This paper presents a new methodology to transform UML State diagram to High Level Petri Net (HLPN). Genetic Algorithms (GAs) has been used to generate software specifications set which covers maximum states and transitions in state diagram. Then, these states are used to improve the logical ordering of HLPN. Not all states will be used to represent the HLPN, but those that are selected by GA which reduce the number of HLPN details and avoiding the redundant of states.