operational flexibility, enabling independent manipulation of product quality and energy consumption. Analytical Bayesian inference achieves 24–30% reduction in predictive uncertainty with cross-validation consistency <0.06. The framework produces performance–confidence trade-off maps across the design space, supporting systematic selection of operating strategies based on specified confidence thresholds. By reducing optimization time from 23 to 5 h while quantifying prediction reliability, the proposed approach offers an alternative to empirical tuning practices and provides clearer visibility into performance–flexibility interactions for hybrid NRU design.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"206 ","pages":"Article 109533"},"PeriodicalIF":3.9,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145836368","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}

引用次数: 0

Comparative performance of stochastic and deep learning models in forecasting crude oil prices under market shocks 随机和深度学习模型在市场冲击下原油价格预测中的比较性能

IF 3.9 2区工程技术 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Chemical Engineering

Pub Date : 2025-12-19 DOI: 10.1016/j.compchemeng.2025.109534

Busra Agan Celik, Murat Yulek, Serdar Celik

This study provides a comprehensive evaluation of crude oil price forecasting by systematically comparing stochastic models (GBM, FBM, BB), machine learning and time-series approaches (LSTM, ARIMA, VAR), and two hybrid strategies LSTM-GBM (time-varying drift estimation) and Bridge-ARIMA (terminal-level anchoring) across short-, medium-, and long-term horizons. Using daily price data spanning 34 years (1990–2024), the analysis encompasses major global crises, including the Asian Financial Crisis, Dot-com Bubble, Global Commodity Boom, Global Financial Crisis, Arab Spring, Oil Price Collapse, COVID-19 pandemic, and the Russia–Ukraine conflict. Stochastic models exhibited sensitivity to the number of simulation runs, with FBM yielding the most robust forecasts, BB demonstrating stability through mean-reversion, and GBM showing larger errors over extended horizons. LSTM consistently outperformed alternative methods, capturing nonlinear dynamics and long-term dependencies with high accuracy across all horizons. ARIMA performed adequately for short-term forecasts but declined over longer periods, whereas VAR consistently underperformed, reflecting its limited capacity to capture complex market dynamics. The hybrid methods further enhance performance: Bridge-ARIMA reduces bias by aligning simulated paths with an ARIMA-implied terminal, improving crisis-period tracking, whereas LSTM-GBM leverages learned drift to refine path forecasts and risk metrics (e.g., VaR). These results underscore the complementary advantages of stochastic and deep learning approaches and suggest that hybrid frameworks can enhance both volatility assessment and trend prediction. The findings offer practical insights for energy market planning, risk management, and strategic investment, particularly under crisis-driven volatility.

本研究通过系统地比较随机模型（GBM、FBM、BB）、机器学习和时间序列方法（LSTM、ARIMA、VAR），以及两种混合策略LSTM-GBM（时变漂移估计）和Bridge-ARIMA（终端级锚定），对原油价格预测进行了全面评估。该报告利用34年（1990年至2024年）的每日价格数据，分析了亚洲金融危机、互联网泡沫、全球大宗商品繁荣、全球金融危机、阿拉伯之春、油价暴跌、COVID-19大流行、俄乌冲突等重大全球危机。随机模型显示出对模拟运行次数的敏感性，其中FBM产生最稳健的预测，BB通过均值回归显示出稳定性，而GBM在较长时间内显示出较大的误差。LSTM始终优于其他方法，在所有视界以高精度捕获非线性动态和长期依赖关系。ARIMA在短期预测中表现良好，但在长期预测中表现下降，而VAR的表现一直不佳，反映出其捕捉复杂市场动态的能力有限。混合方法进一步提高了性能：Bridge-ARIMA通过将模拟路径与arima隐含的终端对齐来减少偏差，改善危机期跟踪，而LSTM-GBM利用学习漂移来优化路径预测和风险指标（如VaR）。这些结果强调了随机和深度学习方法的互补优势，并表明混合框架可以增强波动性评估和趋势预测。研究结果为能源市场规划、风险管理和战略投资提供了实用的见解，特别是在危机驱动的波动下。

