<p>This paper deals with the complex problem of designing an optimal supervisor in safe Petri nets (PNs). Reducing the structural complexity of the supervisory model through a method with less computational burden is discussed. First, a class of linear constraints is introduced as semi-identical constraints, and we prove that disjunctive generalized mutual exclusion constraints, known as OR-GMEC, can be equivalently transformed into a set of them. The main idea is founded on disjunctive properties and the over-state concept. Second, we show that it is easy to obtain an equivalent whose implementation has less structural complexity. Third, a simplification method is proposed to calculate a simpler supervisor using several merging rules. Finally, computationally efficient techniques based on the constraint forbidden over-state concept are provided for applying the way to enforcing OR-GMEC. The obtained controller is maximally permissive, and its model is a non-ordinary PN. The advantages of the method in terms of computational and structural complexity are shown through some attractive examples.</p>