Journal of Process Control最新文献_第10页

A physics-based multi-regime approach for estimation of head losses in operating hydropower plants 运行水电站水头损失估算的基于物理的多状态方法

IF 3.9 2区计算机科学 Q2 AUTOMATION & CONTROL SYSTEMS

Journal of Process Control

Pub Date : 2025-10-01 Epub Date: 2025-09-03 DOI: 10.1016/j.jprocont.2025.103528

Augustin Alonso , Gerard Robert , Gildas Besançon

In this paper, the problem of estimating head losses in the hydraulic feeding system of a hydropower plant is considered. Accurate head loss assessment is crucial for performance monitoring, efficiency optimization, and predictive maintenance of these critical energy infrastructures. To this end, a nonlinear state-space model based on fundamental physical principles is first established. Recognizing the challenges of observability with a full complex model, this paper proposes a multi-regime modelling strategy, where the full model is particularized into simplified forms suitable for different operational scenarios (normal operation, quasi-static conditions, and plant shutdown). This approach facilitates the estimation of specific head loss coefficients or their combinations. Various estimation techniques are then explored and applied to these models, primarily based on Kalman filters for state-observer approaches and direct least squares for regression-based methods, all integrating real-time measurements. The efficacy of these methods is validated through comprehensive simulations and tests using operational data collected from an industrial hydropower facility.

本文研究了某水电站水力给水系统水头损失的估算问题。准确的水头损失评估对于这些关键能源基础设施的性能监测、效率优化和预测性维护至关重要。为此，首先建立了基于基本物理原理的非线性状态空间模型。考虑到完整复杂模型的可观测性挑战，本文提出了一种多状态建模策略，将完整模型具体化为适合不同操作场景（正常运行、准静态条件和工厂关闭）的简化形式。这种方法有助于估计具体的水头损失系数或其组合。然后探索各种估计技术并将其应用于这些模型，主要基于状态观测器方法的卡尔曼滤波器和基于回归的方法的直接最小二乘，所有这些方法都集成了实时测量。通过对某工业水电设施运行数据的综合仿真和试验，验证了这些方法的有效性。

{"title":"A physics-based multi-regime approach for estimation of head losses in operating hydropower plants","authors":"Augustin Alonso , Gerard Robert , Gildas Besançon","doi":"10.1016/j.jprocont.2025.103528","DOIUrl":"10.1016/j.jprocont.2025.103528","url":null,"abstract":"<div><div>In this paper, the problem of estimating head losses in the hydraulic feeding system of a hydropower plant is considered. Accurate head loss assessment is crucial for performance monitoring, efficiency optimization, and predictive maintenance of these critical energy infrastructures. To this end, a nonlinear state-space model based on fundamental physical principles is first established. Recognizing the challenges of observability with a full complex model, this paper proposes a multi-regime modelling strategy, where the full model is particularized into simplified forms suitable for different operational scenarios (normal operation, quasi-static conditions, and plant shutdown). This approach facilitates the estimation of specific head loss coefficients or their combinations. Various estimation techniques are then explored and applied to these models, primarily based on Kalman filters for state-observer approaches and direct least squares for regression-based methods, all integrating real-time measurements. The efficacy of these methods is validated through comprehensive simulations and tests using operational data collected from an industrial hydropower facility.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"154 ","pages":"Article 103528"},"PeriodicalIF":3.9,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144932537","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}

引用次数: 0

An offline-to-online reinforcement learning framework with trajectory-guided exploration for industrial process control 基于轨迹导向探索的工业过程控制离线到在线强化学习框架

IF 3.9 2区计算机科学 Q2 AUTOMATION & CONTROL SYSTEMS

Journal of Process Control

Pub Date : 2025-10-01 Epub Date: 2025-08-30 DOI: 10.1016/j.jprocont.2025.103535

Jiyang Chen, Na Luo

Reinforcement learning (RL) in industrial process control faces critical challenges, including limited data availability, unsafe exploration, and the high cost of high-fidelity simulators. These issues limit the practical adoption of RL in process control systems. To address these limitations, this paper presents a comprehensive framework that combines offline pre-training with online finetuning. Specifically, the framework first employs offline RL method to learn conservative policies from historical data, preventing overestimation of unseen actions. It then transitions to fine-tuning using online RL method with a mixed replay buffer that gradually shifts from offline to online data. To further enhance safety during online exploration, this work introduces a trajectory-guided strategy that leverages timestamped sub-optimal expert demonstrations. Rather than replacing agent actions entirely, the proposed method computes a weighted combination of agent and expert actions based on a decaying intervention rate. Both components are designed as modular additions that can be integrated into existing actor-critic algorithms without structural modifications. Case studies on penicillin fermentation and simulated moving bed (SMB) processes demonstrate that the proposed framework outperforms baseline algorithms in terms of learning efficiency, stability, computation costs, and operational safety.

