Journal of Process Control最新文献

Model predictive control (MPC) allows for dealing with multivariable interactions, known future changes and dynamic satisfaction of constraints. Standard MPC has a cost function that aims at keeping selected controlled variables at constant setpoints. This work considers systems where the steady-state optimal active constraints change during operation. This situation is not handled optimally by standard MPC which uses fixed controlled variables for the unconstrained degrees of freedom. We propose a simple framework that detects the constraint changes and updates the controlled variables accordingly. The unconstrained controlled variables are chosen to be the reduced cost gradients, which when controlled to zero minimizes the steady-state economic cost. In this paper, the nullspace method for self-optimizing control is used to estimate the cost gradient using a static combination of the measurements. This estimated gradient is also used for detecting the current set of active constraints, which in particular allows for giving up constraints that were previously active. The proposed framework, here referred to as “region-based MPC”, is shown to be optimal for linear constrained systems with a quadratic economic cost function, and it allows for good economic performance in nonlinear systems in a neighborhood of the considered design points.

Recently, deep learning algorithms have been successfully applied to industrial fault detection because they are better at automatically extracting complex features and processing high-dimensional data than traditional methods. However, most existing deep learning-based fault detection methods only concentrate on extracting features from industrial process data without considering the crucial long-term temporal features and higher-order statistical information. To address this challenge, we proposed a novel enhanced higher-order pooling-based network (EHOPN) for industrial fault detection. First, the data pre-processing of the network is presented to capture the dynamic features of time-series process data and unify the high-dimensional data scale. Second, the EHOPN utilizes channel and temporal second-order pooling techniques to gather temporal and channel statistics information, facilitating the backbone network’s ability to capture complex inter-dependencies and long-term dynamics. Additionally, the high-order feature aggregation module is presented to aggregate global and local features, enhancing the network’s generalization ability. The proposed industrial fault detection approach is evaluated on the Tennessee Eastman benchmark and a real-world heavy-plate production process. Experimental results show that the proposed method is significantly better than comparison models in four evaluation metrics: accuracy, precision, recall, and F1-score, further proving the effectiveness of EHOPN.

Alarm floods are leading issues that compromise the efficiency of industrial alarm systems and are identified as major causes of many industrial accidents. As an advanced technique to handle alarm floods, sequence alignment based similarity analysis has been developed to match alarm flood sequences, and thus can help with further root cause identification and early warning of alarm floods. However, existing methods based on biological sequence alignment algorithms ignore the relations between alarm occurrences, and thus may cause incorrect matches or mismatches of alarms when comparing two flood sequences. Accordingly, this paper proposes a new alarm flood similarity analysis method based on global vectors and Move–Split–Merge (MSM) distance. The contributions are mainly twofold: (1) An alarm encoding model based on modified global vectors is devised to convert alarm sequences into numerical vectors that reflect the correlations of alarms; (2) a similarity analysis method based on the modified MSM distance is proposed for comparison of encoded alarm flood sequences of unequal lengths. The effectiveness of the proposed method is demonstrated through a case study with a publicly accessible industrial model for Vinyl Acetate Monomer.

State estimation for the Markov jump linear system (MJLS) is a intractable task when the unpredictable measurement loss exists. Although the conventional methods, such as interacting multiple-model method, are widely used in MJLS, their performance still depends on the known transition probability matrix (TPM). In this article, a novel adaptive state estimation method is proposed for MJLS with unknown measurement loss and TPM based on variational Bayesian inference. Specifically, under system state dynamic and measurement loss are independent, the system state, measurement loss probability and TPM are jointly inferred. In particular, when the stochastic measurement loss occurs, a selective learning mechanism is used to the updating of TPM. The efficiency and superiority of the proposed method is verified by a numerical example and a fermenter process compared with the existing methods.

Presently, the demand for precise process monitoring during geological drilling has increased dramatically. However, there exists complex dynamic characteristics due to the various forms of changes in operation conditions. A large number of false alarms are usually triggered when using the conventional static-based monitoring methods. In this paper, two types of dynamic behaviors are comprehensively considered, including the dynamic behaviors caused by the operating parameters adjustment and the operating mode switching, and then, a full condition monitoring method is proposed for the drilling process based on just-in-time learning (JITL)-aided slow feature analysis (SFA). On one hand, the JITL local modeling strategy is improved and adopted to deal with the dynamic behavior due to the operating mode switching. Specifically, a sequence spatiotemporal similarity analysis method is developed to improve the local modeling performance. On the other hand, the SFA-based concurrent monitoring of static deviations and dynamic anomalies is realized to cope with the dynamic behavior due to the operating parameters adjustment. Several industrial cases based on actual drilling data are conducted, which illustrate the effectiveness and superiority of the proposed method.

