Pub Date : 2025-12-13DOI: 10.1016/j.jprocont.2025.103607
Lukas Munser , Ángeles Hoyo , Felix Petzke , Jaime A. Moreno , Stefan Streif
The detection and isolation of various faults in controlled environment agriculture is a notoriously complicated task, since various biological, sensory, and mechanical phenomena may interact with each other. In the present work, a residual-based approach is presented which enables the detection, isolation and quantification of different types of faults. For this purpose, observers are designed that can approximate the residuals despite model inaccuracies and measurement noise. The approach is demonstrated through experiments in a small-scale vertical farming unit whereby it is possible to distinguish between different fault types during operation.
{"title":"Residual-based fault detection and isolation in control environment agriculture","authors":"Lukas Munser , Ángeles Hoyo , Felix Petzke , Jaime A. Moreno , Stefan Streif","doi":"10.1016/j.jprocont.2025.103607","DOIUrl":"10.1016/j.jprocont.2025.103607","url":null,"abstract":"<div><div>The detection and isolation of various faults in controlled environment agriculture is a notoriously complicated task, since various biological, sensory, and mechanical phenomena may interact with each other. In the present work, a residual-based approach is presented which enables the detection, isolation and quantification of different types of faults. For this purpose, observers are designed that can approximate the residuals despite model inaccuracies and measurement noise. The approach is demonstrated through experiments in a small-scale vertical farming unit whereby it is possible to distinguish between different fault types during operation.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"157 ","pages":"Article 103607"},"PeriodicalIF":3.9,"publicationDate":"2025-12-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145749776","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-11DOI: 10.1016/j.jprocont.2025.103595
Rafael D. de Oliveira, Johannes Jäschke
Dynamic Flux Balance Analysis (dFBA) models are powerful metabolic models that have a large potential for application in bioprocess control and optimisation. However, dFBA has an embedded linear FBA optimisation problem with degenerate solutions that give rise to multiple possible state trajectories that all satisfy the model. To address this uncertainty, we propose using a robust control approach based on Multi-stage Economic Nonlinear Model Predictive Control, which allows handling the degenerate solutions of the FBA problem without being too conservative. We propose to add a regularisation term to the FBA problem, to ensure a unique solution and generate the uncertainty scenarios by varying the regularization weights. The scenarios generated in that way then correspond to different solutions of the FBA problem. Then, the KKT conditions of the regularised FBA problem are imposed as equality constraints on the optimal control problem, which is solved using a direct collocation approach. Our methodology is evaluated through a case study on the optimal control of a fed-batch bioreactor for Escherichia coli growth, subject to a constraint on acetate concentration. The results demonstrate that the proposed MS-ENMPC approach, combined with the dFBA model, effectively satisfies the constraints despite uncertainties in the system trajectories.
{"title":"Multi-stage Economic Nonlinear Model Predictive Control of bioreactors using dynamic flux balance analysis models","authors":"Rafael D. de Oliveira, Johannes Jäschke","doi":"10.1016/j.jprocont.2025.103595","DOIUrl":"10.1016/j.jprocont.2025.103595","url":null,"abstract":"<div><div>Dynamic Flux Balance Analysis (dFBA) models are powerful metabolic models that have a large potential for application in bioprocess control and optimisation. However, dFBA has an embedded linear FBA optimisation problem with degenerate solutions that give rise to multiple possible state trajectories that all satisfy the model. To address this uncertainty, we propose using a robust control approach based on Multi-stage Economic Nonlinear Model Predictive Control, which allows handling the degenerate solutions of the FBA problem without being too conservative. We propose to add a regularisation term to the FBA problem, to ensure a unique solution and generate the uncertainty scenarios by varying the regularization weights. The scenarios generated in that way then correspond to different solutions of the FBA problem. Then, the KKT conditions of the regularised FBA problem are imposed as equality constraints on the optimal control problem, which is solved using a direct collocation approach. Our methodology is evaluated through a case study on the optimal control of a fed-batch bioreactor for <em>Escherichia coli</em> growth, subject to a constraint on acetate concentration. The results demonstrate that the proposed MS-ENMPC approach, combined with the dFBA model, effectively satisfies the constraints despite uncertainties in the system trajectories.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"157 ","pages":"Article 103595"},"PeriodicalIF":3.9,"publicationDate":"2025-12-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145737899","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-05DOI: 10.1016/j.jprocont.2025.103604
Yanxin Zhang, Shaoyuan Li
The dynamic evolution of suspended sediment plume for deep-sea mining poses significant challenges for long-term prediction, owing to its inherently nonlinear transport behavior, unknown key parameters, and changing monitoring conditions after mining. To address these issues, this study proposes a framework integrating prediction, sensing, and refinement. Specifically, an incremental Physics-Informed Neural Network (PINN) enhanced with the Learning without Forgetting (LwF) strategy is developed to enable adaptive parameter updates while preserving prior physical knowledge. Furthermore, the sensor layout is optimized to enhance local observability. Numerical results demonstrate that, compared with the traditional PINN model, the proposed method effectively reduces prediction errors by 18.6% and achieves accurate prediction of the dynamic suspended sediment plume.