工业过程控制中的强化学习（RL）面临着严峻的挑战，包括有限的数据可用性、不安全的探索和高保真模拟器的高成本。这些问题限制了RL在过程控制系统中的实际应用。为了解决这些限制，本文提出了一个将离线预训练与在线微调相结合的综合框架。具体而言，该框架首先采用离线强化学习方法从历史数据中学习保守策略，防止对未见动作的高估。然后过渡到使用在线RL方法进行微调，使用混合重播缓冲区逐渐从离线数据转移到在线数据。为了进一步提高在线勘探过程中的安全性，这项工作引入了一种轨迹引导策略，该策略利用了时间戳次优专家演示。该方法不是完全取代代理行为，而是基于衰减的干预率计算代理和专家行为的加权组合。这两个组件都被设计为模块化的附加组件，可以集成到现有的演员批评算法中，而无需进行结构修改。青霉素发酵和模拟移动床（SMB）过程的案例研究表明，所提出的框架在学习效率、稳定性、计算成本和操作安全性方面优于基线算法。

{"title":"An offline-to-online reinforcement learning framework with trajectory-guided exploration for industrial process control","authors":"Jiyang Chen, Na Luo","doi":"10.1016/j.jprocont.2025.103535","DOIUrl":"10.1016/j.jprocont.2025.103535","url":null,"abstract":"<div><div>Reinforcement learning (RL) in industrial process control faces critical challenges, including limited data availability, unsafe exploration, and the high cost of high-fidelity simulators. These issues limit the practical adoption of RL in process control systems. To address these limitations, this paper presents a comprehensive framework that combines offline pre-training with online finetuning. Specifically, the framework first employs offline RL method to learn conservative policies from historical data, preventing overestimation of unseen actions. It then transitions to fine-tuning using online RL method with a mixed replay buffer that gradually shifts from offline to online data. To further enhance safety during online exploration, this work introduces a trajectory-guided strategy that leverages timestamped sub-optimal expert demonstrations. Rather than replacing agent actions entirely, the proposed method computes a weighted combination of agent and expert actions based on a decaying intervention rate. Both components are designed as modular additions that can be integrated into existing actor-critic algorithms without structural modifications. Case studies on penicillin fermentation and simulated moving bed (SMB) processes demonstrate that the proposed framework outperforms baseline algorithms in terms of learning efficiency, stability, computation costs, and operational safety.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"154 ","pages":"Article 103535"},"PeriodicalIF":3.9,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144917021","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}

引用次数: 0

Model selection and parameter optimization of model predictive control for building radiant systems 建筑辐射系统模型预测控制的模型选择与参数优化

IF 3.9 2区计算机科学 Q2 AUTOMATION & CONTROL SYSTEMS

Journal of Process Control

Pub Date : 2025-10-01 Epub Date: 2025-08-05 DOI: 10.1016/j.jprocont.2025.103512

Qiong Chen , Wenjing Wang , Nan Li

In this work, we present a comprehensive parametric investigation quantifying the influence of reduced-order models (ROMs) of varying fidelity on both dynamic and steady-state performance of a Model Predictive Control (MPC) loop for building thermal regulation. Through simulations using the full-order model and ROMs of orders 1–6, we systematically determined that ROMs of order 4 or higher achieve temperature overshoots within 0.2 °C, settling times under 15 min, and steady-state errors below 0.5 °C when paired with prediction horizons of 12–24 steps and control horizons ≥ 2, thus matching full-order MPC performance while reducing computation by up to 70 %. In contrast, the lowest-order ROM (ROM1) requires a prediction horizon ≤ 12 and a control horizon ≥ 3 to limit overshoot to 1.0 °C and static error to 1.2 °C. Furthermore, the original model and high-order ROMs maintain robust control (overshoot < 0.5 °C, settling time < 10 min) across manipulated-variable rate weights from 0.1 to 1.0 and manipulated-output weights from 0.5 to 2.0, whereas ROM1 exhibits strong sensitivity, operating acceptably only near MV-rate ≈ 0.2 and MO weight ≈ 1.0. These quantitative guidelines enable practitioners to balance computational cost and control accuracy by selecting an appropriately ordered ROM and tuning horizons and weightings within the identified numerical ranges.

