Kazuya Asano

Hot strip mill tension–looper control based on decentralization and coordination

Abstract Since tension and looper control is the key to successful operations in the hot strip finishing mill, several multivariable controllers have been applied. However, this leads to complicated controllers, which are hard to tune on site. This paper presents a new control method based on decentralization and coordination. First, interactions between tension and the looper are evaluated to ensure the applicability of decentralized control. Then, a decentralized controller based on the two-degree-of-freedom Internal Model Control (IMC) structure is applied. Moreover, impedance control is introduced to retrieve the coordinated action of the looper. The control scheme enables stepwise commissioning of control systems, which is very helpful for on-line tuning.

Application of moving horizon estimation based fault detection to cold tandem steel mill

For a class of model uncertainty descriptions, plant/model mismatch can be directly incorporated into a model based fault detection scheme using Moving Horizon Estimation. The model uncertainty is represented by a set of bounded parameters which can be used to alter the dynamics of the model through injection of the measured output as well as inputs. The uncertainty class includes gain uncertainty as well as uncertainty in pole locations. Using a bank of filters, detection of exclusive fault scenarios can be accomplished. The proposed method is compared to other methods employing an adaptive threshold, and is demonstrated on a simulation example of a cold tandem steel mill.

Interaction measure of tension-thickness control in tandem cold rolling

Abstract Tandem cold rolling should be considered a multi-input–multi-output process, but it has generally been controlled by a series of single-input–single-output control loops. To meet tight tolerances for product quality, multivariable control has been applied to the process. However, this inevitably leads to more complicated controllers, which are hard to tune on site. Decentralized control can trade off between simplicity and performance deterioration of control systems. In this paper, interactions between interstand tension and thickness control in tandem cold rolling are investigated, using the structured singular value to determine the structure of decentralized control for the process.

Robust molten steel level control for continuous casting

This paper presents a new control method for stabilizing the molten steel level in the molding process of a continuous casting mill. A disturbance observer is employed to estimate a disturbance flow which breaks the equilibrium of mass flow in the mold. Then, the manipulated value is calculated to counterbalance the effect of the estimated disturbance flow. This control scheme enables a straightforward and intuitively comprehensible way of designing and tuning controllers, in addition to high performance. The control method has been successfully applied to the molten level control systems in several continuous casting mills.

A HYBRID SYSTEM APPROACH TO TENSION CONTROL IN HOT ROLLING

Abstract Since tension and looper control is the key to successful operations in the hot strip finishing mill, several multivariable controllers have been proposed for it. These controllers, however, can be applied only after the transient response in the start-up phase of each rolling has settled. This paper presents a new control design method for tension control in the start-up phase. A hybrid system approach is taken to describe the discontinuous dynamics of the system and to optimize the tension and looper trajectories simultaneously throughout the start-up phase. Simulation results are shown to compare the control performance with the conventional ad hoc control.

Neural Network Model for Recognition of Characters Stenciled on Slabs

Abstract The No. 3 continuous casting plant at Chiba Works of Kawasaki Steel Corporation identifies each slab by stenciling a code on the surface of each slab. The stenciled characters are occasionally degraded by blurring or spots of paint and may become unreadable for operators who handle the slabs. In order to detect such defects of characters and to assure the quality of the identification codes, a character recognition system has been developed by using a three-layer neural network model. The characters are extracted from the image taken by MOS cameras by means of a pattern matching method. After preprocessing, the images of each character are provided to the neural network model which produces a criterion of similarity between the characters and the standard patterns. An alarm is given to operators when the quality of characters has been judged to be defective by the neural network model. The correct classification rate of the characters is more than 99%, and the rate of overlooking defective characters is 0%. The system has been successfully used for the quality assurance of slab marking.

A MPC for Start-up Phase Tension and Looper Control in Hot Strip Finishing Mills Using Continuation Approach

Abstract TThis paper gives a design method for a model predictive control (MPC) approach based on a unified performance index throughout the start-up phase tension and looper control which consists of the non-contact and contact modes in order to suppress the deviation of the strip tension while the looper contacts with the strip as quickly as possible. We will formulate the control problem by using a MPC for a piecewise affine (PWA) system with the terminal condition and an unknown terminal time. However, in order to realize the feedback control using a receding horizon strategy, we have to solve nonlinear equations in an on-line manner as precisely as possible. Therefore, the paper gives a method using a continuation method for solving the nonlinear equations efficiently. The efficiency of the proposed method is shown through numerical simulations.

Interaction Measure of Interstand Tension and Thickness Control in Tandem Cold Rolling Based on the Structured Singular Value

Abstract Tandem cold rolling should be considered a multi-input multi-output process, but it has been controlled by a series of single-input single-output control loops. To meet tight tolerances for product quality, multivariable control has been applied to the process. However, it inevitably leads to more complicated controllers which are hard to be tuned. Decentralized control can trade off between simplicity and performance deterioration of control systems. In this paper, interactions between interstand tension and thickness control in tandem cold rolling are investigated using the structured singular value to detennine the structure of decentralized control for the process.

Mold level stabilization using a standing wave model in continuous casting

Abstract In continuous steel casting plants, it is very important to stabilize the molten steel level in the mold for quality of the final products and productivity. Standing waves in the mold cause periodic molten steel level fluctuations, which can deteriorate the product quality and productivity. Nevertheless, there has not been any effective countermeasure for them, because they are different from other disturbances in the sense that they are not changes in the flow rate of the molten steel and hence should not be controlled by manipulating the mass flow. This paper proposes a standing wave model and a mold stabilization technique that can prevent harmful influence of standing waves. This technique can prevent interactions between the mold level control and the standing waves in the mold by removing standing wave components from the original mold level signal. On-line experimental results have shown good performance and this technique has been in practical use in a continuous steel casting process in JFE.

HPN modeling, optimization and control law extraction for continuous steel processing

This paper presents a new modeling and controller design technique for the steel sheet processing line based on Hybrid Petri Nets (HPN), Mixed Logical Dynamical Systems (MLDS) expressions and system identification techniques for hybrid systems. In the steel sheet processing line, the dynamics of sheets is represented by differential equation. On the other hand, some operations such as welding of the steel sheets are described by logical formulas. HPN can harmoniously integrate these two different aspects. The HPN expression can be transformed into a suitable formulation for numerical optimization. Based on this modeling, the closed-loop control can be realized by receding horizon scheme. At the next step, control laws are extracted from the results obtained by receding horizon scheme. The method identifies probability weighted ARX (PrARX) functions as a controller. Some simulations are conducted to demonstrate that the identified controller can achieve fast and near-optimal control.