Abstract: In order to guarantee high control accuracy of ultra-fast cooling zone delivery temperature and coiling temperature， the run-out table cooling control system was optimized according to the major problems emerging from the tuning process of ultra-fast cooling. Aiming at the longitudinal temperature fluctuations after ultra-fast cooling， a new self-learning method for the heat transfer coefficients of ultra-fast cooling was proposed. The strip surface re-reddening phenomenon after ultra-fast cooling process was simulated with the finite difference method， based on which a compensation for strip surface re-reddening was added to run-out table cooling control system. In order to eliminate the influence of strip conveying velocity fluctuations on temperature， the strip conveying velocity diagram and temperature were recalculated for all strip segments which had entered the cooling sections， as soon as a new strip segment came to the measuring point of finish mills delivery temperature (FDT). With the methods mentioned above， the new run-out table cooling control system are optimized and field application showed that the strip temperature control accuracy are improved obviously.

Key words: hot strip mill, ultra-fast cooling, re-reddening, coiling temperature, self-learning