Petri net-based techniques for modelling analysis,and performance evalution

Abstract

The primary objective of this Thesis is to develop Petri net-based techniques for the modelling, analysis, and performance evaluation of distributed processing systems. To illustrate these techniques, we focus on two important classes of distributed systems^ namely, flexible manufacturing systems (FMSs) and multiprocessors. In the area of flexible manufacturing systems, we propose techniques, all based on nets, for modelling, logical analysis, and quantitative performance evaluation. We first present an elegant scheme for hierarchical construction of classical Petri net or coloured Petri net models of FMSs. Using these hierarchical models and the framework of linear invariants of Petri nets and coloured Petri nets, we provide a theoretical basis for the logical analysis, including deadlocks, of these systems. Next, we propose generalized stochastic Petri nets (GSPNs) and Petri nets with deterministic and stochastic firing times (DSPNs) as effective performance modelling tools for FMSs. To illustrate the above techniques, we discuss real-life systems, such as, (i) the General Electric FMS, (ii) a three stage automated transfer line, (iii) a flexible manufacturing cell with multiple robots, and (iv) several other representative FMSs. In the area of multiprocessors, we consider the performance evaluation of a real-world fault-tolerant multiprocessor, FTMP, used in real-time air traffic control applications. Using GSPNs and DSPNs, we present several performance models that capture architectural and operational features of the FTMP, not modelled by the existing queueing network models. The investigations carried out in this Thesis have significant potential in the design and operation of other classes of distributed systems as well.