Pub Date : 2025-01-01DOI: 10.1016/j.jprocont.2024.103353
Jing Li , Defeng He , Xiuli Wang , Yu Kang
This paper proposes a new explicit model predictive control (EMPC) scheme of constrained nonlinear systems with unknown but bounded input disturbances. Firstly, support vector machine is used to learn internal and external approximations of the feasible state space of the EMPC. Then, the control surface on the feasibility of EMPC is constructed by a backpropagation neural network (BPNN). The finite horizon optimal control solution to the EMPC can be computed from real-time data by training the control surface. The proposed EMPC is also suitable for nonlinear systems with higher dimensions in terms of reducing online computational burdens and enhancing control accuracy. Next, the Hoeffding's Inequality is used to ensure that the EMPC law computed by the BPNN approximation complies with the specified range with a high level of confidence. Moreover, some conditions are obtained to guarantee the stability and recursive feasibility of the EMPC with probabilistic assurances. Finally, a 160 MW boiler-turbine system is employed to verify the effectiveness and applications of the proposed method.
{"title":"BP neural network-based explicit MPC of nonlinear boiler-turbine systems","authors":"Jing Li , Defeng He , Xiuli Wang , Yu Kang","doi":"10.1016/j.jprocont.2024.103353","DOIUrl":"10.1016/j.jprocont.2024.103353","url":null,"abstract":"<div><div>This paper proposes a new explicit model predictive control (EMPC) scheme of constrained nonlinear systems with unknown but bounded input disturbances. Firstly, support vector machine is used to learn internal and external approximations of the feasible state space of the EMPC. Then, the control surface on the feasibility of EMPC is constructed by a backpropagation neural network (BPNN). The finite horizon optimal control solution to the EMPC can be computed from real-time data by training the control surface. The proposed EMPC is also suitable for nonlinear systems with higher dimensions in terms of reducing online computational burdens and enhancing control accuracy. Next, the Hoeffding's Inequality is used to ensure that the EMPC law computed by the BPNN approximation complies with the specified range with a high level of confidence. Moreover, some conditions are obtained to guarantee the stability and recursive feasibility of the EMPC with probabilistic assurances. Finally, a 160 MW boiler-turbine system is employed to verify the effectiveness and applications of the proposed method.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"145 ","pages":"Article 103353"},"PeriodicalIF":3.3,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143096452","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 : 2025-01-01DOI: 10.1016/j.jprocont.2024.103343
Damiano Rotondo , Marcin Pazera , Marcin Witczak
The paper deals with the issue of integrating fault estimation and fault-tolerant control for constrained continuous-time linear parameter varying systems with ellipsoidally bounded external disturbances. The unappealing effect related to the above coupling is that control and estimation influence each other. Thus, the main goal is to prevent such an unacceptable performance of the controlled system by providing a new integration strategy. The proposed framework is based on an output feedback approach, which is based on two staged: off-line – a low-complexity optimization task related to the fault estimation and control based on linear matrix inequalities, as well as on-line – a deterministic model predictive control for linear parameter-varying system. The effectiveness of this two-stage approach is illustrated with simulations based on a quadruple tank system.
{"title":"Integrated fault estimation and fault-tolerant control for constrained LPV systems subject to bounded disturbances","authors":"Damiano Rotondo , Marcin Pazera , Marcin Witczak","doi":"10.1016/j.jprocont.2024.103343","DOIUrl":"10.1016/j.jprocont.2024.103343","url":null,"abstract":"<div><div>The paper deals with the issue of integrating fault estimation and fault-tolerant control for constrained continuous-time linear parameter varying systems with ellipsoidally bounded external disturbances. The unappealing effect related to the above coupling is that control and estimation influence each other. Thus, the main goal is to prevent such an unacceptable performance of the controlled system by providing a new integration strategy. The proposed framework is based on an output feedback approach, which is based on two staged: <em>off-line</em> – a low-complexity optimization task related to the fault estimation and control based on linear matrix inequalities, as well as <em>on-line</em> – a deterministic model predictive control for linear parameter-varying system. The effectiveness of this two-stage approach is illustrated with simulations based on a quadruple tank system.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"145 ","pages":"Article 103343"},"PeriodicalIF":3.3,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143096546","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-01-01DOI: 10.1016/j.jprocont.2024.103345
Leonardo M. De Marco , Jorge Otávio Trierweiler , Fábio César Diehl , Marcelo Farenzena
Severe slugging is a prevalent issue in offshore oil wells that significantly hampers oil production. While active pressure control has proven effective in mitigating this problem, determining optimal setpoints remains a manual and labor-intensive procedure. This study introduces the Input-Output Cross Autocorrelation Diagram (IO-CAD), which examines both input and output autocorrelations to provide a more comprehensive assessment of control loops compared to previous methods that only use output autocorrelation data. Four indicators based on IO-CAD were developed and tested in two case studies involving offshore oil production. They were compared to an oscillation detection method published in the literature, as oscillations in the control loop may indicate the slugging flow. Early detection of slugging patterns is crucial in offshore oil production to prevent severe slugging and stabilize control loops. The results demonstrated that the IO-CAD indicators are robust against setpoint changes, disturbances, and noise in the control loop performance assessment, while the oscillation detection method from the literature is sensitive to measurement and process noise, as well as control loop oscillations.