在这项工作中，我们提出了一项全面的参数研究，量化了不同保真度的降阶模型（ROMs）对用于建筑热调节的模型预测控制（MPC）回路的动态和稳态性能的影响。通过使用全阶模型和1-6阶ROMs的模拟，我们系统地确定了4阶或更高阶ROMs在0.2°C内实现温度超调，沉降时间小于15 min，当与12-24步的预测层和控制层≥ 2相匹配时，稳态误差小于0.5°C，从而匹配全阶MPC性能，同时减少计算量高达70% %。相比之下，最低阶ROM （ROM1）需要预测水平≤ 12，控制水平≥ 3才能将超调限制在1.0°C，将静态误差限制在1.2°C。此外，原始模型和高阶rom保持鲁棒控制(超调<；0.5℃，沉降时间<；10 min)在0.1 - 1.0和0.5 - 2.0的操纵变量权值范围内，而ROM1表现出很强的灵敏度，仅在mv率≈ 0.2和MO权值≈ 1.0附近可接受地工作。这些定量指导方针使从业者能够通过选择适当有序的ROM和在确定的数值范围内调整视野和权重来平衡计算成本和控制精度。

{"title":"Model selection and parameter optimization of model predictive control for building radiant systems","authors":"Qiong Chen , Wenjing Wang , Nan Li","doi":"10.1016/j.jprocont.2025.103512","DOIUrl":"10.1016/j.jprocont.2025.103512","url":null,"abstract":"<div><div>In this work, we present a comprehensive parametric investigation quantifying the influence of reduced-order models (ROMs) of varying fidelity on both dynamic and steady-state performance of a Model Predictive Control (MPC) loop for building thermal regulation. Through simulations using the full-order model and ROMs of orders 1–6, we systematically determined that ROMs of order 4 or higher achieve temperature overshoots within 0.2 °C, settling times under 15 min, and steady-state errors below 0.5 °C when paired with prediction horizons of 12–24 steps and control horizons ≥ 2, thus matching full-order MPC performance while reducing computation by up to 70 %. In contrast, the lowest-order ROM (ROM1) requires a prediction horizon ≤ 12 and a control horizon ≥ 3 to limit overshoot to 1.0 °C and static error to 1.2 °C. Furthermore, the original model and high-order ROMs maintain robust control (overshoot < 0.5 °C, settling time < 10 min) across manipulated-variable rate weights from 0.1 to 1.0 and manipulated-output weights from 0.5 to 2.0, whereas ROM1 exhibits strong sensitivity, operating acceptably only near MV-rate ≈ 0.2 and MO weight ≈ 1.0. These quantitative guidelines enable practitioners to balance computational cost and control accuracy by selecting an appropriately ordered ROM and tuning horizons and weightings within the identified numerical ranges.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"154 ","pages":"Article 103512"},"PeriodicalIF":3.9,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144770853","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}

引用次数: 0

Performance grade similarity-based generalized zero-shot operating performance assessment of industrial processes with insufficient samples 样本不足情况下基于性能等级相似性的工业过程广义零弹运行性能评价

IF 3.9 2区计算机科学 Q2 AUTOMATION & CONTROL SYSTEMS

Journal of Process Control

Pub Date : 2025-10-01 Epub Date: 2025-08-14 DOI: 10.1016/j.jprocont.2025.103523

Siqi Wang , Yan Liu , Lulu Fu , Fei Chu , Fuli Wang , Chenhui Bao

The process operating performance assessment (POPA) is vital for enhancing economic production in industrial processes. This study addresses the challenge in POPA of assessment unseen performance grades with zero samples, while also dealing with insufficient data for seen performance grades. We propose PGSGZSIS, a performance grade similarity-based generalized zero-shot method that integrates accessible superficial expert knowledge with a multi-expert voting mechanism to construct a performance grade similarity matrix (PGSM). The PGSM is validated by seen-data-driven expert reliability calculation, reducing dependency on deep expert knowledge while enhancing objectivity through data quantification. Additionally, an auxiliary set augmentation strategy based on feature similarity is introduced, constructing an auxiliary dataset by screening samples from similar operational conditions to address scarce seen samples. By constructing the PGSM and augmenting seen samples with auxiliary data, our approach not only alleviates the issue of insufficient seen samples but also tackles the generalized zero-shot learning (GZSL) problem for POPA. Experimental results validate the effectiveness of the proposed method in a hydrometallurgical process.

