Improved shape control performance of a Sendzimir mill using wavelet radial basis function network and fuzzy logic actuator

Abstract

A shape control system based on a wavelet radial basis function network for a Sendzimir mill (ZRM) and fuzzy control are developed to improve the shape control performance of a conventional ZRM system. The conventional shape recognition system for a ZRM adopted an incomplete multi-layer perceptron neural network system that was constructed two decades ago. The poor shape recognition of this system leads to actuator saturation and shape control performance deterioration. Therefore, the full automatic operation of a ZRM is often stopped, and manual input need to be performed. This affects the quality, causes a decline in the productivity of the steel strip and an unnecessary waste of manpower. In this paper, a wavelet radial basis network is developed to replace the multi-layer perceptron network and consequently improve shape recognition performance. A modified fuzzy controller is also constructed to prevent actuator saturation that occurs in a conventional shape control system owing to the use of a fixed gain-based fuzzy controller. A comparative simulation based on the data measured from an actual ZRM plant demonstrates the efficacy of the proposed shape control system.

Keywords

Sendzmir mill shape control multi-layer perceptron neural network wavelet radial basis function network fuzzy controller

Get full access to this article

View all access options for this article.

References

Grimble

Fotakis

. The design of strip shape control systems for Sendzmir mills. IEEE Trans A C 1982; 27(3): 656–666.

Gunawardede GWDM. Static model development for the Sendzimir cold rolling mill. Sheffield City Polytechnic, Sheffield, 1982.

Ringwood

. Shape control systems for Sendzimir steel mills. IEEE Trans Control System Tech 2000; 8(1): 70–86.

Matsushita T, Ogawa M and Ogino T. Nippon stainless technology report, vol. 17, 1982.

Larkiola

Myllykosi

Korhonen

. The role of neural networks in the optimization of rolling processes. J Mater Process Tech 1998; 80: 16–23.

Peng F, Hu YR and Ng A. Tests of inflatable structure shape control using genetic algorithm and neural network. In: 46th AIAA/ASME/ASCE/AHS/ASC structures, structural dynamics and materials conference, Austin, Texas, 2005, pp.1–9.

Jung

Kwang

. Fuzzy-control simulation of cross-sectional shape in six-high cold rolling mills. J Mater Process Tech 1996; 62(1): 61–69.

Jung

. Simulation of fuzzy shape control for cold-rolled strip with randomly irregular strip shape. J Mater Process Tech 1997; 63(1): 248–253.

Hattori S, Nakajima M and Katayama Y. Hitachi Review, vol. 41, 1992.

10.

Hattori S, Nakajima M and Morooka Y. Hitachi Review, vol. 42, 1993.

11.

Zhang

Walter

Lee

WNW

. Wavelet neural networks for function learning. IEEE Trans Signal Proc 1995; 43(6): 1485–1497.

12.

Chen

Yang

Dong

. Time-series using a local linear wavelet neural network. Neurocomputing 2006; 69: 449–465.

13.

Park

Sandberg

. Approximation and radial-basis-function networks. Neural Comput 1993; 5(2): 305–316.

14.

Sundararajan

Saratchandran . Performance evaluation of a sequential minimal radial basis function (RBF) neural network learning algorithm. IEEE Trans Neural Netw 1998; 9(2): 308–318.

15.

Pearson

WKJ

. Shape measurement and control. J Inst Metals 1964; 93: 1964–1965.

16.

Kumar

Ravi

Carr

. Software development cost estimation using wavelet neural networks. J Syst Softw 2008; 81(11): 1853–1867.