{"title":"Comparative performance of stochastic and deep learning models in forecasting crude oil prices under market shocks","authors":"Busra Agan Celik, Murat Yulek, Serdar Celik","doi":"10.1016/j.compchemeng.2025.109534","DOIUrl":"10.1016/j.compchemeng.2025.109534","url":null,"abstract":"<div><div>This study provides a comprehensive evaluation of crude oil price forecasting by systematically comparing stochastic models (GBM, FBM, BB), machine learning and time-series approaches (LSTM, ARIMA, VAR), and two hybrid strategies LSTM-GBM (time-varying drift estimation) and Bridge-ARIMA (terminal-level anchoring) across short-, medium-, and long-term horizons. Using daily price data spanning 34 years (1990–2024), the analysis encompasses major global crises, including the Asian Financial Crisis, Dot-com Bubble, Global Commodity Boom, Global Financial Crisis, Arab Spring, Oil Price Collapse, COVID-19 pandemic, and the Russia–Ukraine conflict. Stochastic models exhibited sensitivity to the number of simulation runs, with FBM yielding the most robust forecasts, BB demonstrating stability through mean-reversion, and GBM showing larger errors over extended horizons. LSTM consistently outperformed alternative methods, capturing nonlinear dynamics and long-term dependencies with high accuracy across all horizons. ARIMA performed adequately for short-term forecasts but declined over longer periods, whereas VAR consistently underperformed, reflecting its limited capacity to capture complex market dynamics. The hybrid methods further enhance performance: Bridge-ARIMA reduces bias by aligning simulated paths with an ARIMA-implied terminal, improving crisis-period tracking, whereas LSTM-GBM leverages learned drift to refine path forecasts and risk metrics (e.g., VaR). These results underscore the complementary advantages of stochastic and deep learning approaches and suggest that hybrid frameworks can enhance both volatility assessment and trend prediction. The findings offer practical insights for energy market planning, risk management, and strategic investment, particularly under crisis-driven volatility.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"206 ","pages":"Article 109534"},"PeriodicalIF":3.9,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145836366","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}

引用次数: 0

Domain-enforced and operator-in-the-loop neural simulation platform for techno-enviro-economic performance enhancement of gas turbine system 面向燃气轮机系统技术环境经济性能提升的领域强制和操作者在环神经仿真平台

IF 3.9 2区工程技术 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Chemical Engineering

Pub Date : 2025-12-16 DOI: 10.1016/j.compchemeng.2025.109529

Waqar Muhammad Ashraf , Abdulelah S. Alshehri , Abdulrahman bin Jumah , Ghulam Moeen Uddin , Muhammad Akhtar , Vivek Dua

The domain-consistent and operator-centric artificial intelligence (AI) adoption has remained slow in the industrial operation of power systems, including gas power plants. This paper presents a domain-enforced and operator-in-the-loop neural simulation platform that is built upon embedding the neural surrogate models in the nonlinear optimisation framework and is implemented to analyse the operation of 395 MW capacity gas turbine system. Feed forward architecture-based neural network models like Artificial Neural Network (ANN), Data Information integrated Neural Network (DINN) and Kolmogorov-Arnold Networks (KAN) are trained to predict performance variables of gas turbine system (Power-MW, Turbine Heat Rate-kJ/kWh, Thermal Efficiency-%). KAN achieved slightly higher predictive performance on test dataset (R² ≥ 0.96) better than those of ANN (R² ≥ 0.92) and DINN (R² ≥ 0.93). Mahalanobis distance-based constraint introduces operator-in-the-loop and enforces the data-driven domain for estimating domain-consistent and energy-efficient optimised operating levels to produce a set value of power from gas turbine system. The failure modes of operation of the open-source neural simulation platform are also discussed to guide operators in estimating domain-consistent optimal operating levels. The domain-enforced neural simulation platform can reduce 2.9 kton/y [0.3 kton/y, 5.4 kton/y] of CO₂ emissions and may cut the annual operating cost of $ 0.95 m [$ 0.3 m, $ 1.6 m] from the operation of gas turbine system. We anticipate that the developed AI-powered simulation platform may adapt to the dynamic industrial power generation environment and enhance the access of AI and optimisation tools for data-informed decision-making for industrial applications.