过程运行绩效评价（POPA）是提高工业过程经济生产的重要手段。本研究解决了POPA中零样本评估未见绩效等级的挑战，同时也处理了未见绩效等级数据不足的问题。提出了一种基于性能等级相似度的广义零射击方法PGSGZSIS，该方法将可访问的表面专家知识与多专家投票机制相结合，构建了性能等级相似度矩阵（PGSM）。通过可视化数据驱动的专家可靠性计算对PGSM进行验证，减少了对深度专家知识的依赖，同时通过数据量化提高了客观性。此外，引入了一种基于特征相似度的辅助集增强策略，通过筛选相似操作条件下的样本来构建辅助数据集，以解决稀缺的可见样本。该方法通过构造PGSM和用辅助数据扩充见过样本，不仅缓解了见过样本不足的问题，而且解决了POPA的广义零次学习问题。实验结果验证了该方法在湿法冶金过程中的有效性。

{"title":"Performance grade similarity-based generalized zero-shot operating performance assessment of industrial processes with insufficient samples","authors":"Siqi Wang , Yan Liu , Lulu Fu , Fei Chu , Fuli Wang , Chenhui Bao","doi":"10.1016/j.jprocont.2025.103523","DOIUrl":"10.1016/j.jprocont.2025.103523","url":null,"abstract":"<div><div>The process operating performance assessment (POPA) is vital for enhancing economic production in industrial processes. This study addresses the challenge in POPA of assessment unseen performance grades with zero samples, while also dealing with insufficient data for seen performance grades. We propose PGSGZSIS, a performance grade similarity-based generalized zero-shot method that integrates accessible superficial expert knowledge with a multi-expert voting mechanism to construct a performance grade similarity matrix (PGSM). The PGSM is validated by seen-data-driven expert reliability calculation, reducing dependency on deep expert knowledge while enhancing objectivity through data quantification. Additionally, an auxiliary set augmentation strategy based on feature similarity is introduced, constructing an auxiliary dataset by screening samples from similar operational conditions to address scarce seen samples. By constructing the PGSM and augmenting seen samples with auxiliary data, our approach not only alleviates the issue of insufficient seen samples but also tackles the generalized zero-shot learning (GZSL) problem for POPA. Experimental results validate the effectiveness of the proposed method in a hydrometallurgical process.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"154 ","pages":"Article 103523"},"PeriodicalIF":3.9,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144841765","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}

引用次数: 0

Memory-guided reconstruction for generalized zero-shot industrial fault diagnosis 广义零距工业故障诊断的记忆引导重构

IF 3.9 2区计算机科学 Q2 AUTOMATION & CONTROL SYSTEMS

Journal of Process Control

Pub Date : 2025-10-01 Epub Date: 2025-08-25 DOI: 10.1016/j.jprocont.2025.103531

Zhengwei Hu , Wei Xiang , Jingchao Peng , Haitao Zhao

Recently, zero-shot learning (ZSL) has emerged as a promising method in the industrial fault diagnosis. However, restricted by the strong bias problem, unseen class faults tend to be classified as seen class faults in the generalized zero-shot learning (GZSL) task. To address this issue, a novel method called memory-guided reconstruction (MGR) is proposed for generalized zero-shot industrial fault diagnosis. In MGR, memory prototypes of seen classes are first learned by a self-organizing map (SOM) and stored in a memory module. During the training, the encoding of a sample is obtained from the encoder as a query. Instead of directly reconstructing from this query, a support memory aggregated from relevant memory prototypes of the query is delivered to the decoder for reconstruction. A specific memory alignment matrix is designed to measure the consistency between the query and support memory. At the test stage, unseen classes tend to produce higher reconstruction errors than seen classes because the support memory is acquired from seen class memory prototypes. A new “identify-classify” learning paradigm is adopted: identify the domain (i.e. seen or unseen) of the test sample based on the strengthened reconstruction error, and further classifythe sample within the identified domain. Extensive experiments on the benchmark dataset demonstrate the significant superiority of MGR. Moreover, MGR achieves competitive results compared to supervised learning methods. The code of MGR is available at https://github.com/htz-ecust/memory-guided-autoencoder.

近年来，零采样学习（zero-shot learning， ZSL）作为一种很有前途的故障诊断方法出现在工业故障诊断中。然而，在广义零次学习（GZSL）任务中，受强偏差问题的限制，未见类错误容易被分类为见类错误。为了解决这一问题，提出了一种用于广义零点工业故障诊断的记忆引导重构方法。在MGR中，可视类的内存原型首先通过自组织映射（SOM）学习并存储在内存模块中。在训练过程中，作为查询从编码器获得样本的编码。不是直接从该查询进行重构，而是将从查询的相关内存原型聚合的支持内存传递给解码器进行重构。设计了一个特定的内存对齐矩阵来度量查询和支持内存之间的一致性。在测试阶段，不可见的类往往比可见的类产生更高的重构错误，因为支持内存是从可见的类内存原型中获得的。采用了一种新的“识别-分类”学习范式：基于增强的重构误差识别测试样本的域（即可见或不可见），并在识别域内进一步对样本进行分类。在基准数据集上的大量实验证明了MGR的显著优越性。此外，与监督学习方法相比，MGR取得了具有竞争力的结果。MGR的代码可在https://github.com/htz-ecust/memory-guided-autoencoder上获得。

