A new mathematical model constituting temperature schedule was established in order to realize flexible rolling of heavy and medium plate, in which the architecture of artificial neuron network (ANN) with optimal ability to predict mechanical properties was first constituted, and then, temperature schedule was constituted by genetic algorithm (GA). It is shown that mechanical properties can be well-predicted by recursive functions and ANN, and the precision predicted by ANN is higher than that by recursive functions. Finishing temperature and final cooling temperature among schedule ones are most important factors for mechanical properties. Mechanical properties calculated by temperature schedule and recursive mechanical-property functions agreed well with desired results. The temperature schedule constituted can be used to produce some steel plates with the same compositions but different strengths.