在包括燃气电厂在内的电力系统的工业运行中，领域一致和以运营商为中心的人工智能（AI）的采用仍然缓慢。本文在非线性优化框架中嵌入神经代理模型的基础上，建立了一个域强制和操作者在环神经仿真平台，并对395 MW燃气轮机系统的运行进行了分析。基于前馈结构的神经网络模型，如人工神经网络（ANN）、数据信息集成神经网络（DINN）和Kolmogorov-Arnold网络（KAN），被训练来预测燃气轮机系统的性能变量（功率- mw、涡轮热率- kj /kWh、热效率-%）。KAN在测试数据集上的预测性能略高于ANN （R2≥0.92）和DINN （R2≥0.93）。Mahalanobis基于距离的约束引入了操作员在环，并加强了数据驱动域，用于估计域一致和节能优化的操作水平，以产生燃气轮机系统的功率设定值。讨论了开源神经仿真平台的运行故障模式，指导操作者估计域一致的最优运行水平。该领域强制神经仿真平台可减少2.9千吨/年[0.3千吨/年，5.4千吨/年]的二氧化碳排放量，并可从燃气轮机系统的运行中减少每年95万美元[30万美元，160万美元]的运行成本。我们预计开发的人工智能仿真平台可以适应动态的工业发电环境，并增强人工智能和优化工具的访问，为工业应用提供数据知情决策。

{"title":"Domain-enforced and operator-in-the-loop neural simulation platform for techno-enviro-economic performance enhancement of gas turbine system","authors":"Waqar Muhammad Ashraf , Abdulelah S. Alshehri , Abdulrahman bin Jumah , Ghulam Moeen Uddin , Muhammad Akhtar , Vivek Dua","doi":"10.1016/j.compchemeng.2025.109529","DOIUrl":"10.1016/j.compchemeng.2025.109529","url":null,"abstract":"<div><div>The domain-consistent and operator-centric artificial intelligence (AI) adoption has remained slow in the industrial operation of power systems, including gas power plants. This paper presents a domain-enforced and operator-in-the-loop neural simulation platform that is built upon embedding the neural surrogate models in the nonlinear optimisation framework and is implemented to analyse the operation of 395 MW capacity gas turbine system. Feed forward architecture-based neural network models like Artificial Neural Network (ANN), Data Information integrated Neural Network (DINN) and Kolmogorov-Arnold Networks (KAN) are trained to predict performance variables of gas turbine system (Power-MW, Turbine Heat Rate-kJ/kWh, Thermal Efficiency-%). KAN achieved slightly higher predictive performance on test dataset (R<sup>2</sup> ≥ 0.96) better than those of ANN (R<sup>2</sup> ≥ 0.92) and DINN (R<sup>2</sup> ≥ 0.93). Mahalanobis distance-based constraint introduces operator-in-the-loop and enforces the data-driven domain for estimating domain-consistent and energy-efficient optimised operating levels to produce a set value of power from gas turbine system. The failure modes of operation of the open-source neural simulation platform are also discussed to guide operators in estimating domain-consistent optimal operating levels. The domain-enforced neural simulation platform can reduce 2.9 kton/y [0.3 kton/y, 5.4 kton/y] of CO<sub>2</sub> emissions and may cut the annual operating cost of $ 0.95 m [$ 0.3 m, $ 1.6 m] from the operation of gas turbine system. We anticipate that the developed AI-powered simulation platform may adapt to the dynamic industrial power generation environment and enhance the access of AI and optimisation tools for data-informed decision-making for industrial applications.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"206 ","pages":"Article 109529"},"PeriodicalIF":3.9,"publicationDate":"2025-12-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145797270","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}

引用次数: 0

Intelligent fault diagnosis in hybrid chemical processes under limited samples based on multi-feature fusion learning 基于多特征融合学习的有限样本混合化工过程故障智能诊断

IF 3.9 2区工程技术 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Chemical Engineering