{"title":"Memory-guided reconstruction for generalized zero-shot industrial fault diagnosis","authors":"Zhengwei Hu , Wei Xiang , Jingchao Peng , Haitao Zhao","doi":"10.1016/j.jprocont.2025.103531","DOIUrl":"10.1016/j.jprocont.2025.103531","url":null,"abstract":"<div><div>Recently, zero-shot learning (ZSL) has emerged as a promising method in the industrial fault diagnosis. However, restricted by the strong bias problem, unseen class faults tend to be classified as seen class faults in the generalized zero-shot learning (GZSL) task. To address this issue, a novel method called memory-guided reconstruction (MGR) is proposed for generalized zero-shot industrial fault diagnosis. In MGR, memory prototypes of seen classes are first learned by a self-organizing map (SOM) and stored in a memory module. During the training, the encoding of a sample is obtained from the encoder as a query. Instead of directly reconstructing from this query, a support memory aggregated from relevant memory prototypes of the query is delivered to the decoder for reconstruction. A specific <em>memory alignment matrix</em> is designed to measure the consistency between the query and support memory. At the test stage, unseen classes tend to produce higher reconstruction errors than seen classes because the support memory is acquired from seen class memory prototypes. A new “<em>identify-classify</em>” learning paradigm is adopted: <em>identify</em> the domain (i.e. seen or unseen) of the test sample based on the strengthened reconstruction error, and further <em>classify</em>the sample within the identified domain. Extensive experiments on the benchmark dataset demonstrate the significant superiority of MGR. Moreover, MGR achieves competitive results compared to supervised learning methods. The code of MGR is available at <span><span>https://github.com/htz-ecust/memory-guided-autoencoder</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"154 ","pages":"Article 103531"},"PeriodicalIF":3.9,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144893754","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}

引用次数: 0

Addressing external distorted heterogeneity: Input–output disentangled causal representation for mixed time series 处理外部扭曲异质性：混合时间序列的输入-输出解纠缠因果表示

IF 3.9 2区计算机科学 Q2 AUTOMATION & CONTROL SYSTEMS

Journal of Process Control

Pub Date : 2025-10-01 Epub Date: 2025-08-11 DOI: 10.1016/j.jprocont.2025.103521

Liujiayi Zhao , Baoxue Li , Chunhui Zhao

In causal analysis, it is common for industrial systems to have a mixture of continuous and discrete variables, called distribution heterogeneity. In fact, discrete variables typically serve as external inputs to modulate continuous variables in these systems. Existing methods for causal discovery encounter the External Distorted Heterogeneity challenge. The challenge is defined as the difficulty of correcting the statistical relationships distorted by discrete inputs, interfering with the identification of causality within systems. To overcome the challenge, we propose a method called Input–Output Disentangled Causal Representation. The key idea is to reveal the continuous external control effects from discrete inputs and exclude the control effects from observed outputs to decouple the inference of causality. Technically, a reversible external control converter is designed to recover the continuous control effects from discrete input signals through affine processes, bridging the heterogeneity. In addition, we construct an additive causal model to distinguish between causal effects from inputs and outputs, capturing disentangled representations in a unified space through feature distribution alignment and discrimination. Dual predictions are designed to exclude the regulatory influences from observed outputs using gradient truncation, thereby decoupling the inference of causality. The proposed method demonstrates robust causal identification accuracy across diverse datasets and scenarios, outperforming existing approaches in high-dimensional input–output systems. These results highlight its potential for industrial applications in the causal discovery of input–output systems.

在因果分析中，工业系统通常有连续变量和离散变量的混合，称为分布异质性。事实上，在这些系统中，离散变量通常作为外部输入来调制连续变量。现有的因果发现方法遇到了外部扭曲异质性的挑战。这一挑战被定义为纠正被离散输入扭曲的统计关系的困难，干扰了系统内因果关系的识别。为了克服这一挑战，我们提出了一种称为“输入-输出解纠缠因果表示”的方法。关键思想是从离散输入中揭示连续的外部控制效应，并从观察到的输出中排除控制效应，以解耦因果关系的推断。从技术上讲，设计了一个可逆的外部控制转换器，通过仿射过程从离散输入信号中恢复连续控制效果，桥接异质性。此外，我们构建了一个加性因果模型来区分输入和输出的因果效应，通过特征分布对齐和区分来捕获统一空间中的解纠缠表示。双重预测的目的是使用梯度截断来排除来自观察输出的调节影响，从而解耦因果关系的推断。所提出的方法在不同的数据集和场景中证明了强大的因果识别准确性，优于高维输入输出系统中的现有方法。这些结果突出了它在投入产出系统因果发现方面的工业应用潜力。