{"title":"Input-Output Cross Autocorrelation Diagram (IO-CAD) for control loop performance assessment in offshore oil production","authors":"Leonardo M. De Marco , Jorge Otávio Trierweiler , Fábio César Diehl , Marcelo Farenzena","doi":"10.1016/j.jprocont.2024.103345","DOIUrl":"10.1016/j.jprocont.2024.103345","url":null,"abstract":"<div><div>Severe slugging is a prevalent issue in offshore oil wells that significantly hampers oil production. While active pressure control has proven effective in mitigating this problem, determining optimal setpoints remains a manual and labor-intensive procedure. This study introduces the Input-Output Cross Autocorrelation Diagram (IO-CAD), which examines both input and output autocorrelations to provide a more comprehensive assessment of control loops compared to previous methods that only use output autocorrelation data. Four indicators based on IO-CAD were developed and tested in two case studies involving offshore oil production. They were compared to an oscillation detection method published in the literature, as oscillations in the control loop may indicate the slugging flow. Early detection of slugging patterns is crucial in offshore oil production to prevent severe slugging and stabilize control loops. The results demonstrated that the IO-CAD indicators are robust against setpoint changes, disturbances, and noise in the control loop performance assessment, while the oscillation detection method from the literature is sensitive to measurement and process noise, as well as control loop oscillations.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"145 ","pages":"Article 103345"},"PeriodicalIF":3.3,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143096454","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 : 2025-01-01DOI: 10.1016/j.jprocont.2024.103354
Mingwei Jia , Chao Yang , Zhouxin Pan , Qiang Liu , Yi Liu
The development of soft sensors in process industries necessitates learning the dynamic variable relationships caused by physicochemical reactions, whilst avoiding noise interference that degrades prediction performance and explainability. To address this, an adversarial relationship graph learning soft sensor is proposed, comprising both relationship learning and prediction modules. Irrelevant and false variable relationships caused by noise are treated as negative information, quantified through mutual information loss. They are captured by alternately adversarial training the self-attention network and graph autoencoder. By excluding negative information, a suitable variable relationship graph is constructed. The graph convolutional network then mines information from the data and relationships for accurate prediction. Two practical cases verify the model’s physical consistency and demonstrate superior performance compared to several common models.
{"title":"Adversarial relationship graph learning soft sensor via negative information exclusion","authors":"Mingwei Jia , Chao Yang , Zhouxin Pan , Qiang Liu , Yi Liu","doi":"10.1016/j.jprocont.2024.103354","DOIUrl":"10.1016/j.jprocont.2024.103354","url":null,"abstract":"<div><div>The development of soft sensors in process industries necessitates learning the dynamic variable relationships caused by physicochemical reactions, whilst avoiding noise interference that degrades prediction performance and explainability. To address this, an adversarial relationship graph learning soft sensor is proposed, comprising both relationship learning and prediction modules. Irrelevant and false variable relationships caused by noise are treated as negative information, quantified through mutual information loss. They are captured by alternately adversarial training the self-attention network and graph autoencoder. By excluding negative information, a suitable variable relationship graph is constructed. The graph convolutional network then mines information from the data and relationships for accurate prediction. Two practical cases verify the model’s physical consistency and demonstrate superior performance compared to several common models.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"145 ","pages":"Article 103354"},"PeriodicalIF":3.3,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143096453","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 : 2025-01-01DOI: 10.1016/j.jprocont.2024.103342
Tian Wang , Ping Wang , Feng Yang , Shuai Wang , Qiang Fang , Meng Chi
Fault-tolerant control is crucial for ensuring flight safety in aircraft. However, existing methods for fault diagnosis in nonlinear systems face challenges such as data sparsity, limited generalization, and lack of explainability. To address these challenges, this paper proposes a multi-large language model (LLM) collaboration framework for few-shot link prediction in evolutionary fault diagnosis event graphs. The framework consists of two modules: the Clustering Language Model (LMc) and the Prediction Language Model (LMP). LMc utilizes the semantic understanding capabilities of LLMs to cluster entities and decompose large-scale graph data into smaller subgraphs, mitigating the impact of data sparsity on link prediction. LMP leverages the reasoning capabilities of LLMs to perform link prediction within each subgraph and fuses the prediction results to enhance accuracy and generalization. The completion of the link serves as a means to an end, which is to conduct fault diagnosis reasoning on a more detailed knowledge graph, thereby significantly improving the accuracy of fault diagnosis. Experimental results demonstrate that the proposed framework outperforms traditional embedding models and existing meta-learning methods on multiple datasets, particularly for sparse and background-rich datasets. This approach offers a novel solution for fault diagnosis in nonlinear systems, with significant theoretical and practical value.
