Semicrystalline polymer materials play an essential role in the industry. The modeling and prediction of polymer microstructures has obvious environmental and economic interests. In this study, a mesoscale modeling of spherulite pattern formation is constructed. For this purpose, the spherulite growing is modeled by a multi-phase fields model and by a cellular automaton. The heterogeneous nucleation of the spherulites is modeled using the concept of random local nucleation time which allows to optimize the computational time.