Abstract: The process control of coiling temperature can be described mainly by mathematics, but laminar cooling is such a complex non-linear process that it can't be accurately described by mathematics especially at low temperature. According to the measured data of laminar cooling process during plate hot rolling in PanSteel, a decision table is given with sampled, in which the rough set theory is introduced to fuzz up linguistically the sample data so as to mine the implicit association rule by reducing attributes and rejecting redundant rules in accordance to the actual support/confidence level of association rule for linguistic data. Then, the conventional mathematic model for laminar cooling can be optimized by developing a fuzzily dynamic model. The actual operation with measured data shows that this method can improve the controlling precision of coiling temperature by 1%-2% and it has a great application potential.