{"title":"Multi large language model collaboration framework for few-shot link prediction in evolutionary fault diagnosis event graphs","authors":"Tian Wang , Ping Wang , Feng Yang , Shuai Wang , Qiang Fang , Meng Chi","doi":"10.1016/j.jprocont.2024.103342","DOIUrl":"10.1016/j.jprocont.2024.103342","url":null,"abstract":"<div><div>Fault-tolerant control is crucial for ensuring flight safety in aircraft. However, existing methods for fault diagnosis in nonlinear systems face challenges such as data sparsity, limited generalization, and lack of explainability. To address these challenges, this paper proposes a multi-large language model (LLM) collaboration framework for few-shot link prediction in evolutionary fault diagnosis event graphs. The framework consists of two modules: the Clustering Language Model (LMc) and the Prediction Language Model (LMP). LMc utilizes the semantic understanding capabilities of LLMs to cluster entities and decompose large-scale graph data into smaller subgraphs, mitigating the impact of data sparsity on link prediction. LMP leverages the reasoning capabilities of LLMs to perform link prediction within each subgraph and fuses the prediction results to enhance accuracy and generalization. The completion of the link serves as a means to an end, which is to conduct fault diagnosis reasoning on a more detailed knowledge graph, thereby significantly improving the accuracy of fault diagnosis. Experimental results demonstrate that the proposed framework outperforms traditional embedding models and existing meta-learning methods on multiple datasets, particularly for sparse and background-rich datasets. This approach offers a novel solution for fault diagnosis in nonlinear systems, with significant theoretical and practical value.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"145 ","pages":"Article 103342"},"PeriodicalIF":3.3,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143096540","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 : 2025-01-01DOI: 10.1016/j.jprocont.2024.103344
Qingkai Meng, Milad Shahvali, Stelios Vrachimis, Marios M. Polycarpou
As a typical nonlinear process control infrastructure, the safety and reliability of drinking water transport systems (DWTS) are affected by various factors, including their complex interconnected structures and external environments. This paper proposes a fault-tolerant control scheme for DWTS that ensures their states remain within safe boundaries despite the presence of disturbances, uncertainties and faults. Firstly, considering the impacts of random consumer behavior, unpredictable process and actuator faults, the DWTS is modeled as an interconnected stochastic nonlinear system. Secondly, combining the backstepping technique with control barrier functions, a sufficient and necessary condition for guaranteeing system safety is derived. Thirdly, by minimizing a loss function constructed based on dynamic programming, we synthesize a distributed controller using neural networks and theoretically prove the safety guarantees provided by our approach. Lastly, simulations are conducted to validate the effectiveness of the proposed approach on our benchmark water transport system.