{"title":"An incremental physics-informed neural network for rapid prediction of suspended sediment plume for deep-sea mining","authors":"Yanxin Zhang, Shaoyuan Li","doi":"10.1016/j.jprocont.2025.103604","DOIUrl":"10.1016/j.jprocont.2025.103604","url":null,"abstract":"<div><div>The dynamic evolution of suspended sediment plume for deep-sea mining poses significant challenges for long-term prediction, owing to its inherently nonlinear transport behavior, unknown key parameters, and changing monitoring conditions after mining. To address these issues, this study proposes a framework integrating prediction, sensing, and refinement. Specifically, an incremental Physics-Informed Neural Network (PINN) enhanced with the Learning without Forgetting (LwF) strategy is developed to enable adaptive parameter updates while preserving prior physical knowledge. Furthermore, the sensor layout is optimized to enhance local observability. Numerical results demonstrate that, compared with the traditional PINN model, the proposed method effectively reduces prediction errors by 18.6% and achieves accurate prediction of the dynamic suspended sediment plume.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"157 ","pages":"Article 103604"},"PeriodicalIF":3.9,"publicationDate":"2025-12-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145694511","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-01DOI: 10.1016/j.jprocont.2025.103592
Chonggao Hu , Ridong Zhang , Furong Gao
Traditional two-dimensional (2D) model predictive model iterative learning control strategies can only rely on feedback to passively deal with time delays, repetitive disturbances, and non-repetitive disturbances of batch processes. To address these shortcomings, this paper proposes a two-dimensional model predictive iterative learning control strategy using an improved state space model structure with new error compensation (2D-EC-MPILC). Firstly, a two-dimensional extended non-minimal state space (2D-ENMSS) model is established, which can provide more degrees of freedom for controller design. Secondly, a novel error compensation (EC) strategy is proposed to correct the tracking error value of the current batch. The novel 2D-EC-MPILC controller is designed with both additional tuning degrees and batch-wise error correction, ensuring an improved control performance. The proposed algorithm is tested on the holding pressure control system of an injection molding process and the temperature control system of a nonlinear batch reactor.
{"title":"New design of two-dimensional model predictive iterative learning control with novel error compensation for batch processes","authors":"Chonggao Hu , Ridong Zhang , Furong Gao","doi":"10.1016/j.jprocont.2025.103592","DOIUrl":"10.1016/j.jprocont.2025.103592","url":null,"abstract":"<div><div>Traditional two-dimensional (2D) model predictive model iterative learning control strategies can only rely on feedback to passively deal with time delays, repetitive disturbances, and non-repetitive disturbances of batch processes. To address these shortcomings, this paper proposes a two-dimensional model predictive iterative learning control strategy using an improved state space model structure with new error compensation (2D-EC-MPILC). Firstly, a two-dimensional extended non-minimal state space (2D-ENMSS) model is established, which can provide more degrees of freedom for controller design. Secondly, a novel error compensation (EC) strategy is proposed to correct the tracking error value of the current batch. The novel 2D-EC-MPILC controller is designed with both additional tuning degrees and batch-wise error correction, ensuring an improved control performance. The proposed algorithm is tested on the holding pressure control system of an injection molding process and the temperature control system of a nonlinear batch reactor.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"157 ","pages":"Article 103592"},"PeriodicalIF":3.9,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145646040","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-01DOI: 10.1016/j.jprocont.2025.103591
Abhilash Dev, Sharad Bhartiya, Sachin Patwardhan
Sample-based nonlinear state estimation is an intensive problem to solve and significantly contributes to the computation delay in real-time applications. This paper explores a novel neural network (NN) based state estimation method in order to reduce the computation time required by sample-based state estimators during online deployment. The proposed method obtains estimates for the predicted as well as the filtered states, wherein the NN is trained using data obtained from simulations of constrained state estimators. The efficacy of the method is demonstrated by learning constrained Unscented Kalman Filter (UKF) on a six-state benchmark Williams Otto reactor with four measurements using a deep feedforward NN. The NN is then used as the proposal density in lieu of UKF for a Particle Filter (PF) implementation. Another example consists of exploring the role of the trained NN as a state estimator in a nonlinear internal model control (NIMC) application for tracking of economic setpoints in the benchmark Willaims-Otto reactor. Both these applications show that the proposed approach significantly outperforms the sample-based estimator in terms of the computation time while matching its performance, thereby reducing the latency related to sample-based state estimators.
