Modeling Of Software System’s Requirements With Continuous And Discrete Variables

With the rapid development of computer information technology, people propose rigorous andrelative complete software requirements which increase the difficulty of the software modeling.In the past, people paid much attention to the behavior of the software system and realized the fu-nction with formal modeling methods. The current software systems are regarded as a hybrid sy-stem with discrete and continuous variables while modeling hybrid system has made some prog-ress and it’s not extensively used in the field of software development, modeling and design.Combining the powerful ability and advantage of the description of discrete system by Petri ne-ts and continuous system by artificial neural networks (ANN), this paper proposed an enhancedPetri nets model to model the hybrid system in order to provide a novel way of software require-ments modeling and the main contribution of this paper is introduced as the following:1) Extracting valid input and output data flow information from the logfile of the software syst-emWe obtain the software system’s input and output data flow information in the form of discreteand continuous variables from the system’s logfile by the keyword search, and then establish thePetri nets model based on discrete variables and ANN model based on continuous variables withthe purpose of satisfying the system’s requirements.2) Transforming the ANN model into the EPN modelPetri nets model and ANN model have no direct relationship between each other owing to theyare two different models and may have some difference in modeling the same system. By the def-inition of a new Petri nets naming enhanced Petri nets, it can model the nonlinear system. The E-PN model which the ANN model is transformed by the improving transformational rules can ha-ve the same ability of approximate arbitrary nonlinear function as ANN model and achieve themodeling of the nonlinear systems and hybrid systems.3) Synthesizing of Petri nets modelsThe existing synthesizing methods of Petri nets are shared composition and synchronous comp-osition which can’t fully reserve the properties of subsystems. However, we can define new tran-sitions with new functions by an EPN model in order to synthesize different Petri nets models a-nd realize system’s uniform modeling.An example of water tank controlling system is redesigned and analysed in this paper. The sy-stem is first modeled by the method propsed in this paper and then we analyze the output of the EPN model and the original system model with the same input. By comparing the EPN’s state re-achable graph and original system states, the result shows that our method has the same perform-ance as the original model and is more efficient to model hybrid system.