Friction stir welding is a relatively new way to join solid materials without melting using a nonconsumable tool, which has many applications in different industries including automotive, shipbuilding, and aerospace. ...
详细信息
Friction stir welding is a relatively new way to join solid materials without melting using a nonconsumable tool, which has many applications in different industries including automotive, shipbuilding, and aerospace. Destructive testing is an integral part of engineering science, which costs a lot. Reducing the number of destructive tests via numerical calculations to determine the quality of welded parts is valuable. On the other hand, advances in computer technology and embedded sensing systems in different domains have made it possible to collect a variety of data in huge volume at an unbelievable velocity, which provides an opportunity and at the same time a challenge to engineers and practitioners to utilize this rich source of information efficiently. Functional data as a rich form of structured data allows for high dimensionality modeling and analysis of the data. In this paper, we develop a fully functional linear regression model to quantify and predict the quality of the process outputs by reducing the number of destructive tests and presenting a change-point detection model to avoid using the model when a change has occurred in one of the components of the process. Important issues such as autocorrelation and correlation are taken into account in the presented model. The functional variables of the model are solved by polynomial basis function expansions. The results of the experimental tests indicate that the proposed method performs well in detecting out-of-control conditions as well as estimating the change-point location. The obtained value of the multiple correlation coefficient 0.98 and the corresponding F-value equal to 652.95 support these results.
As a core part of industrial production, the management of energy consumption (EC) in discrete manufacturing workshops (DMWs) is critical to improving productivity, reducing costs, and minimizing environmental impact....
详细信息
ISBN:
(纸本)9798350350319;9798350350302
As a core part of industrial production, the management of energy consumption (EC) in discrete manufacturing workshops (DMWs) is critical to improving productivity, reducing costs, and minimizing environmental impact. This paper provides an in-depth study of EC in DMWs, which is categorized into two main groups: work EC and public EC. The work EC includes basic energy, manufacturing energy and transportation energy, and each EC is analyzed and modeled in detail. Different analysis methods are proposed for the two levels of concern, process level and technological level. Based on these analyses, a shop-floor level energy calculation method based on simplified power curves and a data-driven power modeling approach are proposed for constructing accurate shop-floor energy models. Taking typical manufacturing and transportation equipment such as NC machine tools and AGVs as examples, the various power characteristics in the work EC are analyzed and the corresponding models are established. The experimental results show that a better power modeling accuracy can be obtained by using the data-driven method.
In this study, a novel nonlinear generalized predictive control (NGPC) method is proposed to tackle the tracking control problem which considers a mathematical model in conjunction with the virtual unmodeled dynamics ...
详细信息
In this study, a novel nonlinear generalized predictive control (NGPC) method is proposed to tackle the tracking control problem which considers a mathematical model in conjunction with the virtual unmodeled dynamics decomposition compensation technology. First, data-modeling technology is adopted to take advantage of the input and output data information from the process to decompose the virtual unmodeled dynamics into the form of posterior measurement unmodeled dynamics and an unknown increment. Then, feedforward compensator is designed. The essential difference between the proposed datamodeling algorithm and existing methods is that the effect of the measurement unmodeled dynamics is eliminated by the feedforward compensator. For the increment of the virtual unmodeled dynamics, a novel estimation algorithm is established through a data driven approach;A nonlinear compensator for the unknown increment of the virtual unmodeled dynamics is designed to suppress the effect of the increment of the virtual unmodeled dynamics on the closed-loop system. Second, the developed compensator is combined with the GPC algorithm to construct a nonlinear generalized predictive controller. Theoretical analysis proves that the proposed modeling and control algorithm have bounded-input bounded-output (BIBO) stability. Finally, an experimental study is conducted, which indicates the effectiveness of the proposed algorithm. (c) 2023 Published by Elsevier Ltd.
As an effective feature extraction method, slow feature analysis (SFA) has been successfully applied in the domain of chemical process monitoring. However, as an unsupervised method, the basic SFA ignores the key info...