{"title":"Efficient computation for sample-based state estimators using deep neural networks","authors":"Abhilash Dev, Sharad Bhartiya, Sachin Patwardhan","doi":"10.1016/j.jprocont.2025.103591","DOIUrl":"10.1016/j.jprocont.2025.103591","url":null,"abstract":"<div><div>Sample-based nonlinear state estimation is an intensive problem to solve and significantly contributes to the computation delay in real-time applications. This paper explores a novel neural network (NN) based state estimation method in order to reduce the computation time required by sample-based state estimators during online deployment. The proposed method obtains estimates for the predicted as well as the filtered states, wherein the NN is trained using data obtained from simulations of constrained state estimators. The efficacy of the method is demonstrated by learning constrained Unscented Kalman Filter (UKF) on a six-state benchmark Williams Otto reactor with four measurements using a deep feedforward NN. The NN is then used as the proposal density in lieu of UKF for a Particle Filter (PF) implementation. Another example consists of exploring the role of the trained NN as a state estimator in a nonlinear internal model control (NIMC) application for tracking of economic setpoints in the benchmark Willaims-Otto reactor. Both these applications show that the proposed approach significantly outperforms the sample-based estimator in terms of the computation time while matching its performance, thereby reducing the latency related to sample-based state estimators.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"156 ","pages":"Article 103591"},"PeriodicalIF":3.9,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145617872","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-12-01DOI: 10.1016/j.jprocont.2025.103596
Akhtar Jan , Rehan Ali Shah , Ebraheem Alzahrani , Zehba Raizah
<div><div>Enzymatic reactions operate under a variety of biochemical responses in cellular systems and are frequently affected by intrinsic variations imposed by molecular uncertainty. This study examined both deterministic and stochastic models for a two-step enzymatic process in which the primary substrate <span><math><mrow><msub><mrow><mi>S</mi></mrow><mrow><mn>1</mn></mrow></msub><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow></mrow></math></span> and the inhibitory substrate <span><math><mrow><msub><mrow><mi>S</mi></mrow><mrow><mn>2</mn></mrow></msub><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow></mrow></math></span> collaborate to generate intermediate complexes <span><math><mrow><msub><mrow><mi>C</mi></mrow><mrow><mn>1</mn></mrow></msub><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow></mrow></math></span> and <span><math><mrow><msub><mrow><mi>C</mi></mrow><mrow><mn>2</mn></mrow></msub><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow></mrow></math></span>. The overall dynamics of the reaction are represented by combining the enzyme–substrate kinetics and the resulting product production. The invariant region of the system, the existence and non-negativity of solutions, the classification of equilibrium states and their stability characteristics are the main issues of a qualitative investigation. For both the models, the threshold parameter <span><math><msub><mrow><mi>T</mi></mrow><mrow><mn>0</mn></mrow></msub></math></span> is determined. The system’s sensitivity to stochastic influences is shown by the comparison analysis, which reveals that random perturbations typically lower the system’s effective activation threshold by <strong>25–35%</strong> when compared to the deterministic calculation. Additionally, sensitivity studies showed that stochastic noise parameters (<span><math><mrow><msub><mrow><mi>σ</mi></mrow><mrow><mn>3</mn></mrow></msub><mo>=</mo><msub><mrow><mi>σ</mi></mrow><mrow><mn>4</mn></mrow></msub><mo>=</mo><mn>0</mn><mo>.</mo><mn>1</mn></mrow></math></span>) generate around <strong>15–25%</strong> variation in the rates of complex formation; however, altering the initial enzyme level (<span><math><mrow><msub><mrow><mi>E</mi></mrow><mrow><mn>0</mn></mrow></msub><mo>=</mo><mn>0</mn><mo>.</mo><mn>3</mn></mrow></math></span>) impacts the reaction progress by about <strong>40%</strong>. Numerical simulations, performed over <span><math><mrow><mi>t</mi><mo>=</mo><mn>150</mn></mrow></math></span> with rate constants <span><math><mrow><msub><mrow><mi>k</mi></mrow><mrow><mn>1</mn></mrow></msub><mo>=</mo><mn>0</mn><mo>.</mo><mn>3</mn></mrow></math></span>, <span><math><mrow><msub><mrow><mi>k</mi></mrow><mrow><mn>2</mn></mrow></msub><mo>=</mo><mn>0</mn><mo>.</mo><mn>4</mn></mrow></math></span>, <span><math><mrow><msub><mrow><mi>k</mi></mrow><mrow><mn>3</mn></mrow></msub><mo>=</mo><mn>0</mn><mo>.</mo><mn>2</mn></mrow></math></span>, and <span><math><mrow><msub><mrow><mi>k</mi></mrow><mrow><mn>4</mn></mrow></msub><mo>=</mo><mn>0</mn><mo>.</mo><mn>5</mn></
