Journal of Process Control最新文献

Deep neural networks (DNNs) can result in suboptimal monitoring performance due to nonlinearity, dynamics, and local characteristics in modern complex industrial processes. To surmount these limitations, this paper first proposes a novel data construction method to model the short-term autocorrelation and spatial correlations as a three-dimensional matrix and then reorder the elements of it to better encode the local and temporal structures. Subsequently, we design a new structure called Long-range Discriminative Attention (LDA) based on the self-attention mechanism to enlarge the receptive field of the original convolutional neural networks (CNNs) to extract global features. Finally, we propose a monitoring model named Long-range Discriminative Attention Autoencoder (LDCA) based on LDA to extract structural features between long-range and local variables from the constructed matrix. The effectiveness of the method in fault detection is verified by numerical examples and a three-phase flow process.

The complexity of the industrial processes, large-scale plants and the massive use of distributed control systems and sensors are challenges which open ways for alternative monitoring systems. The decentralized monitoring methods are one option to deal with these complex challenges. These methods are based on process decomposition, i.e., dividing the plant variables into blocks, and building statistical data models for every block to perform local monitoring. After that, the local monitoring results are integrated through a decision fusion algorithm for a global output concerning the process. However, decentralized process monitoring has to deal with a critical issue: a proper process decomposition, or block division, using only available data. Knowledge of the plant is rarely available, so data-driven approaches can help to manage this issue. Moreover, this is the first and key step to developing decentralized monitoring models and several alternative approaches are available. In this work a comparative study is carried out regarding decentralized fault monitoring methods, comparing several alternative proposals for process decomposition based on data. These methods are based on information theory, regression and clustering, and are compared in terms of their monitoring performance. When the blocks are obtained, CVA (Canonical Variate Analysis) based local dynamic monitors are set up to characterize the local process behavior, while also considering the dynamic nature of the industrial plants. Finally, the Bayesian Inference Index (BII) is implemented, based on these local monitoring, to achieve a global outcome regarding fault detection for the whole process. To further compare their performance from the application viewpoint, the Tennessee Eastman (TE) process, a well-known industrial benchmark, is used to illustrate the efficiencies of all the discussed methods. So, a systematically comparison have been carried out involving different data-driven methods for process decomposition to implement a decentralized monitoring scheme. The results are focused on providing a reference for practitioners as guidelines for successful decentralized monitoring strategies.

This paper presents the implementation of the pressure control system for the vacuum vessel of Damavand Tokamak. PID controllers within the framework of multiple-model control are utilized for controller design, aiming to safely achieve the desired setpoint for the pressure of the vacuum vessel. The chamber pressure is measured in real-time using a Cold Cathode Pirani gauge and transferred as feedback to the controller. There is a gas injection system to adjust the chamber pressure. Multiple process models are derived for the vacuum vessel pressure based on a system identification approach using the experimental data from the process. The derived models are employed in designing the PID controllers. The designed controllers are implemented on the tokamak gas injection system. The experimental results demonstrate that the designed controllers effectively track the desired pressure profile.

Proper maintenance and management of equipment are essential for producing high-quality wafers. Anomalies in equipment lead to the production of low-quality wafers. This study proposes a method to maintain and manage etching equipment in semiconductor manufacturing utilizing a virtual metrology (VM) model. Leveraging acquired equipment data, the VM model predicts electrical resistance measurement values to monitor the equipment state. Engineers determine the equipment state by inspecting the electrical resistance values and consistency of variance in the measurement data derived from the VM model. However, conventional complex machine learning models frequently yield predicted values with limited variability, making it challenging to detect abnormal equipment states. To address this issue, we propose a novel method, double bagging trees with weighted sampling, which guarantees the predicted values follow a proper distribution and exhibit a variance that aligns with the actual measurement values. The proposed method provides reliable predictions about the equipment state. A case study utilizing a real-world semiconductor manufacturing dataset is presented to demonstrate the effectiveness of the proposed approach. The VM model provides timely information about the state of equipment, which helps engineers maintain and manage equipment efficiently.

Electrolytic hydrogen production serves as not only a vital source of green hydrogen but also a key strategy for addressing renewable energy consumption challenges. For the safe production of hydrogen through Alkaline water electrolyzer (AWE), dependable process monitoring technology is essential. However, random noise can easily contaminate the AWE process data collected in industrial settings, presenting new challenges for monitoring methods. In this study, we develop the variational Bayesian sparse principal component analysis (VBSPCA) method for process monitoring. VBSPCA methods based on Gaussian prior and Laplace prior are derived to obtain the sparsity of the projection matrix, which corresponds to ${\ell }_2$ regularization and ${\ell }_1$ regularization, respectively. The correlation of dynamic latent variables is then analyzed by sparse autoregression and fault variables are diagnosed by fault reconstruction. The effectiveness of the method is verified by an industrial hydrogen production process, and the test results demonstrated that both Gaussian prior and Laplace prior based VBSPCA can effectively detect and diagnose critical faults in AWEs.