{"title":"Addressing external distorted heterogeneity: Input–output disentangled causal representation for mixed time series","authors":"Liujiayi Zhao , Baoxue Li , Chunhui Zhao","doi":"10.1016/j.jprocont.2025.103521","DOIUrl":"10.1016/j.jprocont.2025.103521","url":null,"abstract":"<div><div>In causal analysis, it is common for industrial systems to have a mixture of continuous and discrete variables, called distribution heterogeneity. In fact, discrete variables typically serve as external inputs to modulate continuous variables in these systems. Existing methods for causal discovery encounter the External Distorted Heterogeneity challenge. The challenge is defined as the difficulty of correcting the statistical relationships distorted by discrete inputs, interfering with the identification of causality within systems. To overcome the challenge, we propose a method called Input–Output Disentangled Causal Representation. The key idea is to reveal the continuous external control effects from discrete inputs and exclude the control effects from observed outputs to decouple the inference of causality. Technically, a reversible external control converter is designed to recover the continuous control effects from discrete input signals through affine processes, bridging the heterogeneity. In addition, we construct an additive causal model to distinguish between causal effects from inputs and outputs, capturing disentangled representations in a unified space through feature distribution alignment and discrimination. Dual predictions are designed to exclude the regulatory influences from observed outputs using gradient truncation, thereby decoupling the inference of causality. The proposed method demonstrates robust causal identification accuracy across diverse datasets and scenarios, outperforming existing approaches in high-dimensional input–output systems. These results highlight its potential for industrial applications in the causal discovery of input–output systems.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"154 ","pages":"Article 103521"},"PeriodicalIF":3.9,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144809632","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}

引用次数: 0

Harnessing monotonicity to design an adaptive PI passivity-based controller for a fuel-cell system 利用单调性设计燃料电池系统自适应PI无源控制器

IF 3.9 2区计算机科学 Q2 AUTOMATION & CONTROL SYSTEMS

Journal of Process Control

Pub Date : 2025-10-01 Epub Date: 2025-08-06 DOI: 10.1016/j.jprocont.2025.103511

Carlo A. Beltrán , Rafael Cisneros , Diego Langarica-Cordoba , Romeo Ortega , Luis H. Díaz-Saldierna

In this paper, a controller is designed to regulate the output voltage of a fuel-cell (FC) system comprising a proton-exchange membrane FC feeding a purely resistive load through a boost converter. The controller aims to maintain voltage regulation regardless of uncertainties in the resistive load. Leveraging the monotonicity of the FC polarization curve, it is demonstrated that the non-linear system can be controlled with a simple proportional–integral (PI) action through the PI-passivity-based control (PI-PBC) methodology. The result is subsequently extended to an adaptive version, enabling it to address parametric uncertainties, including inductor parasitic resistance, load variations, and fuel cell polarization curve parameters. The overall system is proved to be stable by regulating the output voltage under parametric uncertainty. Experimental results validate the proposed controller.

本文设计了一种控制器来调节燃料电池（FC）系统的输出电压，该系统由质子交换膜FC通过升压转换器提供纯电阻负载。控制器的目的是保持电压调节，而不考虑电阻性负载的不确定性。利用FC极化曲线的单调性，通过PI-无源控制（PI- pbc）方法，证明了非线性系统可以用一个简单的比例积分（PI）作用来控制。结果随后扩展到自适应版本，使其能够解决参数不确定性，包括电感寄生电阻、负载变化和燃料电池极化曲线参数。在参数不确定的情况下，通过调节输出电压，证明了整个系统是稳定的。实验结果验证了该控制器的有效性。

引用次数: 0

A contrastive generative network with feature-attribute consistency for zero-shot fault diagnosis in process industries 面向过程工业零爆故障诊断的特征-属性一致性对比生成网络