{"title":"Mathematical modeling of stability in enzyme-catalyzed cooperative reactions through deterministic and stochastic approaches","authors":"Akhtar Jan , Rehan Ali Shah , Ebraheem Alzahrani , Zehba Raizah","doi":"10.1016/j.jprocont.2025.103596","DOIUrl":"10.1016/j.jprocont.2025.103596","url":null,"abstract":"<div><div>Enzymatic reactions operate under a variety of biochemical responses in cellular systems and are frequently affected by intrinsic variations imposed by molecular uncertainty. This study examined both deterministic and stochastic models for a two-step enzymatic process in which the primary substrate <span><math><mrow><msub><mrow><mi>S</mi></mrow><mrow><mn>1</mn></mrow></msub><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow></mrow></math></span> and the inhibitory substrate <span><math><mrow><msub><mrow><mi>S</mi></mrow><mrow><mn>2</mn></mrow></msub><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow></mrow></math></span> collaborate to generate intermediate complexes <span><math><mrow><msub><mrow><mi>C</mi></mrow><mrow><mn>1</mn></mrow></msub><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow></mrow></math></span> and <span><math><mrow><msub><mrow><mi>C</mi></mrow><mrow><mn>2</mn></mrow></msub><mrow><mo>(</mo><mi>t</mi><mo>)</mo></mrow></mrow></math></span>. The overall dynamics of the reaction are represented by combining the enzyme–substrate kinetics and the resulting product production. The invariant region of the system, the existence and non-negativity of solutions, the classification of equilibrium states and their stability characteristics are the main issues of a qualitative investigation. For both the models, the threshold parameter <span><math><msub><mrow><mi>T</mi></mrow><mrow><mn>0</mn></mrow></msub></math></span> is determined. The system’s sensitivity to stochastic influences is shown by the comparison analysis, which reveals that random perturbations typically lower the system’s effective activation threshold by <strong>25–35%</strong> when compared to the deterministic calculation. Additionally, sensitivity studies showed that stochastic noise parameters (<span><math><mrow><msub><mrow><mi>σ</mi></mrow><mrow><mn>3</mn></mrow></msub><mo>=</mo><msub><mrow><mi>σ</mi></mrow><mrow><mn>4</mn></mrow></msub><mo>=</mo><mn>0</mn><mo>.</mo><mn>1</mn></mrow></math></span>) generate around <strong>15–25%</strong> variation in the rates of complex formation; however, altering the initial enzyme level (<span><math><mrow><msub><mrow><mi>E</mi></mrow><mrow><mn>0</mn></mrow></msub><mo>=</mo><mn>0</mn><mo>.</mo><mn>3</mn></mrow></math></span>) impacts the reaction progress by about <strong>40%</strong>. Numerical simulations, performed over <span><math><mrow><mi>t</mi><mo>=</mo><mn>150</mn></mrow></math></span> with rate constants <span><math><mrow><msub><mrow><mi>k</mi></mrow><mrow><mn>1</mn></mrow></msub><mo>=</mo><mn>0</mn><mo>.</mo><mn>3</mn></mrow></math></span>, <span><math><mrow><msub><mrow><mi>k</mi></mrow><mrow><mn>2</mn></mrow></msub><mo>=</mo><mn>0</mn><mo>.</mo><mn>4</mn></mrow></math></span>, <span><math><mrow><msub><mrow><mi>k</mi></mrow><mrow><mn>3</mn></mrow></msub><mo>=</mo><mn>0</mn><mo>.</mo><mn>2</mn></mrow></math></span>, and <span><math><mrow><msub><mrow><mi>k</mi></mrow><mrow><mn>4</mn></mrow></msub><mo>=</mo><mn>0</mn><mo>.</mo><mn>5</mn></","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"156 ","pages":"Article 103596"},"PeriodicalIF":3.9,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145617871","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-21DOI: 10.1016/j.jprocont.2025.103594
Marta Catalão , José Pinto , Cristiana A.V. Torres , Filomena Freitas , Maria A.M. Reis , Rafael S. Costa , Rui Oliveira
