ABSTRACT
This paper is concerned with the idea of modelling the growth of reliability during a Test, Analyse and Fix regime, with a view to bringing out some of the advantages, disadvantages, uses and abuses of such models. The notion of a mathematical model is clarified initially, with some comments on different types of models. The author goes on to discuss Duane modelling and Fix Effectiveness Factors, applied to data obtained from projects at the development stage. The importance of allowing engineering judgement to be taken into account is considered, and the means of doing this, and the advantages of, using Bayesian statistics is discussed. The uses of such techniques are considered, such as management aids to decision making, milestone targets, assessing the health of a project and obtaining an idea of the further development effort needed to meet a requirement or the current system reliability.lt is emphasised that such figures for reliability and other parameters are only an assessment and not a promise, unless the underlying assumptions are met. This would appear to raise questions about the utility of modelling in this situation, but quantification of the reliability is not the only useful activity, and the other uses mentioned above can still make these techniques beneficial. The point is also made that making predictions in any discipline is a difficult task that is fraught with danger.