IF 3.9 2区计算机科学 Q2 AUTOMATION & CONTROL SYSTEMS

Journal of Process Control

Pub Date : 2025-10-01 Epub Date: 2025-08-25 DOI: 10.1016/j.jprocont.2025.103529

Lin Sha , Jiaqi Li , Min Wang , Shihang Yu , Sibo Qiao

In fault diagnosis tasks for process industries, comprehensively identifying all potential fault types poses significant challenges. Therefore, zero-shot fault diagnosis has gradually become a research hotspot. Currently, existing zero-shot fault diagnosis methods commonly face domain shift issues, which limit diagnostic performance. To address this shift, this paper proposes a feature-attribute consistency contrastive generative network (FAC-CGNet). This method combines attribute supervision with a contrastive learning mechanism to simultaneously maintain attribute consistency and decouple the feature space during feature generation. FAC-CGNet constructs an attribute-guided feature generation framework that integrates attribute information into the feature transformation process, ensuring that the generated features in the feature space remain consistent with their corresponding attributes. Furthermore, to prevent excessive overlap of generated features with similar attributes in the feature space, the paper designs a contrastive decoupling module. This module optimizes the feature space distribution through feature separation constraints and further enhances feature representation discrimination by combining a feature concatenation strategy. Finally, experiments on the public TEP dataset show that FAC-CGNet achieves an average accuracy of 83.1% in unknown fault diagnosis and significantly optimizes the feature representations in the feature space, confirming the effectiveness and superiority of the proposed method.

在过程工业的故障诊断任务中，全面识别所有潜在的故障类型是一项重大挑战。因此，零炮故障诊断逐渐成为研究热点。目前，现有的零射故障诊断方法普遍存在域漂移问题，限制了诊断性能。为了解决这一转变，本文提出了一种特征-属性一致性对比生成网络（facc - cgnet）。该方法将属性监督与对比学习机制相结合，在特征生成过程中保持属性一致性，同时对特征空间进行解耦。facc - cgnet构建了属性导向的特征生成框架，将属性信息融入到特征转换过程中，保证在特征空间中生成的特征与其对应的属性保持一致。此外，为了防止生成的具有相似属性的特征在特征空间中过度重叠，本文设计了对比解耦模块。该模块通过特征分离约束优化特征空间分布，并结合特征拼接策略进一步增强特征表示判别能力。最后，在公共TEP数据集上的实验表明，facc - cgnet在未知故障诊断中平均准确率达到83.1%，显著优化了特征空间中的特征表示，验证了所提方法的有效性和优越性。

{"title":"A contrastive generative network with feature-attribute consistency for zero-shot fault diagnosis in process industries","authors":"Lin Sha , Jiaqi Li , Min Wang , Shihang Yu , Sibo Qiao","doi":"10.1016/j.jprocont.2025.103529","DOIUrl":"10.1016/j.jprocont.2025.103529","url":null,"abstract":"<div><div>In fault diagnosis tasks for process industries, comprehensively identifying all potential fault types poses significant challenges. Therefore, zero-shot fault diagnosis has gradually become a research hotspot. Currently, existing zero-shot fault diagnosis methods commonly face domain shift issues, which limit diagnostic performance. To address this shift, this paper proposes a feature-attribute consistency contrastive generative network (FAC-CGNet). This method combines attribute supervision with a contrastive learning mechanism to simultaneously maintain attribute consistency and decouple the feature space during feature generation. FAC-CGNet constructs an attribute-guided feature generation framework that integrates attribute information into the feature transformation process, ensuring that the generated features in the feature space remain consistent with their corresponding attributes. Furthermore, to prevent excessive overlap of generated features with similar attributes in the feature space, the paper designs a contrastive decoupling module. This module optimizes the feature space distribution through feature separation constraints and further enhances feature representation discrimination by combining a feature concatenation strategy. Finally, experiments on the public TEP dataset show that FAC-CGNet achieves an average accuracy of 83.1% in unknown fault diagnosis and significantly optimizes the feature representations in the feature space, confirming the effectiveness and superiority of the proposed method.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"154 ","pages":"Article 103529"},"PeriodicalIF":3.9,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144902696","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}

引用次数: 0

Robust online identification for hybrid multirate systems based on recursive EM algorithm 基于递归EM算法的混合多速率系统鲁棒在线辨识

IF 3.9 2区计算机科学 Q2 AUTOMATION & CONTROL SYSTEMS

Journal of Process Control

Pub Date : 2025-09-01 Epub Date: 2025-07-29 DOI: 10.1016/j.jprocont.2025.103514

Fan Guo , Biao Huang

This paper focuses on robust identification for both linear time-invariant and time-variant multirate systems with time delays subject to outliers. The time delays are time varying and modeled by a Markov chain. Furthermore, the collected output data, which is corrupted by outliers, is described through a Laplace distribution. Parameters for the time-invariant model are estimated utilizing the batch expectation maximization (BEM) algorithm, whereas the recursive EM (REM) algorithm is employed for parameter estimation of the time-variant model. Upon receiving new data, the BEM first incorporates it in the historical batch data set and then iteratively recalculates parameter estimation using the updated data set. In contrast, the REM algorithm uses the parameter values obtained from the preceding step to recursively refine its estimates according to the new data sample. The efficacy of the proposed methods is demonstrated through a numerical example and a simulated continuous fermentation reactor process.

