Theory and design of a hardware stochastic reliability simulator.

Abstract

The work described in this thesis is concerned with the design of special purpose digital circuitry to achieve high-speed simulation of system reliability. Initially, an investigation into alternative simulation methods is undertaken. Analytical and Monte-Carlo techniques are described, and the advantages and disadvantages of each method are discussed. The operating principle of the newly-developed simulator is placed in context with current simulator design. Systems that contain reliable components, arranged in redundant configurations, possess high reliability. A reliability assessment for such systems requires fast simulation. Considerable parallel hardware operation is employed to produce simulation speeds faster than lOOyears/second, independent of the number of system components. A novel technique is also used to gain further speed via asynchronous time scaling. Modelling of complex strategies of repair and preventative maintenance are catered for. Systems can be modelled where facilities for repair are limited and system components compete, according to some priority policy, for available repair men. The use of these additional modelling features is achieved without reducing simulation speed. Components that make up the system under investigation are modelled by general purpose hardware modules. Each component module is equipped for considerable modelling detail, providing a simulator that can model systems with wide-ranging characteristics. The operation of component modules is determined by a micro-programmable control unit. The interconnection of components to form a system is dealt with by logical network techniques. Although logical networks are employed, observation of system behaviour is not solely by monitoring network events. Important to the operation of the simulator is the generation of random event signals of prescribed distribution. Considerable care has been taken in the design of the event signal generators, and their operation is both mathematically and statistically investigated to verify their performance. Finally the simulator is applied to a range of reliability simulation problems. The systems initially considered are analysed to confirm the simulator's operation. Later, systems containing complex repair schemes are modelled, and the economic or reliability performance of the system is optimised.