کاربرد مدل سازی احتمالی برای مدیریت طول عمر دیگهای بخار هسته ای در رژیم خزش : قسمت 1

Monte Carlo probabilistic simulation has been applied to a large population of nominally identical components in an AGR boiler operating in the creep regime. The components have a history of defectiveness. The R5 procedure is used to calculate creep-fatigue crack growth rates within a probabilistic programme. The inspection process is also modelled probabilistically. The overall result is a ‘prediction’ of past inspection results which can be used to tune the parameters of the model. The model then makes genuine predictions of the required level of remediations in future overhauls by predicting the inspection results. The probabilistic treatment of both the structural calculations and the inspection process jointly has been shown to assist in clarifying the interpretation of the inspections and ascertaining the true state of the plant.

Some of the UK's Advanced Gas Cooled (AGR) reactors are already operating beyond their original design lifetime, and all the AGRs may be expected to do so in due course. At full power the reactor coolant gas temperature is around 650 °C as it enters the boilers. Consequently, creep is a potentially life limiting mechanism for some boiler components. The accurate prediction of creep lives is hampered by the large scatter in creep material properties. For the purposes of underwriting nuclear safety, bounding material properties are assumed in conservative assessment methodologies. The degree of conservatism in such safety related assessments can be such that, whilst entirely appropriate for ensuring safety, they give no realistic picture of plant lifetimes. In common with boilers in conventional power plant, AGR boiler surfaces consist of large numbers of very similar tubes and associated features. This lends itself naturally to a probabilistic treatment. The failure of small numbers of tubes is tolerable from the nuclear safety perspective. Indeed the occasional occurrence of steam leaks is anticipated and managed by repairs, replacements or by plugging tubes. This need have no commercial or safety implications so long as the rate of leaks and the number of tubes requiring remediation remains small. One of the purposes of a probabilistic treatment is to predict future leak and remediation rates, and hence the likely commercial impact and the degree of challenge to the safety envelope. Two cases may be distinguished depending upon the availability of inspection evidence, [1] Extensive in-service inspection evidence exists and there is a history of cracking as well as some steam leaks; [2] Little or no in-service inspection evidence exists and hence the potential defectiveness is unknown other than indirectly from a few steam leaks. Case [1] is considered in this paper. The focus in this case is on the probabilistic modelling of creep-fatigue crack growth and the probabilistic modelling of the inspection process. Modelling the combination of these two factors statistically is unusual, though crucial to the outcome in this case. It is intended that subsequent work will treat case [2] for which the focus will be the probabilistic modelling of either creep rupture or creep-fatigue crack initiation (possibly followed by crack growth). There is also, of course, the case where there is in-service inspection evidence but with no history of cracking. However this more benign case is unlikely to merit detailed probabilistic modelling and hence is not considered further.

Structural assessments in the creep regime are subject to large uncertainties. Difficult in-reactor inspections have limited reproducibility. A probabilistic treatment of both these issues jointly can assist in ascertaining the true state of the plant.