本文主要研究具有异常值的线性时不变和时变多速率系统的鲁棒辨识问题。时滞是时变的，用马尔可夫链建模。此外，收集的输出数据被异常值破坏，通过拉普拉斯分布进行描述。定常模型的参数估计采用批期望最大化算法（BEM），时变模型的参数估计采用递归EM （REM）算法。当接收到新数据时，BEM首先将其合并到历史批处理数据集中，然后使用更新后的数据集迭代地重新计算参数估计。相比之下，REM算法使用前一步获得的参数值，根据新的数据样本递归地改进其估计。通过数值算例和模拟连续发酵反应器过程验证了所提方法的有效性。

引用次数: 0

Optimal controlled variable switching for global self-optimizing control of active-set change processes 活动集变化过程全局自优化控制的最优控制变量切换

IF 3.9 2区计算机科学 Q2 AUTOMATION & CONTROL SYSTEMS

Journal of Process Control

Pub Date : 2025-09-01 Epub Date: 2025-07-31 DOI: 10.1016/j.jprocont.2025.103509

Lingjian Ye , Feifan Shen , Zeyu Yang , Xiaofeng Yuan

For global self-optimizing control (SOC) of active-set change processes, we propose two approaches for optimal switching of the controlled variables (CVs), namely, the descriptor function method and the partial switching method. The descriptor function method designs self-optimizing CVs for each critical region, the change of operating region is monitored by the descriptor function, such that whether to the switch CVs is decided. This method is an extension of the previous switching strategy based on the local SOC, but constructs the descriptor function using the generalized global SOC (g

^{2}

SOC) approach. In the second partial switching method, however, only part of the CVs that relate to varying active constraints are switched, while others are kept invariant over all critical regions, which are also solved using the g

^{2}

SOC approach. In this method, common max/min selectors are employed to automatically switch the CVs, whenever necessary. Finally, practical design procedure and optimality of the proposed switching methods are illustrated using three simulated examples.

针对活动集变化过程的全局自优化控制（SOC），提出了两种控制变量（cv）的最优切换方法，即描述函数法和部分切换法。描述函数法为每个关键区域设计自优化的cv，通过描述函数监测工作区域的变化，从而决定是否切换cv。该方法是先前基于局部SOC的切换策略的扩展，但使用广义全局SOC （g2SOC）方法构建描述符函数。然而，在第二种部分切换方法中，只有与变化的主动约束相关的部分CVs被切换，而其他CVs在所有关键区域保持不变，这也是使用g2SOC方法解决的。在这种方法中，常用的最大/最小选择器用于在必要时自动切换CVs。最后，通过三个仿真实例说明了所提出的切换方法的实际设计过程和最优性。

{"title":"Optimal controlled variable switching for global self-optimizing control of active-set change processes","authors":"Lingjian Ye , Feifan Shen , Zeyu Yang , Xiaofeng Yuan","doi":"10.1016/j.jprocont.2025.103509","DOIUrl":"10.1016/j.jprocont.2025.103509","url":null,"abstract":"<div><div>For global self-optimizing control (SOC) of active-set change processes, we propose two approaches for optimal switching of the controlled variables (CVs), namely, the <em>descriptor function method</em> and the <em>partial switching method</em>. The descriptor function method designs self-optimizing CVs for each critical region, the change of operating region is monitored by the descriptor function, such that whether to the switch CVs is decided. This method is an extension of the previous switching strategy based on the local SOC, but constructs the descriptor function using the generalized global SOC (g<span><math><msup><mrow></mrow><mrow><mn>2</mn></mrow></msup></math></span>SOC) approach. In the second partial switching method, however, only part of the CVs that relate to varying active constraints are switched, while others are kept invariant over all critical regions, which are also solved using the g<span><math><msup><mrow></mrow><mrow><mn>2</mn></mrow></msup></math></span>SOC approach. In this method, common max/min selectors are employed to automatically switch the CVs, whenever necessary. Finally, practical design procedure and optimality of the proposed switching methods are illustrated using three simulated examples.</div></div>","PeriodicalId":50079,"journal":{"name":"Journal of Process Control","volume":"153 ","pages":"Article 103509"},"PeriodicalIF":3.9,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144739222","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}

引用次数: 0