Many previous studies have investigated the economic production of polyhydroxyalkanoates (PHA) by natural microbiomes. A key underlying strategy is the feast and famine (F/F) feeding regimen for bacteria selection. For this purpose, a sequencing batch reactor (SBR) is commonly operated in a sequence of F/F cycles until an evolved microbiome is attained with high PHA storage capacity. The effectiveness of this process is critically dependent on control parameters such as the hydraulic retention time (HRT), organic loading rate (OLR) and carbon-to-nitrogen ratio (C/N) applied at each cycle. This study evaluates for the first time a physics-informed neural network (PINN) for model predictive control (MPC) of microbiome evolution in a SBR. A PINN model was trained on historical data collected in a SBR operated over 93 days and 31 cycles. Carbon (acetate), Nitrogen (ammonium), Volatile Suspended Solids (VSS) and intracellular PHA concentration data were used to train and validate the PINN. Subsequently, a second SBR experiment was conducted under automatic control of the PINN over a period of 36 days and 12 cycles. A transfer learning method was implemented leverage on in-process data to minimize process-model mismatch. The results showed a systematic cycle-to-cycle prediction error decrease. The intracellular PHA concentration systematic increased from 0.51 % (w/w) to 16.5 % (w/w) at the 12th cycle (32-fold increase). The final evolved microbiome, collected at the 12th cycle, was inoculated in a production reactor yielding a final intracellular PHA content of 52.86 % (w/w) and volumetric concentration of 8.93 g PHA/L. Overall, the PINN-MPC method has shown high potential to efficiently explore the reactor design space and to implement in autonomy efficient strategies for natural microbiome evolution.
许多先前的研究已经研究了天然微生物群对聚羟基烷酸酯（PHA）的经济生产。一个关键的潜在策略是细菌选择的盛宴和饥荒（F/F）喂养方案。为此目的，测序间歇式反应器（SBR）通常以F/F循环的顺序运行，直到进化的微生物组达到具有高PHA存储能力。该工艺的有效性主要取决于控制参数，如水力停留时间（HRT）、有机载荷率（OLR）和碳氮比（C/N）。本研究首次评估了用于SBR微生物组进化模型预测控制（MPC）的物理信息神经网络（PINN）。根据SBR运行93天31个周期收集的历史数据，对PINN模型进行了训练。碳（乙酸）、氮（铵）、挥发性悬浮物（VSS）和细胞内PHA浓度数据用于训练和验证PINN。随后，在PINN自动控制下进行第二次SBR实验，为期36天，12个周期。利用进程内数据实现迁移学习方法，最大限度地减少过程模型不匹配。结果表明，系统的周期间预测误差减小。第12个循环时，细胞内PHA浓度从0.51 % （w/w）增加到16.5 % (w/w)，增加了32倍。在第12个循环中收集最终进化的微生物组，在生产反应器中接种，最终细胞内PHA含量为52.86 % (w/w)，体积浓度为8.93 g PHA/L。总体而言，PINN-MPC方法在有效探索反应器设计空间和实现自然微生物群进化的自主高效策略方面显示出很高的潜力。
{"title":"Bioprocess model-predictive control with physics-informed neural networks: Driving microbiome evolution toward high polyhydroxyalkanoates production capacity","authors":"Marta Catalão , José Pinto , Cristiana A.V. Torres , Filomena Freitas , Maria A.M. Reis , Rafael S. Costa , Rui Oliveira","doi":"10.1016/j.jprocont.2025.103594","DOIUrl":"10.1016/j.jprocont.2025.103594","url":null,"abstract":"<div><div>Many previous studies have investigated the economic production of polyhydroxyalkanoates (PHA) by natural microbiomes. A key underlying strategy is the feast and famine (F/F) feeding regimen for bacteria selection. For this purpose, a sequencing batch reactor (SBR) is commonly operated in a sequence of F/F cycles until an evolved microbiome is attained with high PHA storage capacity. The effectiveness of this process is critically dependent on control parameters such as the hydraulic retention time (HRT), organic loading rate (OLR) and carbon-to-nitrogen ratio (C/N) applied at each cycle. This study evaluates for the first time a physics-informed neural network (PINN) for model predictive control (MPC) of microbiome evolution in a SBR. A PINN model was trained on historical data collected in a SBR operated over 93 days and 31 cycles. Carbon (acetate), Nitrogen (ammonium), Volatile Suspended Solids (VSS) and intracellular PHA concentration data were used to train and validate the PINN. Subsequently, a second SBR experiment was conducted under automatic control of the PINN over a period of 36 days and 12 cycles. A transfer learning method was implemented leverage on in-process data to minimize process-model mismatch. The results showed a systematic cycle-to-cycle prediction error decrease. The intracellular PHA concentration systematic increased from 0.51 % (w/w) to 16.5 % (w/w) at the 12th cycle (32-fold increase). The final evolved microbiome, collected at the 12th cycle, was inoculated in a production reactor yielding a final intracellular PHA content of 52.86 % (w/w) and volumetric concentration of 8.93 g PHA/L. Overall, the PINN-MPC method has shown high potential to efficiently explore the reactor design space and to implement in autonomy efficient strategies for natural microbiome evolution.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"156 ","pages":"Article 103594"},"PeriodicalIF":3.9,"publicationDate":"2025-11-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145571421","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-21DOI: 10.1016/j.jprocont.2025.103593
