Quantification of the reliability required of non-destructive inspection of PWR pressure vessels

Abstract

This paper presents a method of quantifying the reliability required of non-destructive inspections of PWR pressure vessels. It gives a strategy for improving the effectiveness of ultrasonic non-destructive testing in assuring the integrity of a PWR vessel and allows targets of inspection reliability to be set in order to achieve the requisite level of vessel integrity. To do this the failure rate of PWR pressure vessels is predicted on the basis of a probabilistic fracture mechanics model. We use various models of the reliability of non-destructive inspection to discover the minimum level of reliability which is consistent with the desired integrity of the structure and to demonstrate how improvements can be made most effective. The reliability of inspection is usually modelled by a function giving the probability of leaving an unacceptable defect in the vessel. This function B(a) is really the “unreliability” of inspection and so 1 - B(a) gives the usual reliability. A reliable inspection is one which detects and correctly classifies defects according to some criterion usually based on size. A reliable inspection must use a technique which is intrinsically capable of detecting and sizing defects in the required size range and it must be reliably applied in practice. We find that, based on certain stated assumptions, that an inspection reliability of 80% of detecting and correctly sizing a defect of 15 mm through-wall extent yields a predicted failure rate of 10 −7 per vessel year. The failure rate includes a frequency of a major accident such as a large loss of coolant (LOCA) of frequency 10 −4 per vessel year. The predicted failure rate can be reduced to 10 −8 per vessel year if the sizing accuracy of the technique is improved so that the chance of undersizing a 15 mm defect falls from 0.19 to about 0.01. However, the failure rate of the vessel is not predicted to decrease further with any subsequent improvement in sizing accuracy unless there is also an improvement in the asymptote of the reliability of inspection. This asymptote is due to factors beyond the capability of the technique such as, for example, human error.