Pub Date : 2025-12-15 DOI: 10.1016/j.compchemeng.2025.109531

Min Yin , Youqing Wang , Xin Ma , Yining Dong

In recent years, intelligent fault diagnosis—particularly the feasibility of handling hybrid variables—has garnered increasing research attention due to the growing complexity of chemical processes. In many industrial settings, hybrid variables that include both continuous and discrete elements are frequently observed, reflecting the complexity of modern process systems. However, the scarcity of fault samples in such systems has led to the emergence of the few-shot learning problem. Insights from continuous-variable systems suggest that information augmentation is an effective strategy for addressing this issue. To this end, this study proposes a novel information augmentation approach based on Multi-Feature Fusion Networks (MFNets). Specifically, numerical, trend, and manipulation features are extracted from hybrid data using sliding time windows, the Gramian Angular Field (GAF) algorithm, and Gaussian blur techniques, respectively. These multi-view features are then integrated through a shared fully connected layer designed to capture complex interdependencies across views. Furthermore, an independent cross-fusion learning loss function is introduced to model both the consistency and complementarity among feature interactions. Experimental results confirm that the proposed MFNets method demonstrates superior adaptability to few-shot scenarios, enhanced noise robustness, and improved fault diagnosis accuracy compared to existing baseline methods.

近年来，由于化学过程的复杂性日益增加，智能故障诊断，特别是处理混合变量的可行性，引起了越来越多的研究关注。在许多工业环境中，经常观察到包括连续和离散元素的混合变量，反映了现代过程系统的复杂性。然而，由于故障样本的稀缺性，导致了系统中出现了少次学习问题。从连续变量系统的见解表明，信息增强是解决这一问题的有效策略。为此，本研究提出了一种基于多特征融合网络（MFNets）的信息增强方法。具体来说，分别使用滑动时间窗、格拉曼角场（GAF）算法和高斯模糊技术从混合数据中提取数值特征、趋势特征和操作特征。然后，这些多视图特性通过一个共享的完全连接层集成，该层旨在捕获视图之间复杂的相互依赖关系。此外，引入独立的交叉融合学习损失函数，对特征交互之间的一致性和互补性进行建模。实验结果表明，与现有的基线方法相比，本文提出的MFNets方法具有更强的自适应性，增强了噪声鲁棒性，提高了故障诊断的准确性。

{"title":"Intelligent fault diagnosis in hybrid chemical processes under limited samples based on multi-feature fusion learning","authors":"Min Yin , Youqing Wang , Xin Ma , Yining Dong","doi":"10.1016/j.compchemeng.2025.109531","DOIUrl":"10.1016/j.compchemeng.2025.109531","url":null,"abstract":"<div><div>In recent years, intelligent fault diagnosis—particularly the feasibility of handling hybrid variables—has garnered increasing research attention due to the growing complexity of chemical processes. In many industrial settings, hybrid variables that include both continuous and discrete elements are frequently observed, reflecting the complexity of modern process systems. However, the scarcity of fault samples in such systems has led to the emergence of the few-shot learning problem. Insights from continuous-variable systems suggest that information augmentation is an effective strategy for addressing this issue. To this end, this study proposes a novel information augmentation approach based on Multi-Feature Fusion Networks (MFNets). Specifically, numerical, trend, and manipulation features are extracted from hybrid data using sliding time windows, the Gramian Angular Field (GAF) algorithm, and Gaussian blur techniques, respectively. These multi-view features are then integrated through a shared fully connected layer designed to capture complex interdependencies across views. Furthermore, an independent cross-fusion learning loss function is introduced to model both the consistency and complementarity among feature interactions. Experimental results confirm that the proposed MFNets method demonstrates superior adaptability to few-shot scenarios, enhanced noise robustness, and improved fault diagnosis accuracy compared to existing baseline methods.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"207 ","pages":"Article 109531"},"PeriodicalIF":3.9,"publicationDate":"2025-12-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145923404","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}

引用次数: 0

Data-driven source term estimation of hazardous gas leakages in complex chemical industrial parks 复杂化工园区有害气体泄漏的数据驱动源项估算

IF 3.9 2区工程技术 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Chemical Engineering

Pub Date : 2025-12-13 DOI: 10.1016/j.compchemeng.2025.109530

Chuantao Ni , Ziqiang Lang , Bing Wang , Ang Li , Chenxi Cao , Wenli Du , Feng Qian

