The dynamic slab allocation problem is to assign the surplus slabs, which are dynamically obtained from steel- making and casting process, to the customer-orders, potential-orders and self-designed-orders in a given period. Therefore, a 0-1 integer programming model is formulated. Due to its NP-hardness, a multiple neighborhoods based scatter search algorithm is proposed to solve the problem approximately. Furthermore, to avoid search process being tapped in local optima, a random perturbation strategy is introduced. Both randomly generated instances and practical instances obtained from a large iron-and-steel enterprise are used to test the performance of the proposed algorithm. Compared with commercial soft named CPLEX, the proposed algorithm can get near-optimal solution in a reasonable CPU time. The experiments on the practical data show that the proposed algorithm outperforms the manual methods on solution quality and computation time.