详细信息
As an effective feature extraction method, slow feature analysis (SFA) has been successfully applied in the domain of chemical process monitoring. However, as an unsupervised method, the basic SFA ignores the key information carried by prior fault data so that many complicated faults cannot be sensitively detected. To address this issue, an enhanced SFA method, termed Information Enhanced SFA (IE-SFA), is proposed by integrating available fault discriminant information to achieve more sensitive detection of complicated chemical process faults. The proposed method builds a primary-auxiliary SFA modeling framework. On the one hand, the normal training data are analyzed to establish the primary SFA model. On the other, prior fault data are integrated to develop an auxiliary monitoring model for providing extra fault-sensitive information. In the auxiliary model, a sample-variable weighted Fisher discriminant analysis method is performed on normal data, fault data, and their respective slow feature components, thereby extracting fault-sensitive features for enhancing the fault detection capability. For full-view fault monitoring, the Bayesian fusion strategy is used to fuse the outcomes from both primary and auxiliary models. The applications on a benchmark Tennessee Eastman chemical process and a three-phase flow process demonstrate that the proposed IE-SFA method provides superior fault detection performance compared to the basic SFA method.
SENTINEL (SEnsor NeTwork INtelligent Emissions Locator) is an application developed in R Shiny to support emerging user groups of lower cost fenceline sensors, such as those monitoring volatile organic compound or met...
详细信息
SENTINEL (SEnsor NeTwork INtelligent Emissions Locator) is an application developed in R Shiny to support emerging user groups of lower cost fenceline sensors, such as those monitoring volatile organic compound or methane concentrations inside and near industrial facilities or for emergency response applications. During deployment, sensors collect a large quantity of high-frequency pollutant concentration data, time-aligned meteorological information, and sensor performance indicators. These sensors can collect a quantity of data that is overwhelming for users to process and understand without designated software. The SENTINEL application provides users with a consistent framework for processing, analyzing, and visualizing fenceline sensor data. SENTINEL temporally aggregates data for synthesized analysis and interpretation. Quality assurance screening automatically removes anomalous datapoints and a baseline correction algorithm reduces background drift in pollutant concentration data. SENTINEL offers streamlined sensor dataanalysis through a user-friendly graphical user interface that supports interpretation of source emission data and sensor-triggered field samples.
Electrochemical reactors play a critical role in various industrial sectors, including energy storage, chemical production, and environmental engineering. The complexity of these systems-arising from coupled electroch...
详细信息
Electrochemical reactors play a critical role in various industrial sectors, including energy storage, chemical production, and environmental engineering. The complexity of these systems-arising from coupled electrochemical reactions with mass, heat and charge transport phenomena-poses significant challenges in modeling, analysis, and control. Machine learning (ML) has emerged as a promising tool for addressing these challenges by providing data-driven solutions to complex processmodeling, optimization, and advanced control. This tutorial review explores the state-of-the-art applications of ML in electrochemical reactor systems, including ML-based modeling techniques and ML-based advanced control strategies, followed by the discussions of practical challenges and their solutions. An electrochemical carbon dioxide (CO2) reduction reactor is used as a demonstration example to show the effectiveness of various modeling and control methods. In addition, an integrated data infrastructure platform is presented for the digitalization and control of the electrochemical CO2 reduction reactor. By identifying current gaps and future opportunities, this article provides a roadmap for leveraging ML tools to improve the analysis and operation of electrochemical reactors.
Accurate quality variable inference by process variables is the core of industrial inferential sensor modeling, where recent advancements have seen deep learning (DL) models achieving remarkable success. However, inte...