Hazardous gas leakage in chemical industrial parks (CIPs) can cause irreversible damage to the environment and human health. When this happens, it is crucial to perform source term estimation (STE) timely and accurately and take effective measures to reduce or prevent possible losses. To achieve real-time STE, machine learning (ML)-based STE methods have recently been developed, aiming to build a ML model to represent the relationship between sensor measurements and STE outcome to facilitate real-time applications. However, the problem with these methods is that they often cannot handle cases when sensor measurements are beyond the scope of the training dataset. To address this limitation, in the present study, a novel approach is developed in which ML is used to generate a surrogate representation of complex atmospheric transport and dispersion processes by utilizing data from a high-fidelity computational fluid dynamics (CFD) model. This surrogate ML model captures the forward relationship between hazardous gas leakage locations and rates and the resulting sensor observations, enabling efficient nonlinear optimisation for off-line STE. In addition, the study, for the first time, introduces the concept of the incremental linear response matrix to address issues with potential system nonlinearities. These approaches are evaluated on a pseudo-real concentration dataset generated by CFD simulated ethane leakage scenarios in a CIP with complex obstacles. The findings validate the effectiveness of the proposed approaches and demonstrate their superiority over existing ML-based STE methods, particularly in scenarios that extend beyond the training data.

化工园区有害气体泄漏对环境和人体健康造成不可逆转的损害。在这种情况下，及时准确地进行源项估计（STE），并采取有效措施减少或防止可能的损失至关重要。为了实现实时STE，最近开发了基于机器学习（ML）的STE方法，旨在建立一个ML模型来表示传感器测量和STE结果之间的关系，以促进实时应用。然而，这些方法的问题在于，当传感器测量超出训练数据集的范围时，它们通常无法处理这种情况。为了解决这一限制，在本研究中，开发了一种新的方法，利用高保真计算流体动力学（CFD）模型的数据，使用ML来生成复杂大气传输和扩散过程的替代表示。该代理ML模型捕获了有害气体泄漏位置和速率与传感器观测结果之间的正向关系，从而实现了离线STE的有效非线性优化。此外，该研究还首次引入了增量线性响应矩阵的概念来解决潜在的系统非线性问题。通过CFD模拟复杂障碍物CIP中乙烷泄漏情景生成的伪真实浓度数据集对这些方法进行了评估。研究结果验证了所提出方法的有效性，并证明了它们优于现有的基于ml的STE方法，特别是在扩展到训练数据之外的场景中。

{"title":"Data-driven source term estimation of hazardous gas leakages in complex chemical industrial parks","authors":"Chuantao Ni , Ziqiang Lang , Bing Wang , Ang Li , Chenxi Cao , Wenli Du , Feng Qian","doi":"10.1016/j.compchemeng.2025.109530","DOIUrl":"10.1016/j.compchemeng.2025.109530","url":null,"abstract":"<div><div>Hazardous gas leakage in chemical industrial parks (CIPs) can cause irreversible damage to the environment and human health. When this happens, it is crucial to perform source term estimation (STE) timely and accurately and take effective measures to reduce or prevent possible losses. To achieve real-time STE, machine learning (ML)-based STE methods have recently been developed, aiming to build a ML model to represent the relationship between sensor measurements and STE outcome to facilitate real-time applications. However, the problem with these methods is that they often cannot handle cases when sensor measurements are beyond the scope of the training dataset. To address this limitation, in the present study, a novel approach is developed in which ML is used to generate a surrogate representation of complex atmospheric transport and dispersion processes by utilizing data from a high-fidelity computational fluid dynamics (CFD) model. This surrogate ML model captures the forward relationship between hazardous gas leakage locations and rates and the resulting sensor observations, enabling efficient nonlinear optimisation for off-line STE. In addition, the study, for the first time, introduces the concept of the incremental linear response matrix to address issues with potential system nonlinearities. These approaches are evaluated on a pseudo-real concentration dataset generated by CFD simulated ethane leakage scenarios in a CIP with complex obstacles. The findings validate the effectiveness of the proposed approaches and demonstrate their superiority over existing ML-based STE methods, particularly in scenarios that extend beyond the training data.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"206 ","pages":"Article 109530"},"PeriodicalIF":3.9,"publicationDate":"2025-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145797288","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}