{"title":"Fault-tolerant safe control for water networks: A backstepping neural control barrier function approach","authors":"Qingkai Meng, Milad Shahvali, Stelios Vrachimis, Marios M. Polycarpou","doi":"10.1016/j.jprocont.2024.103344","DOIUrl":"10.1016/j.jprocont.2024.103344","url":null,"abstract":"<div><div>As a typical nonlinear process control infrastructure, the safety and reliability of drinking water transport systems (DWTS) are affected by various factors, including their complex interconnected structures and external environments. This paper proposes a fault-tolerant control scheme for DWTS that ensures their states remain within safe boundaries despite the presence of disturbances, uncertainties and faults. Firstly, considering the impacts of random consumer behavior, unpredictable process and actuator faults, the DWTS is modeled as an interconnected stochastic nonlinear system. Secondly, combining the backstepping technique with control barrier functions, a sufficient and necessary condition for guaranteeing system safety is derived. Thirdly, by minimizing a loss function constructed based on dynamic programming, we synthesize a distributed controller using neural networks and theoretically prove the safety guarantees provided by our approach. Lastly, simulations are conducted to validate the effectiveness of the proposed approach on our benchmark water transport system.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"145 ","pages":"Article 103344"},"PeriodicalIF":3.3,"publicationDate":"2025-01-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143096451","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 : 2024-12-01DOI: 10.1016/j.jprocont.2024.103339
Yun Li , Neil Yorke-Smith , Tamas Keviczky
The way how the uncertainties are represented by sets plays a vital role in the performance of robust optimization (RO). This paper presents a novel approach leveraging machine learning (ML) techniques to construct data-driven uncertainty sets from historical uncertainty data for RO problems. The proposed method integrates Density-Based Spatial Clustering of Applications with Noise (DBSCAN), Gaussian Mixture Model (GMM), and Principle Component Analysis (PCA) systematically to eliminate the influence of uncertainty scenarios with low occurrence probability and generate a nonconvex uncertainty set that is a union of multiple basic subsets (box or ellipsoid) without sacrificing its computational tractability. In addition to presenting a comprehensive algorithm for uncertainty set development, this paper offers detailed guidelines for parameter tuning and performance analysis. By harnessing the well-established ML packages scikit-learn, a Python-based toolkit for implementing the proposed approach is also provided. Furthermore, a computationally efficient solution for a two-stage linear RO problem with the proposed data-driven uncertainty set is derived, alongside establishing a probabilistic guarantee of constraint satisfaction for out-of-sample uncertainties. Extensive numerical experiments, conducted on both synthetic and real-world datasets as well as an optimization-based control problem, are performed to demonstrate the efficacy of the proposed methodology.
{"title":"Machine learning enabled uncertainty set for data-driven robust optimization","authors":"Yun Li , Neil Yorke-Smith , Tamas Keviczky","doi":"10.1016/j.jprocont.2024.103339","DOIUrl":"10.1016/j.jprocont.2024.103339","url":null,"abstract":"<div><div>The way how the uncertainties are represented by sets plays a vital role in the performance of robust optimization (RO). This paper presents a novel approach leveraging machine learning (ML) techniques to construct data-driven uncertainty sets from historical uncertainty data for RO problems. The proposed method integrates Density-Based Spatial Clustering of Applications with Noise (DBSCAN), Gaussian Mixture Model (GMM), and Principle Component Analysis (PCA) systematically to eliminate the influence of uncertainty scenarios with low occurrence probability and generate a nonconvex uncertainty set that is a union of multiple basic subsets (box or ellipsoid) without sacrificing its computational tractability. In addition to presenting a comprehensive algorithm for uncertainty set development, this paper offers detailed guidelines for parameter tuning and performance analysis. By harnessing the well-established ML packages <span>scikit-learn</span>, a Python-based toolkit for implementing the proposed approach is also provided. Furthermore, a computationally efficient solution for a two-stage linear RO problem with the proposed data-driven uncertainty set is derived, alongside establishing a probabilistic guarantee of constraint satisfaction for out-of-sample uncertainties. Extensive numerical experiments, conducted on both synthetic and real-world datasets as well as an optimization-based control problem, are performed to demonstrate the efficacy of the proposed methodology.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"144 ","pages":"Article 103339"},"PeriodicalIF":3.3,"publicationDate":"2024-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142744901","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2024-11-26DOI: 10.1016/j.jprocont.2024.103341
Meng Zhou , Yan Wu , Jing Wang , Tarek Raïssi , Vicenç Puig
This paper proposes a fault detection strategy based on a two-step interval estimation method for T–S fuzzy systems with unmeasurable premise variables. First, an observer is designed to achieve robust point estimation under Lipschitz conditions. Then, the estimated error bounds are analyzed and optimized using the performance conditions to enable interval estimation. Furthermore, the residual threshold is derived from the interval estimation to achieve robust fault detection. Finally, an activated sludge process in a wastewater treatment is considered to validate the proposed method. Simulation results demonstrate that the proposed approach can provide more accurate state interval estimation and outperforms standard observer design methods in addressing fault detection problems compared with existing methods.
