Study on Optimization of Learning Structure in ROT Process of Hot Strip Mills

We propose an algorithm that optimizes the learning structure of the heat flux coefficient through statistical similarity analysis to improve the temperature control problems in the run-out table (ROT) process of hot strip mills. The p-values are calculated for each segment of the learning structure to determine the similarity of the average values. Based on the proposed algorithms, we perform optimization tests with various steel types, thicknesses, and temperature grades produced in the ROT process. From the test results, the existing learning divisions are reduced by 62.2%. Using hot strip mill data, we show that the average of the bank fitness is significantly improved by the proposed learning structure.