The thesis considers several connected problems in the study of reliability of systems with unreliable components. Initially it reviews the dichotomic case introducing a new characterization which clarifies the importance of the k-out-of-n structures. This characterization is applied to two problems, failure to operate and to idle and stress-strength modelling. Whilst most previous work has concerned itself with coherent structures, being thought of as reasonable systems, the latter section of the chapter considers the case of non-coherent structures. The following chapter moves from the dichotomic models to multilevel models, considering possible . extensions and producing a hierarchy within the possible definitions. Chapter 3 considers the stochastic properties of such systems, again attension is paid to the k-out-of-n structures and non-coherent models. The final two chapters consider component performance. In chapter 4 the system is assumed to be hierarchical,, so that a component is part of a subsystem, and it is assumed the whole subsystem may be replaced. The effect on the component is examined. The final chapter considers optimal age replacement of a component considering 3 possible alternative criteria.

Connectionism is an approach currently being used in the field of cognitive science to investigate intelligence. Connectionist models are based upon the information processing properties of neurons in the brain and consist of very many, simple, interconnected processing elements which operate upon simple signals in parallel. The main objective of the work reported here is to show that connectionism may be applied to areas other than cognitive science. A simulation program has been implemented in which a connectionist network performs the real-time planning and control activities required to supervise the movement and processing of parts in a manufacturing system. The concurrency of connectionism is exploited in such a way that production of a part type may be characterised as “parallel”, where many machine tools are specified to perform each one of the operations required to transform raw material into finished product. The connectionist network is able to control all of these machines simultaneously and in real-time so that many parts can be at any stage of completion in a production facility. The precise routing of a part through the production facility is not specified in advance. Instead, the machine to which a part is scheduled next and the route by which it reaches this machine are decided when the part completes its current operation. These decisions are based upon the availabilities of machines at the time the decisions are made. The connectionist network is able to make these decisions, for every part in the production facility, in negligible time. The benefit of this approach is shown when the breakdown of machines is simulated. The network is able to react autonomously to breakdown by scheduling and routing parts around the affected machine. The necessary computations can be performed in real-time so that breakdown does not cause the manufacturing system to halt while production is re-planned.

No description available.

No description available.

The use of dynamical system models is commonplace in many areas of science and engineering. One is often interested in whether the attracting solutions in these models are robust to perturbations of the equations of motion. This question is extremely important in situations where it is undesirable to have a large response to perturbations for reasons of safety. An especially interesting case occurs when the perturbations are aperiodic and their exact form is unknown. Unfortunately, there is a lack of theory in the literature that deals with this situation. It would be extremely useful to have a practical technique that provides an upper bound on the size of the response for an arbitrary perturbation of given size. Estimates of this form would allow the simple determination of safety criteria that guarantee the response falls within some pre-specified safety limits. An excellent area of application for this technique would be engineering systems. Here one is frequently faced with the problem of obtaining safety criteria for systems that in operational use are subject to unknown, aperiodic perturbations. In this thesis I show that such safety criteria are easy to obtain by using the concept of persistence of hyperbolicity. This persistence result is well known in the theory of dynamical systems. The formulation I give is functional analytic in nature and this has the advantage that it is easy to generalise and is especially suited to the problem of unknown, aperiodic perturbations. The proof I give of the persistence theorem provides a technique for obtaining the safety estimates we want and the main part of this thesis is an investigation into how this can be practically done. The usefulness of the technique is illustrated through two example systems, both of which are forced oscillators. Firstly, I consider the case where the unforced oscillator has an asymptotically stable equilibrium. A good application of this is the problem of ship stability. The model is called the escape equation and has been argued to capture the relevant dynamics of a ship at sea. The problem is to find practical criteria that guarantee the ship does not capsize or go through large motions when there are external influences like wind and waves. I show how to provide good criteria which ensure a safe response when the external forcing is an arbitrary, bounded function of time. I also consider in some detail the phased-locked loop. This is a periodically forced oscillator which has an attracting periodic solution that is synchronised (or phase-locked) with the external forcing. It is interesting to consider the effect of small aperiodic variations in the external forcing. For hyperbolic solutions I show that the phase-locking persists and I give a method by which one can find an upperbound on the maximum size of the response.

No description available.

No description available.

This work reports the development of an approach to computer-aided control system design (CACSD). Of primary concern is the quality of co-operation between the designer and his computer. In order to achieve an effective co-operation, we propose a conceptual view of the CACSD problem to the designer, and develop a framework for the computer. The conceptual view is named Generalized Cooperative Search (GCS), which suggests the designer to treat a CACSD problem as a search problem, in which he and the computer co-operate to locate satisfactory designs among sets of candidates. The framework is Interactive Multi-Objective Programming (IMOP), which defines the level of abstraction as well as the organization of the design facilities in the computer. While IMOP methods and control system design methods provide readily usable tools organized by the framework, a design strategy is developed to guide the designer in their effective use. The strategy is a two-level plan of the design process which helps the designer to conduct his search problems as ones of IMOP. He constructs parametrized sets of candidate designs using the design methods (innovates in level I) and matches his design wishes with the design possibilities amongst these candidates using the IMOP methods (trade-offs in level II). The advantages of this approach are (1) a proper emphasis is put on an effective designer-computer co-operation, the designer is supported rather than forced to follow any planned course of actions; (2) a uniform trade-off among design objectives receive their long due attention and (3) it is pluralistic and different design methods may be combined for their respective strengths. The work conducted is likely to be the first integrated approach to designing control systems by search, and may even be the first which supports the principal design stages (formulation, generation and evaluation) in a co-ordinated as well as most general manner. The proposed approach has a promising prospect of being one for the general areas of computer-aided design in engineering and decision support systems in scientific management.

No description available.

No description available.