In the framework of automated manufacturing systems (AMS), Petri nets are widely used to model, analyze, and control them. Resolving deadlocks is of paramount significance because their emergence reduces the system throughput to zero. Supervisory control technique is the most widely adopted method to resolve them. A control policy can be converted into satisfying a set of inequalities, each of which corresponds to a siphon in a Petri net structure. The number of siphons can be exponential in the worst case, so does the number of inequalities. Taking into account the independent and dependent inequalities, this paper proposes a method to remove all the dependent inequalities while preserving only the independent ones. This method can significantly reduce the size of a supervisory controller. Examples are presented to illustrate the effectiveness and efficiency of this method.