An assumption-based inference system for solving scheduling problems in the steel-making process is proposed. Scheduling problems that are difficult to solve by an analytical method have been mainly solved by searching state space. However, the cost of solving the state-space problem is expensive. An assumption-based inference system that make it possible to search state space effectively and avoid combinational explosion by the pair coils exchange method was developed. To evaluate the approach, the system was applied to a steel production line for prototype use. It was confirmed that this system improved product quality.