Process monitoring detects faults and issues alerts when faults occur. It has become an integral part of ensuring the safety and quality of industrial processes. Existing mainstream process-monitoring methods often separate data from knowledge, forming distinct systems. However, data and knowledge exhibit complementary characteristics, and using them together can contribute to enhancing monitoring performance. Furthermore, the importance of fault data has not been adequately emphasized. Within this fault data, valuable fault features contribute significantly to process monitoring. In light of these considerations, we propose a process-monitoring method based on temporal knowledge graphs and supervised contrastive learning,which can fully use knowledge, data, and fault information to improve the monitoring performance of the model. First, a temporal knowledge graph is constructed, in which knowledge and data are organically integrated through qualitative knowledge and quantitative data calculations to enhance the interpretability and accuracy of the graph. Second, spatiotemporal features are extracted from the temporal knowledge graph at multiple levels through differentiable graph pooling. Finally, a monitoring statistic is constructed, and fault information is introduced into the statistic through supervised contrastive learning, using fault information to enhance monitoring performance of the model. The fault detection rate on the float-glass production process reaches 95%.

A new robust nonlinear model predictive control (RNMPC) heuristic is proposed, specifically developed to be i) easy to implement, ii) robust against constraint violations and iii) fast to solve. Our proposed heuristic samples from the disturbance distributions and performs $n$-steps-ahead Monte Carlo (MC) simulations to calculate the back-off where $n$ is a small number, typically one. We show two implementations of our heuristic. The Automatic Back-off Calculation NMPC (ABC-NMPC) uses MC simulations on a process model to calculate the back-off, and explicitly states the back-off in a standard NMPC problem. Our second implementation, the MC Single-Stage NMPC (MCSS-NMPC), directly includes the disturbance distribution in the optimization problem, making it an implicit back-off method. Our methods are robust against constraint violation in the next time-step, under certain assumptions. In the presented case-study, our proposed RNMPC methods outperform the popular multi-stage NMPC in terms of robustness and/or computational cost. We suggest several further modifications to our RNMPC methods to improve performance, at the cost of increased complexity.

The application of multivariate statistical analysis in process monitoring has emerged as a significant research topic, with a focus on consideration of data correlations. The present study investigates an anomaly detection method based on autoregressive double latent variables probabilistic (ADLVP) model for industrial dynamic processes. Specifically, the ADLVP model incorporates two distinct types of latent variables (LVs) to capture the internal relationships within the data from both quality-correlated and uncorrelated perspectives. Moreover, the model employs autoregressive modeling to characterize the double latent variables with time-dependence, enabling them to unveil more intricate higher-order autocorrelations between industrial measurements. The model parameters and the double latent variables can be iteratively determined using the expectation maximization (EM) algorithm, upon which the statistics for process monitoring are devised. Finally, the proposed method is validated in two industrial studies, and experimental results demonstrate that the ADLVP model outperforms its counterparts in dynamic processes monitoring.

Zero-shot learning (ZSL) can diagnose unseen faults without corresponding training data, which has aroused the researchers’ interest. However, a prevailing challenge in most existing ZSL approaches is their limited effectiveness in distinguishing similar unseen faults. This paper proposed a category-tree-guided hierarchical knowledge transfer zero-shot fault diagnosis (CTZSD) method, which is a coarse-to-fine zero-shot fault diagnosis framework to alleviate this problem. To embody the similarities between fault categories, the concept of fault category tree is proposed, for which a data-attribute collaborative tree construction mechanism (DATC) is designed. Rather than relying solely on semantic knowledge, DATC involves data, which carries richer information, to complement the category similarity evaluation. A hierarchical knowledge transfer zero-shot fault diagnosis mechanism (HKT) is subsequently developed, utilizing the established category tree to gradually narrow down the options, thereby promoting the recognition of similar unseen faults. The mechanism treats the diagnostic outcomes and model parameters from coarse-grained tasks as knowledge and transfers them to fine-grained tasks for guidance, realizing a coarse-to-fine diagnosis. Aiming at providing discriminative information to further distinguish similar unseen faults, attention modules are integrated within HKT. These modules assess attribute weight, thereby directing the model’s focus toward the discriminative attributes of similar unseen faults. Experiments on a real TPP industrial process demonstrate that the proposed CTZSD outperforms other traditional ZSL methods in distinguishing similar unseen faults, improving the average accuracy by at least 19.7%.

The accurate soft sensing of f-CaO content in cement clinker is crucial for the cement industry. However, existing methods require improvements in terms of effectiveness, practicality, and computational efficiency for industrial applications. Responding to these needs, this paper proposes a self-adaptive multisource information fusion framework (SA-MSIFF) for f-CaO content soft sensing. The SA-MSIFF utilizes a dynamic rotary kiln model for independent and real-time calcination state estimation and mechanistic feature generation, along with a dilated 3D convolution and attention-based network for direct feature extraction from flame image sequences. Subsequently, a temporal–spatial feature extraction and fusion (TSFE&F) network is introduced to utilize the multisource feature series for f-CaO content soft sensing. Offline experiments validate the SA-MSIFF’s ability to adaptively extract features from multisource information. Compared to its previous version, MSIFF, the SA-MSIFF achieves a considerable 89.65% reduction in framework training time and an 8.22% decrease in soft sensing error. The SA-MSIFF’s effectiveness is also demonstrated in its engineering applications.