Improving lookup table control of a hot coil strip process with online retrainable RBF network

This paper presents an online retrainable radial basis function (RBF) network to control the coiling temperature for a hot coil strip at the Pohang Iron and Steel Company, Pohang, Korea. The proposed RBF network is designed to replace the conventional rule-based lookup table, the output of which is a heat transmission coefficient in the temperature control system. In order to make the controller more adaptable to the changing environments in the steelmaking process, specific interconnection weights were additionally devised for the hidden-to-output weights of a conventional RBF network. These weights were locally adjustable to reduce the immediate temperature error of a coil strip, while the global information of the RBF network trained with offline past data was largely unaltered. As a result, the proposed RBF network substantially alleviated the effect of catastrophic interference-completely forgetting old information in the presence of new inputs. Moreover, a rejection network was incorporated within the proposed control scheme to ensure reliable operation in the actual process. Results applied to the real steelmaking process indicated an improvement of 2.2% in control performance compared to conventional methods.

[1]  Joydeep Ghosh,et al.  Scale-based clustering using the radial basis function network , 1996, IEEE Trans. Neural Networks.

[2]  Sukhan Lee,et al.  A Gaussian potential function network with hierarchically self-organizing learning , 1991, Neural Networks.

[3]  Sungzoon Cho,et al.  Reliable roll force prediction in cold mill using multiple neural networks , 1997, IEEE Trans. Neural Networks.

[4]  Anthony V. Robins,et al.  Catastrophic Forgetting, Rehearsal and Pseudorehearsal , 1995, Connect. Sci..

[5]  Teuvo Kohonen,et al.  The self-organizing map , 1990 .

[6]  Witold Pedrycz,et al.  Conditional fuzzy clustering in the design of radial basis function neural networks , 1998, IEEE Trans. Neural Networks.

[7]  Stephen Grossberg,et al.  Competitive Learning: From Interactive Activation to Adaptive Resonance , 1987, Cogn. Sci..

[8]  Kurt Hornik,et al.  Approximation capabilities of multilayer feedforward networks , 1991, Neural Networks.

[9]  Marios M. Polycarpou,et al.  An analytical framework for local feedforward networks , 1998, IEEE Trans. Neural Networks.

[10]  Robert M. French,et al.  Pseudo-recurrent Connectionist Networks: An Approach to the 'Sensitivity-Stability' Dilemma , 1997, Connect. Sci..

[11]  Wei Gao,et al.  Measurement and control of rolling of a precision moving table , 1997, 1997 IEEE International Conference on Intelligent Processing Systems (Cat. No.97TH8335).

[12]  Wang Qi,et al.  Multilayered mathematical model system for hot continuous rolling process control and its application in real process , 1997, 1997 IEEE International Conference on Intelligent Processing Systems (Cat. No.97TH8335).