Eike Cramer
Real-world (bio)chemical processes often exhibit stochastic dynamics with non-trivial correlations and state-dependent fluctuations. Model predictive control (MPC) often must consider these fluctuations to achieve reliable performance. However, most process models simply add stationary noise terms to a deterministic prediction. This work proposes using conditional normalizing flows as discrete-time models to learn stochastic dynamics. Normalizing flows learn the probability density function (PDF) of the states explicitly, given prior states and control inputs. In addition to standard least squares (LSQ) objectives, this work derives a marginal log-likelihood (MLL) objective based on the explicit PDF and Markov chain simulations. In a reactor study, the normalizing flow MPC setpoint errors in open and closed-loop cases are competitive with a full model-based stochastic MPC. Furthermore, the chance constraints lead to fewer constraint violations than the benchmark controller. The MLL objective yields slightly more stable optimization results than the LSQ, particularly for small scenario sets.
{"title":"Least squares and marginal log-likelihood model predictive control using normalizing flows","authors":"Eike Cramer","doi":"10.1016/j.jprocont.2025.103593","DOIUrl":"10.1016/j.jprocont.2025.103593","url":null,"abstract":"<div><div>Real-world (bio)chemical processes often exhibit stochastic dynamics with non-trivial correlations and state-dependent fluctuations. Model predictive control (MPC) often must consider these fluctuations to achieve reliable performance. However, most process models simply add stationary noise terms to a deterministic prediction. This work proposes using conditional normalizing flows as discrete-time models to learn stochastic dynamics. Normalizing flows learn the probability density function (PDF) of the states explicitly, given prior states and control inputs. In addition to standard least squares (LSQ) objectives, this work derives a marginal log-likelihood (MLL) objective based on the explicit PDF and Markov chain simulations. In a reactor study, the normalizing flow MPC setpoint errors in open and closed-loop cases are competitive with a full model-based stochastic MPC. Furthermore, the chance constraints lead to fewer constraint violations than the benchmark controller. The MLL objective yields slightly more stable optimization results than the LSQ, particularly for small scenario sets.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"156 ","pages":"Article 103593"},"PeriodicalIF":3.9,"publicationDate":"2025-11-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145571423","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-15DOI: 10.1016/j.jprocont.2025.103590
Xiaona Song , Ao Shang , Shuai Song , Danjing Zheng , Choon Ki Ahn
In this work, the issue of exponential admissibility analysis and controller designs for the singular linear parameter-varying partial differential equation (LPV-PDE) system is investigated. First, the reaction process in the chemical tubular reactor is described by a singular LPV-PDE, and the definition for the singular LPV-PDE system to be exponentially admissible is established by the Galerkin method, energy estimates, etc. Second, the parameter-dependent fuzzy dynamic event-triggered mechanism is developed to alleviate communication pressure by reducing the amount of transmitted data. Furthermore, a sliding surface is constructed, then a suitable control law is designed to ensure the reachability of system states, and sufficient conditions for the singular LPV-PDE system to be exponentially admissible are given during the sliding phase. Finally, the effectiveness of the proposed method is verified through simulation studies.
