Modelling condition monitoring inspection using the delay-time concept
In the literature on inspection modelling, the failure distribution traditionally plays a fundamental role in model construction in that it is assumed that system failures occur instantly at random time points from new with a known pdf. of time to failure. Numerous models have been built on this basis. However, Professor Christer challenged this traditional idea and proposed the concept of delay time. The idea, which is an essential part of most engineers' experience, assumes that defects do not just appear as failures, but are present for a while before becoming sufficiently obvious to be noticed and declared as failures. The time lapse from when a defect could first be identified at an inspection to consequential failure has been termed the "delay time". It is this idea which can be captured to reveal the nature and scope for preventive maintenance or inspection. It appears that the concept is now being taken up by many other authors. In this thesis, various models for condition monitoring inspection are built on the basis of delay time analysis. Extensions and further developments are made here to enrich the delay-time modelling. Since the distribution of the delay time is important to delay time modelling, a new approach to estimate the delay time distribution is proposed. This technique, which contrasts with the previous subjective data estimation technique, is based upon objective data. Assuming the distribution of the delay time is known, models of condition monitoring inspection are fully discussed for both perfect and imperfect inspections, and for infinite and finite time horizons. Based upon the models for perfect inspection, algorithms are presented to find the optimal solution. Numerical examples are presented in each Chapter to illustrate how models and algorithms work.