In a hot strip rolling mill (HSRM) process, the prediction of the steel crown is a key factor in improving the quality of the strip steel. In this paper, a new multibatch feature extraction-based method is proposed for predicting the steel crown. Different from the cascaded feature extraction-based method which cannot extract both temporal and local features well, this method parallelly captures the feature between different batches of data using a method based on the multi-channel convolution neural network (MCNN) and long short-term memory (LSTM). The feature extraction is performed in parallel by an LSTM layer fusing variable attention and temporal attention, and a Multi-channel convolutional neural network fusing channel attention and spatial attention, which are used to extract temporal and local features of the input variables, respectively. Then, an LSTM-based fusion layer is used to incorporate both features for the development of the prediction model. The proposed method is applied to a cloud–edge-end collaborative prototype system, where the actual HSRM data is integrated. Based on the fact that an HSRM process commonly runs with the steel header crown data for the model update, an adaptive prediction method is also developed and deployed in the prototype system. It can be seen from the model complexity analysis and application results that the prediction performance improves by 42.70% compared with the cascaded feature extraction-based method, and the adaptive method can ensure a realtime prediction realization.

Variable rate application of pesticides in agriculture can improve pest control and also increase food production. Nevertheless, incorrect spraying poses risks to the environment and human health, as well as may increase the total cost of production. Nowadays, it is quite known the importance of innovation in techniques and technologies to improve the spraying process in a variable rate application for pest control. This work presents a generalized predictive control (GPC) strategy to cope with nonlinearities. An extension of the stability analysis for constrained GPC controller for infinite horizons is also developed, which guarantees the stability of a fuzzy GPC for a limited variation of its tuning parameters $\lambda$ and $\delta$. Results show the usefulness in adding a fuzzy logic system, to cope with nonlinearities, leading to a conception of an advanced fuzzy GPC for a limited variation of its tuning parameters.

Modern glucose sensors deployed in closed-loop insulin delivery systems, so-called artificial pancreas use wireless communication channels. While this allows a flexible system design, it also introduces vulnerability to cyberattacks. Timely detection and mitigation of attacks are imperative for device safety. However, large unknown meal disturbances are a crucial challenge in determining whether the sensor has been compromised or the sensor glucose trajectories are normal. We address this issue from a control-theoretic security perspective. In particular, a time-varying Kalman filter is employed to handle the sporadic meal intakes. The filter prediction error is then statistically evaluated to detect anomalies if present. We compare two state-of-the-art online anomaly detection algorithms, namely the ${\chi }^2$ and CUSUM tests. We establish a robust optimal detection rule for unknown bias injections. Even if the optimality holds only for the restrictive case of constant bias injections, we show that the proposed model-based anomaly detection scheme is also effective for generic non-stealthy sensor deception attacks through numerical simulations.

This work considers to achieve near-optimal operation for a class of batch processes by employing self-optimizing control (SOC). Comparing with a continuous one, a batch process exhibits stronger nonlinearity with dynamics because of the non-steady operation condition. This necessitates a global version of SOC to achieve satisfactory performance. Meanwhile, it also makes the existing global SOC (gSOC) not directly applicable to batch processes due to the causality amongst variables. Therefore, it is necessary to extend the original gSOC to batch processes. In addition to the nonconvexity challenge of the original gSOC problem, the new extension for batch processes has to face even more challenges. Particularly, the causality due to dynamics of batch processes brings in structural constraints on controlled variables (CVs), making a CV selection problem even more difficult. To address these challenges, the gSOC problem is recast in a vectorized formulation and it is proved that the structural constraints considered are linear in the vectorized formulation. Moreover, a novel shortcut method is proposed to efficiently find sub-optimal but more transparent solutions for this problem. The effectiveness of the new approach is validated through a case study of a fed-batch reactor, where CVs are constructed through a combination matrix with a repetitive structure, resulting in a simple SOC scheme. This simplicity facilitates the implementation of the SOC approach and enhances its practical applicability and robustness.

The paper considers the problem of optimally filling a Hele-Shaw cell. The system is subject to viscous fingering effect. It is shown that, despite the threshold terms appearing on the right-hand side of the governing equations, the dynamics can be rewritten using several prime integrals. This allows reforming optimal control problems for the Fourier modes describing the fluid interface into smooth optimization problems, in the sense of Gâteaux derivative. Some numerical experiments illustrate the advantages of using the optimal solutions obtained using this reformulation instead of the currently known time-dependent injection rates.