引用次数: 0

Explainable Neural Network meets Graph Neural Network: Recent advances in process fault detection and diagnosis 可解释神经网络与图神经网络：过程故障检测与诊断的最新进展

IF 3.9 2区工程技术 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Chemical Engineering

Pub Date : 2025-12-13 DOI: 10.1016/j.compchemeng.2025.109528

Yi Liu , Bingbing Shen , David Shan-Hill Wong , Mingwei Jia , Yuan Yao

Modern processes rely on thousands of sensors, yet operators still lack trustworthy tools for measurement-driven fault detection and diagnosis. Deep learning excels at capturing nonlinearity and dynamics, but its opaque decision-making limits use in safety-critical processes. This survey fills that gap by introducing a measurement-oriented taxonomy of explainable neural networks (XNNs) and explainable graph neural networks (XGNNs) from the view of instrumentation and measurement. XNNs are posited as variable-centered instruments that assign calibrated importance scores to individual sensors for different faults. XGNNs are framed as topology-centric instruments, allowing direct measurement of interaction strength and causal propagation among units and control loops. This review delivers step-by-step guidelines that convert historical data into explainable detectors, trackers, and diagnostic meters. A comparison highlights when an XNN suffices and when an XGNN is mandatory, giving instrumentation engineers a decision chart. Different from prior surveys, we show that graphs are the faithful way to integrate P&IDs, material balances, and causal knowledge into deep learning measurements: XGNN explanations map directly onto process diagrams, creating on-screen instruments that display both the alarm and its physical trail. Finally, it concludes by identifying open challenges and recommending future directions for industrial deployment.

现代流程依赖于数千个传感器，但操作人员仍然缺乏可靠的工具来进行测量驱动的故障检测和诊断。深度学习擅长捕捉非线性和动态，但其不透明的决策限制了在安全关键过程中的应用。本研究从仪器和测量的角度引入了面向测量的可解释神经网络（xnn）和可解释图神经网络（xgnn）分类法，填补了这一空白。xnn被设定为可变中心的工具，为不同故障的单个传感器分配校准的重要性分数。xgnn被构建为以拓扑为中心的工具，允许直接测量单元和控制回路之间的相互作用强度和因果传播。本综述提供了将历史数据转换为可解释的检测器、跟踪器和诊断仪表的逐步指南。一个比较突出了XGNN在什么情况下是足够的，什么情况下是必须的，给仪器工程师一个决策图。与之前的调查不同，我们表明图形是将P&； id、物料平衡和因果知识集成到深度学习测量中的忠实方法：XGNN解释直接映射到流程图上，创建显示警报及其物理轨迹的屏幕仪器。最后，它通过确定开放的挑战和建议未来的工业部署方向来结束。

{"title":"Explainable Neural Network meets Graph Neural Network: Recent advances in process fault detection and diagnosis","authors":"Yi Liu , Bingbing Shen , David Shan-Hill Wong , Mingwei Jia , Yuan Yao","doi":"10.1016/j.compchemeng.2025.109528","DOIUrl":"10.1016/j.compchemeng.2025.109528","url":null,"abstract":"<div><div>Modern processes rely on thousands of sensors, yet operators still lack trustworthy tools for measurement-driven fault detection and diagnosis. Deep learning excels at capturing nonlinearity and dynamics, but its opaque decision-making limits use in safety-critical processes. This survey fills that gap by introducing a measurement-oriented taxonomy of explainable neural networks (XNNs) and explainable graph neural networks (XGNNs) from the view of instrumentation and measurement. XNNs are posited as variable-centered instruments that assign calibrated importance scores to individual sensors for different faults. XGNNs are framed as topology-centric instruments, allowing direct measurement of interaction strength and causal propagation among units and control loops. This review delivers step-by-step guidelines that convert historical data into explainable detectors, trackers, and diagnostic meters. A comparison highlights when an XNN suffices and when an XGNN is mandatory, giving instrumentation engineers a decision chart. Different from prior surveys, we show that graphs are the faithful way to integrate P&IDs, material balances, and causal knowledge into deep learning measurements: XGNN explanations map directly onto process diagrams, creating on-screen instruments that display both the alarm and its physical trail. Finally, it concludes by identifying open challenges and recommending future directions for industrial deployment.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"206 ","pages":"Article 109528"},"PeriodicalIF":3.9,"publicationDate":"2025-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145797289","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}