{"title":"Fault detection for T–S fuzzy systems with unmeasurable premise variables based on a two-step interval estimation method","authors":"Meng Zhou , Yan Wu , Jing Wang , Tarek Raïssi , Vicenç Puig","doi":"10.1016/j.jprocont.2024.103341","DOIUrl":"10.1016/j.jprocont.2024.103341","url":null,"abstract":"<div><div>This paper proposes a fault detection strategy based on a two-step interval estimation method for T–S fuzzy systems with unmeasurable premise variables. First, an <span><math><msub><mrow><mi>L</mi></mrow><mrow><mi>∞</mi></mrow></msub></math></span> observer is designed to achieve robust point estimation under Lipschitz conditions. Then, the estimated error bounds are analyzed and optimized using the <span><math><msub><mrow><mi>L</mi></mrow><mrow><mi>∞</mi></mrow></msub></math></span> performance conditions to enable interval estimation. Furthermore, the residual threshold is derived from the interval estimation to achieve robust fault detection. Finally, an activated sludge process in a wastewater treatment is considered to validate the proposed method. Simulation results demonstrate that the proposed approach can provide more accurate state interval estimation and outperforms standard <span><math><msub><mrow><mi>L</mi></mrow><mrow><mi>∞</mi></mrow></msub></math></span> observer design methods in addressing fault detection problems compared with existing methods.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"144 ","pages":"Article 103341"},"PeriodicalIF":3.3,"publicationDate":"2024-11-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142722584","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 : 2024-11-19DOI: 10.1016/j.jprocont.2024.103338
Congxin Li , Liangliang Sun
The volatility of slab weight in steelmaking-continuous casting (SCC) production, attributed to factors such as flexible order demand, is addressed in this paper. A robust optimization mathematical model for charge batch planning (CBP) with uncertain slab weight is established, and a collaborative optimization method using the surrogate Lagrangian relaxation (SLR) framework and improved objective feasibility pump (IOFP) is developed to solve the problem. In the SLR method, new step-size updating conditions are developed, eliminating the need for pre-estimating the optimal dual value. Additionally, only a subset of subproblems that satisfy the optimality conditions of the surrogate needs to be solved to overcome the low optimization efficiency resulting from oscillations in the feasible domain during internal searches in traditional Lagrangian relaxation (LR) methods. The IOFP method is employed to match the structure of the subproblem model of 0–1 mixed integer programming (MIP). During the search for integer solutions, a weighted objective function is added to the auxiliary model to improve the quality of solutions. Furthermore, it combines a variable neighborhood branching method to prevent the algorithm from entering into cycles. Finally, the effectiveness of the proposed model and the performance of the algorithm are validated through simulation experiments.
{"title":"A robust optimization approach for steeling-continuous casting charge batch planning with uncertain slab weight","authors":"Congxin Li , Liangliang Sun","doi":"10.1016/j.jprocont.2024.103338","DOIUrl":"10.1016/j.jprocont.2024.103338","url":null,"abstract":"<div><div>The volatility of slab weight in steelmaking-continuous casting (SCC) production, attributed to factors such as flexible order demand, is addressed in this paper. A robust optimization mathematical model for charge batch planning (CBP) with uncertain slab weight is established, and a collaborative optimization method using the surrogate Lagrangian relaxation (SLR) framework and improved objective feasibility pump (IOFP) is developed to solve the problem. In the SLR method, new step-size updating conditions are developed, eliminating the need for pre-estimating the optimal dual value. Additionally, only a subset of subproblems that satisfy the optimality conditions of the surrogate needs to be solved to overcome the low optimization efficiency resulting from oscillations in the feasible domain during internal searches in traditional Lagrangian relaxation (LR) methods. The IOFP method is employed to match the structure of the subproblem model of 0–1 mixed integer programming (MIP). During the search for integer solutions, a weighted objective function is added to the auxiliary model to improve the quality of solutions. Furthermore, it combines a variable neighborhood branching method to prevent the algorithm from entering into cycles. Finally, the effectiveness of the proposed model and the performance of the algorithm are validated through simulation experiments.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"144 ","pages":"Article 103338"},"PeriodicalIF":3.3,"publicationDate":"2024-11-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142705161","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 : 2024-11-16DOI: 10.1016/j.jprocont.2024.103340