{"title":"Exponential admissibility and control for singular LPV-PDE systems with application to chemical tubular reactor","authors":"Xiaona Song , Ao Shang , Shuai Song , Danjing Zheng , Choon Ki Ahn","doi":"10.1016/j.jprocont.2025.103590","DOIUrl":"10.1016/j.jprocont.2025.103590","url":null,"abstract":"<div><div>In this work, the issue of exponential admissibility analysis and controller designs for the singular linear parameter-varying partial differential equation (LPV-PDE) system is investigated. First, the reaction process in the chemical tubular reactor is described by a singular LPV-PDE, and the definition for the singular LPV-PDE system to be exponentially admissible is established by the Galerkin method, energy estimates, etc. Second, the parameter-dependent fuzzy dynamic event-triggered mechanism is developed to alleviate communication pressure by reducing the amount of transmitted data. Furthermore, a sliding surface is constructed, then a suitable control law is designed to ensure the reachability of system states, and sufficient conditions for the singular LPV-PDE system to be exponentially admissible are given during the sliding phase. Finally, the effectiveness of the proposed method is verified through simulation studies.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"156 ","pages":"Article 103590"},"PeriodicalIF":3.9,"publicationDate":"2025-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145571422","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-11DOI: 10.1016/j.jprocont.2025.103588
Ming Lu , Yingcong Li , Xianke He , Lei He , Ying Zou , Zunhui Yi , Pei Li
In the flotation cell, the flotation process involves complex physicochemical reactions and exhibits unknown, intricate nonlinear dynamic characteristics, making it difficult to establish an accurate model or predict froth flotation grade. To address this issue, this paper proposes a mechanism-based modeling approach for predicting froth flotation grade. First, the working mechanism of the flotation cell is analyzed, the concentrate grade is redefined, and a simplified mechanistic model is derived. Next, preprocessed flotation process data is integrated with an intelligent optimization algorithm to conduct mechanism-guided data-driven modeling. This identifies the optimal system matrix parameters in the mechanistic model under current working conditions, thereby enhancing its adaptability to the actual flotation environment. Finally, the optimized system matrix parameters are incorporated into the mechanistic flotation grade model to construct an online prediction model. The predicted values are used to guide parameter settings during the flotation process. Experiments show that this method has high measurement accuracy, good generalization performance, and strong robustness.
{"title":"The flotation grade prediction model based on mechanism-guided and data-driven approaches","authors":"Ming Lu , Yingcong Li , Xianke He , Lei He , Ying Zou , Zunhui Yi , Pei Li","doi":"10.1016/j.jprocont.2025.103588","DOIUrl":"10.1016/j.jprocont.2025.103588","url":null,"abstract":"<div><div>In the flotation cell, the flotation process involves complex physicochemical reactions and exhibits unknown, intricate nonlinear dynamic characteristics, making it difficult to establish an accurate model or predict froth flotation grade. To address this issue, this paper proposes a mechanism-based modeling approach for predicting froth flotation grade. First, the working mechanism of the flotation cell is analyzed, the concentrate grade is redefined, and a simplified mechanistic model is derived. Next, preprocessed flotation process data is integrated with an intelligent optimization algorithm to conduct mechanism-guided data-driven modeling. This identifies the optimal system matrix parameters in the mechanistic model under current working conditions, thereby enhancing its adaptability to the actual flotation environment. Finally, the optimized system matrix parameters are incorporated into the mechanistic flotation grade model to construct an online prediction model. The predicted values are used to guide parameter settings during the flotation process. Experiments show that this method has high measurement accuracy, good generalization performance, and strong robustness.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"156 ","pages":"Article 103588"},"PeriodicalIF":3.9,"publicationDate":"2025-11-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145520306","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","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