引用次数: 0

Sequential KDE‑guided zero-shot regression under process changes across materials 在不同材料的工艺变化下，顺序KDE引导的零射击回归

IF 3.9 2区工程技术 Q2 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computers & Chemical Engineering

Pub Date : 2025-12-12 DOI: 10.1016/j.compchemeng.2025.109522

Kanta Sato , Manabu Kano

This study presents a zero-shot regression framework that enables product quality prediction for a target process–material combination with no operating history. The framework facilitates scale-up and line transfer while minimizing experimental effort. A kernel density estimator sequentially selects source-process samples whose material features and product quality are similar to those on the target process, so that only relevant source data are reused. For each selected source-process sample, we predict operating conditions on the target process that would realize the same product quality for the same material, assuming that the material properties and product quality remain the same across processes. We then build a prediction model on a combined dataset consisting of the observed target-process samples and the selected source-process samples paired with their predicted target process operating conditions. A case study with two processes and seven materials demonstrates that the proposed method achieved consistent prediction accuracy; the median root mean squared error between prediction and measurement was 0.066 when using only target-process samples, 0.086 when combining target- and source-process samples without sample selection, and 0.055 with the proposed framework. The framework can successfully predict the behavior of new materials on the target process without additional experiments while suppressing negative transfer by automatically selecting and reusing existing data.

本研究提出了一个零射击回归框架，使产品质量预测的目标工艺-材料组合没有操作历史。该框架有助于扩大规模和线路转移，同时最大限度地减少实验工作量。核密度估计器依次选择材料特征和产品质量与目标过程相似的源过程样本，以便只重用相关的源数据。对于每个选定的源工艺样本，我们预测目标工艺的操作条件，假设材料特性和产品质量在各个工艺之间保持相同，从而实现相同材料的相同产品质量。然后，我们在一个组合数据集上建立一个预测模型，该数据集由观察到的目标工艺样本和选择的源工艺样本及其预测的目标工艺操作条件组成。两种工艺和七种材料的实例研究表明，该方法具有一致的预测精度；仅使用目标过程样本时，预测与测量的中位数均方根误差为0.066，不进行样本选择时，将目标和源过程样本结合使用时为0.086，使用所提出的框架时为0.055。该框架可以成功地预测新材料在目标过程中的行为，而无需额外的实验，同时通过自动选择和重用现有数据来抑制负传递。

{"title":"Sequential KDE‑guided zero-shot regression under process changes across materials","authors":"Kanta Sato , Manabu Kano","doi":"10.1016/j.compchemeng.2025.109522","DOIUrl":"10.1016/j.compchemeng.2025.109522","url":null,"abstract":"<div><div>This study presents a zero-shot regression framework that enables product quality prediction for a target process–material combination with no operating history. The framework facilitates scale-up and line transfer while minimizing experimental effort. A kernel density estimator sequentially selects source-process samples whose material features and product quality are similar to those on the target process, so that only relevant source data are reused. For each selected source-process sample, we predict operating conditions on the target process that would realize the same product quality for the same material, assuming that the material properties and product quality remain the same across processes. We then build a prediction model on a combined dataset consisting of the observed target-process samples and the selected source-process samples paired with their predicted target process operating conditions. A case study with two processes and seven materials demonstrates that the proposed method achieved consistent prediction accuracy; the median root mean squared error between prediction and measurement was 0.066 when using only target-process samples, 0.086 when combining target- and source-process samples without sample selection, and 0.055 with the proposed framework. The framework can successfully predict the behavior of new materials on the target process without additional experiments while suppressing negative transfer by automatically selecting and reusing existing data.</div></div>","PeriodicalId":286,"journal":{"name":"Computers & Chemical Engineering","volume":"206 ","pages":"Article 109522"},"PeriodicalIF":3.9,"publicationDate":"2025-12-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145836406","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}

引用次数: 0