Abstract:
This thesis consists of a study of Manitoba Telecom Services' Trouble Diagnosis and Repair System (TDRS). The TDRS consists of three levels of service staff responsible for diagnosing and repairing telephone switching equipment faults (troubles). It is modelled as a network flow model with feedback in which service has several probabilistic outcomes and routing is determined by service outcome. There are several sources of heterogeneity including different criticality levels of troubles having different probabilistic distributions of service outcome and the scheduling of different staffing levels over the shifts of the week. A stochastic logical model of the TDRS is built and then implemented as a stochastic computer simulation program. It is demonstrated that the program produces stable and reasonable measures of the performance of the TDRS. A series of experiments is performed to observe the effect of varying staffing levels on the performance measures of the TDRS, which typifies the sort of questions forwhich the program may be used to gain insight. It is also demonstrated that simulation is a useful tool for studying complex systems that may be intractable to analytic methods. In particular the simulation provides insight into the behaviour of the system over daily and weekly periods by providing a graphical interface with dynamic trace outputs. A utilization analysis of the TDRS is presented that overcomes its non-Markovian nature, in particular its lack of the memoryless property, and its lack of steady state. The utilization analysis has proved to be a useful tool both in verifying the correctness of the computer simulation and in planning experimental work.
