Fundamental Techniques for Intelligent Control. Learning Strategies and Algorithms. System Modeling and Estimation. Dynamic Controls. Optimization Control Techniques. Multivariate Statistics and Quality Control. Fault Detection and Diagnosis. Appendix. Bibliography. Index.

Industrial intelligent control: fundamentals and applications

A review of planning and scheduling methods for hot rolling mills in steel production

The recent advances in artificial intelligence have already begun to penetrate our daily lives. Even though the development is still in its infancy, it has been shown that it can outperform human beings even in terms of intelligence (e.g., AlphaGo by DeepMind), implying a massive potential for its broader application in various industrial sectors. In particular, the growing public interest in industry 4.0, which focuses on revolutionizing the traditional manufacturing scene, has stimulated a deeper investigation of its possible applications in the related industries. Since it has several limitations that hinder its direct usage, research on the convergence of artificial intelligence with other engineering fields, including precision engineering and manufacturing, is ongoing. This overview looks to summarize some of the important achievements made using artificial intelligence in some of the most influential and lucrative manufacturing industries in hopes of transforming the manufacturing sites.

/pdf/recent-advances-of-artificial-intelligence-in-manufacturing-2hh9iwkbbd.pdf

Recent Advances of Artificial Intelligence in Manufacturing Industrial Sectors: A Review

In industrial electro galvanizing lines aged anodes deteriorate zinc coating distribution over the strip width, leading to an increase in electricity and zinc cost. We introduce a data-driven approach in predictive maintenance of anodes to replace the cost- and labor-intensive manual inspection, which is still common for this task. The approach is based on parasitic resistance as an indicator of anode condition which might be aged or mis-installed. The parasitic resistance is indirectly observable via the voltage difference between the measured and baseline (theoretical) voltage for healthy anode. Here we calculate the baseline voltage by means of two approaches: (1) a physical model based on electrical and electrochemical laws, and (2) advanced machine learning techniques including boosting and bagging regression. The data was collected on one exemplary rectifier unit equipped with two anodes being studied for a total period of two years. The dataset consists of one target variable (rectifier voltage) and nine predictive variables used in the models, observing electrical current, electrolyte, and steel strip characteristics. For predictive modelling, we used Random Forest, Partial Least Squares and AdaBoost Regression. The model training was conducted on intervals where the anodes were in good condition and validated on other segments which served as a proof of concept that bad anode conditions can be identified using the parasitic resistance predicted by our models. Our results show a RMSE of 0.24 V for baseline rectifier voltage with a mean ± standard deviation of 11.32 ± 2.53 V for the best model on the validation set. The best-performing model is a hybrid version of a Random Forest which incorporates meta-variables computed from the physical model. We found that a large predicted parasitic resistance coincides well with the results of the manual inspection. The results of this work will be implemented in online monitoring of anode conditions to reduce operational cost at a production site.

/pdf/parasitic-resistance-as-a-predictor-of-faulty-anodes-in-6ajjilmsx4.pdf

Parasitic resistance as a predictor of faulty anodes in electro galvanizing: a comparison of machine learning, physical and hybrid models

Feedforward control of the transverse strip profile in hot-dip galvanizing lines

The hot-dip galvanizing line (HDGL) is a typical order-driven discrete-event process in steelmaking. It has some complicated dynamic characteristics such as a large time-varying delay, strong nonlinearity, and unmeasured disturbance, all of which lead to the difficulty of an online coating weight controller design. We propose a novel neural network based control system to solve these problems. The proposed method has been successfully applied to a real production line at VaLin LY Steel Co., Loudi, China. The industrial application results show the effectiveness and efficiency of the proposed method, including significant reductions in the variance of the coating weight and the transition time.

Development and application of a neural network based coating weight control system for a hot-dip galvanizing line

Making an integrated schedule on both continues casting (CC) and hot rolling (HR) process in steel-making lines is a difficult problem since the balance of material flow, the continuity of production, the hot charge rate as well as the coupling between successive stages should be taken into consideration together. Therefore, few approaches have been published to solve the multistage integrated scheduling problem and applied to the real systems in practice. In this study, a novel modified Extremal Optimization (EO) algorithm combining exact mathematical model on CC stage and a heuristic algorithm on HR stage is proposed to solve the scheduling problem with less computational effort, and has been successfully applied to a real plant. The industrial application results show that compared with manual scheduling, considerable improvement in hot charge rate is achieved, leading to large energy-saving for the steel maker.

Application of extremal optimization approach to the integrated scheduling problem of continuous casting and hot rolling process

To deal with the difficulties of large time-varying delay, seriously nonlinear and strong disturbance in the hot-dip galvanizing line (HDGL) control, this paper proposes a novel ANN based modelling and prediction method for the coating weight, the proposed method is then applied to a real plant at Valin LY Steel Co., Loudi, China. The prediction model is built through data pre-processing, data clustering and model training, with the model deviation being corrected through bias-update. The comparisons between measured coating weight and model prediction of areal HDGL process show the effectiveness of the proposed method, and the rationality of the prediction model is further verified though gain analysis between major input and output of HDGL process.

Neural network based modelling and prediction for the coating weight of HDGL process

In this research, we employ LOGIT model, which is widely used in statistical field, to describe the individual decision-making process instead of shortest path rule in transportation network models. With the idea of congestion effects, we can derive a stationary flow distribution solution analytically under any given network structure which is solved by a numeric iterative approach approximately. Based on that, simulations are made to compare the cumulative the difference of load distribution under shortest path law and our LOGIT law. Meanwhile, four different classes of networks are tested to uncover the relationship between traffic flow and topological structures under LOGIT law. The results indicated that random network and scale-free network outperform than the others two in the measurement of TSC.

Characteristic of transportation network with LOGIT model

Based on the linear dynamic model(LDN) on urban traffic system proposed by Lucas Barcelos et al. in 2010, the MPC method has been successfully used in the signal splitting control area. In this paper, we extend the model to a nonlinear form employing the idea from S Lin et al. which is close to the real world, and the Gradient-based distributed dynamic optimization method (GBDDO) is introduced to distributively solving the large-scale problem. The simulation results show that our method can derive an approving solution.

Zaisheng Pan

Papers

Development and application of a neural network based coating weight control system for a hot-dip galvanizing line

Application of extremal optimization approach to the integrated scheduling problem of continuous casting and hot rolling process

Neural network based modelling and prediction for the coating weight of HDGL process

Characteristic of transportation network with LOGIT model

A novel distributed nonlinear MPC algorithm of signaling split in urban traffic system