详细信息
Accurate quality variable inference by process variables is the core of industrial inferential sensor modeling, where recent advancements have seen deep learning (DL) models achieving remarkable success. However, integrating knowledge of unit operations is critical for improving inferential sensor performance, yet it has received little attention. The main challenge lies in the incompleteness and correctness of industrial knowledge due to its semi-empirical nature and inevitable engineering errors. Addressing this, this article introduces the gradient knowledge network based on the graph neural network's message-passing mechanism within the variational Bayesian inference framework, which naturally copes with the abovementioned issues by fusing observational data. Initially, the prior knowledge about the process variables, which mirrors the graph in graph neural network, is parameterized as Dirichlet distribution based on the analysis of message-passing mechanism. However, the divergence computation and normalization constraints are challenging for model implementation. To navigate these challenges, the Bayesian inference problem is transformed into an optimization problem, subsequently recast as a simulation problem induced by the gradient field, ensuring compatibility with DL backends. Furthermore, a theoretical iteration equation is derived to maintain the normalization constraint. The architecture of the proposed model and its learning algorithm are then detailed. Finally, various experiments are conducted on two real industrial processes to demonstrate the model's efficacy from the perspective of prediction accuracy, sensitivity analysis, and ablation study.
Optical emission spectroscopy (OES) data is essential for virtual metrology, enabling accurate predictions of wafer performance in plasma etching processes. This approach not only reduces the need for physical measure...
详细信息
Optical emission spectroscopy (OES) data is essential for virtual metrology, enabling accurate predictions of wafer performance in plasma etching processes. This approach not only reduces the need for physical measurements of product quality, leading to significant resource savings, but also supports improved decision-making, particularly in processcontrol and quality assurance. To exploit the consecutive nature of OES data, we propose a prediction method based on a functional approach using multivariate functional partial least squares regression, coupled with dimension reduction and a novel outlier detection technique via functional independent component analysis. The proposed approach improves prediction performance by capturing the continuous nature of OES data and effectively extracting the components that describe the data structure. Numerical experiments, including simulation studies and real-world applications of OES data, demonstrate the effectiveness of the proposed method through superior prediction performance, as evidenced by low RMSE and MAE values, particularly in the presence of outliers.
The optimization and control of Vertical Roller Mill (VRM) circuits are critical for industrial processes, yet limited modeling has slowed progress in operator training, error minimization, and laboratory cost savings...
详细信息
The optimization and control of Vertical Roller Mill (VRM) circuits are critical for industrial processes, yet limited modeling has slowed progress in operator training, error minimization, and laboratory cost savings. To overcome limitations, the innovative "Conscious Lab" (CL) was introduced, utilizing industrial datasets and Explainable Artificial Intelligence (XAI) techniques. For the first time, CL combines Shapley Additive Explanations (SHAP) with machine learning models, such as XGBoost and Random Forest, to optimize VRM operations. Differential pressure and feed rate were identified as the most influential parameters of working pressure, essential for maintaining stable operations. Robust linear correlations (coefficients: 0.94 for feed rate, 0.84 for main drive power, and 0.83 for differential pressure) and nonlinear marginal plots (0.95, 0.81, and 0.81) highlighted how increases in these parameters significantly raise working pressure. The XGBoost model achieved remarkable prediction accuracy (0.99 for training and 0.98 for testing/validation) with a low RMSE (0.01), confirmed by 5-fold cross-validation. SHAP analysis further verified the relationship between working pressure and key parameters, aligning with VRM grinding principles. The CL approach introduces a data-driven control system enabling real-time decision-making, process optimization, and improved production efficiency, showcasing the transformative potential of advanced data analytics for industrial applications.
With recent improvements in computer performance, computer-aided studies have become increasingly important. Computer-aided methods have been applied in fat crystallization studies for modeling, simulation, optimizati...
详细信息
With recent improvements in computer performance, computer-aided studies have become increasingly important. Computer-aided methods have been applied in fat crystallization studies for modeling, simulation, optimization, dataanalysis and visualization. In this paper, various methods, such as molecular dynamic simulation, Monte Carlo, cellular automata modeling, finite element analysis, machine learning and computer vision, are introduced. Applications and advances in mechanism explanation, behavior prediction, process optimization, and so forth, are reviewed for fat crystallization. As a powerful and essential tool, computer-aided study should play an important role in the field of lipid research in the future.
暂无评论