Application of Case-based Reasoning and Iterative Learning to Laminar Cooling Process Control

In some previous study, the strip coiling temperature prediction compensator and batch to batch compensator cannot obtain good compensating results due to the manually adjusted weight parameters for index feature of the case-based reasoning (CBR) system. And exact match and effective iteration cannot be done for the lack of initial operating condition matching algorithm. For this reason, a method based on neural network technology is proposed to learn the weights parameters of the index features of CBR system, with an initial operating condition matching algorithm that uses iterative learning technique to improve prediction compensator and the batch to batch compensator. The proposed hybrid intelligent control method is applied to a large domestic steel plant, and the results show that the strip coiling temperature control error decrease 1.63 C and the hit rate increased 14.5 % where the coiling temperature errors are controlled in the range of ± 10 C.