Edward H. Bras, Tobias M. Louw, Steven M. Bradshaw
The adoption of reinforcement learning (RL) in chemical process industries is currently hindered by the use of black-box models that cannot be easily visualized or interpreted as well as the challenge of balancing safe control with exploration. Clearly illustrating the similarities between classical control- and RL theory, as well as demonstrating the possibility of maintaining process safety under RL-based control, will go a long way towards bridging the gap between academic research and industry practice. In this work, a simple approach to the dynamic online adaptation of a non-linear control policy initialised using PI control through RL is introduced. The familiar PI controller is represented as a plane in the state-action space, where the states comprise the error and integral error, and the action is the control input. The plane was recreated using a neural network and this recreated plane served as a readily visualizable initial “warm-started” policy for the RL agent. The actor-critic algorithm was applied to adapt the policy non-linearly during interaction with the controlled process, thereby leveraging the flexibility of the neural network to improve performance. Inherently safe control during training is ensured by introducing a soft active region component in the actor neural network. Finally, the use of cold connections is proposed whereby the state space can be augmented at any stage of training (e.g., through the incorporation of measurements to facilitate feedforward control) while fully preserving the agent’s training progress to date. By ensuring controller safety, the proposed methods are applicable to the dynamic adaptation of any process where stable PI control is feasible at nominal initial conditions.
强化学习（RL）目前在化工流程工业中的应用受到以下因素的阻碍：黑盒模型的使用不便于可视化或解释，以及在安全控制与探索之间取得平衡所面临的挑战。清楚地说明经典控制理论与 RL 理论之间的相似性，并证明在基于 RL 的控制下保持过程安全的可能性，将大大有助于缩小学术研究与行业实践之间的差距。在这项工作中，介绍了一种通过 RL 对使用 PI 控制初始化的非线性控制策略进行动态在线调整的简单方法。我们熟悉的 PI 控制器被表示为状态-动作空间中的一个平面，其中状态包括误差和积分误差，而动作则是控制输入。使用神经网络重新创建了该平面，并将该重新创建的平面作为 RL 代理可视化的初始 "热启动 "策略。在与受控过程交互的过程中，采用行为批判算法对策略进行非线性调整，从而利用神经网络的灵活性提高性能。通过在行动者神经网络中引入软活动区域组件，确保了训练期间的固有安全控制。最后，还提出了使用冷连接的方法，这样就可以在训练的任何阶段对状态空间进行扩展（例如，通过纳入测量数据来促进前馈控制），同时完全保留代理到目前为止的训练进度。通过确保控制器的安全性，所提出的方法适用于在标称初始条件下可进行稳定 PI 控制的任何过程的动态适应。
{"title":"Safe, visualizable reinforcement learning for process control with a warm-started actor network based on PI-control","authors":"Edward H. Bras, Tobias M. Louw, Steven M. Bradshaw","doi":"10.1016/j.jprocont.2024.103340","DOIUrl":"10.1016/j.jprocont.2024.103340","url":null,"abstract":"<div><div>The adoption of reinforcement learning (RL) in chemical process industries is currently hindered by the use of black-box models that cannot be easily visualized or interpreted as well as the challenge of balancing safe control with exploration. Clearly illustrating the similarities between classical control- and RL theory, as well as demonstrating the possibility of maintaining process safety under RL-based control, will go a long way towards bridging the gap between academic research and industry practice. In this work, a simple approach to the dynamic online adaptation of a non-linear control policy initialised using PI control through RL is introduced. The familiar PI controller is represented as a plane in the state-action space, where the states comprise the error and integral error, and the action is the control input. The plane was recreated using a neural network and this recreated plane served as a readily visualizable initial “warm-started” policy for the RL agent. The actor-critic algorithm was applied to adapt the policy non-linearly during interaction with the controlled process, thereby leveraging the flexibility of the neural network to improve performance. Inherently safe control during training is ensured by introducing a soft active region component in the actor neural network. Finally, the use of cold connections is proposed whereby the state space can be augmented at any stage of training (e.g., through the incorporation of measurements to facilitate feedforward control) while fully preserving the agent’s training progress to date. By ensuring controller safety, the proposed methods are applicable to the dynamic adaptation of any process where stable PI control is feasible at nominal initial conditions.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"144 ","pages":"Article 103340"},"PeriodicalIF":3.3,"publicationDate":"2024-11-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142664019","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","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