Dynamic manufacturing systems consisting of multiple stages use a combination of dispatching rules to obtain production schedules in general. A weighted sum method, which assigns different weights to each dispatching rule and prioritizes jobs with a high weighted sum, is widely used especially for LCD and semiconductor manufacturing. A suitable set of weights by considering dynamic system states has to be determined in order to improve the throughput and utilization of systems. Hence, we develop a sequential search framework, with simulation and decision trees, which can generate a good weight set of dispatching rules within a short period of time. We show that the proposed search method performs better than a random search by performing experiments with real fab data.