SUMMARY OF THE SPECIAL SESSION ON LIFE PREDICTION (2009) - Chaired by Shunichi Suzuki, Peter Ford, Anders Jenssen, Peter Scott & Roger Staehle

Prepared by Peter Ford (with input from the other session chairs) & edited by John Hickling. Ulla Ehrnsten & Mike Wright reported on the final discussion.

Introduction

The introductory session was co-chaired by Shunichi Suzuki (who, as Chairman, chose the topic for the special session) and Peter Ford. Ford emphasized the technical management challenges associated with the overall need to predict both crack initiation and propagation times. These times are required to define appropriate inspection intervals.

In order to place the presentations in an historical context, Peter Ford then outlined the progress in meeting these technical management challenges since the 1970’s. This progress was governed initially by the urgent practical need to evolve long-lived mitigation actions, with a further, more recent need to develop quantitative life prediction methodologies required for life extension decisions. Although earlier emphasis was placed on an empirical, semi-quantitative understanding of the effects of, for instance, stress, environment and the metallurgical condition on the stress corrosion cracking susceptibility of a given alloy, there has been in the last 20 years an emphasis on developing an adequate understanding of the mechanisms of the various degradation processes. This can act as a basis, or skeleton, for making operating and regulatory decisions. Both of these approaches to life prediction, the empirical correlation with laboratory or plant data and the more fundamental mechanisms-informed approach, are represented in this life prediction special session

In order to define realistic goals, Ford suggested some “success criteria” against which these various mitigation and life prediction developments might be judged. For instance, the life prediction algorithm should:

Predict the relevant laboratory data and plant experience.
Take into account the plethora of stress, environment and material conditions, and their interactions.
Account for the uncertainty in prediction due to aleatory (random, stochastic events) and epistemic (lack of completeness in knowledge or control) factors.
Be backed up by an adequate quantitative understanding of the mechanism of cracking.
Be acceptable to the regulatory authorities.

In order to focus discussion, it was decided to concentrate on two “case studies”:

PWSCC of nickel-base Aaloys in PWRs
IGSCC and IASCC of austenitic stainless steels in BWRs

Before entering into these two specific topics , however, there were two “plenary” presentations , one by Peter Andresen of GE-GRC entitled “Prediction of SCC/IASCC in hot water by integrating fundamentals, critical measurements & plant characteristics”, and the other by Karen Gott of the Swedish Regulators (SSM) entitled “Life prediction ; regulatory aspects”. The reason for this choice of introductory talks was to stress the technical challenges associated with the various approaches to life prediction (e.g. empirical correlations vs. mechanisms-informed), as well as to emphasize the importance of acceptance by the regulators of whatever approach is finally adopted.

Peter Andresen stressed the complex interactions between the stress, material and environment inputs to life prediction algorithms that are applicable to most of the cracking systems in BWR and PWRs. He reviewed the approaches that have been made in the past including, for instance, analysis both of plant experience and of “well controlled” laboratory studies. To a large extent the uncertainty associated with the scatter in data that arises from such analyses is associated with a lack of definition (and control) of the relevant stress, material and environment conditions. Thus, in his view, it is critical that there be a fundamental understanding of the role of these factors, and that the resultant hypothesis governing life prediction be examined via critical testing under well-controlled conditions. If this is accomplished, then it becomes apparent that there are common features between the cracking sensitivity in apparently disparate systems (e.g. BWRs and PWRs). Thus, not only can there be “lessons-learned” from one system to another, but the engineering-driven assumptions of “threshold immunities”, environment-independent values of fracture resistance, etc. may be challenged.

Karen Gott reviewed from a regulatory perspective the conditions when a life prediction capability is needed. Namely when; (a) a damaged component is to be kept in service, (b) a plant wants to operate under conditions in excess of the original design basis (“Life extension” or “License extension”), (c) a plant increases thermal power (power uprate), and (d) a plant performs extensive modifications. She then elaborated on each of these four items. For instance in the first case, she emphasised that not only was it necessary to define the flaw dimensions and the propagation rate, but that these engineering requirements be backed up by an adequate understanding of the degradation mechanism. Moreover, if that understanding is not available, the regulator is obliged to assume a “worst case” scenario. Similar unnecessary conservative approaches may be necessary if there is uncertainty about the flaw dimensions.

Gott went on to define in some detail the regulatory requirements with regard to an adequate aging management program (AMP), not only for ongoing reactor operations but also for decision making for changes in design basis associated with, for instance, life extension and power uprates. Many of these AMPs include qualitative or quantitative risk-based inspection programmes (which have been required for more than a decade in Sweden). The effects of operational changes on potential degradation mechanisms need to be assessed with respect to AMPs including, for example, flow-accelerated corrosion and irradiation embrittlement, both of which can have synergistic interactions with cracking phenomena.

PWSCC of Ni-base alloys

The session on the Case Study “PWSCC of Ni-base alloys” was chaired by Anders Jenssen and Peter Scott. Three presentations were made:

Vessel Head management at Ringhals (Efsing)
Prediction of SCC Initiation Time of Alloy 600 PWR Components (Amzallag and Benhamou)
PWR Materials Aging Life Prediction – A US Perspective (Fyfitch)

Pål Efsing of Ringhals AB described the background and strategy behind the vessel head replacements for the Ringhals PWRs. After detection of defects in the vessel head penetrations of Ringhals 2 in 1992, the head was eventually replaced in 1996. The new head was manufactured with penetrations in Alloy 690 and Alloy 152/52 weld metals. Regarding Ringhals 3 and 4, shallow surface breaking defects on the inner surface of the penetrations were detected during inspection of the VHPs in 1999 and 2001. Prior to these findings, new regulatory requirements had been put in place in Sweden, which were more rigid regarding demands on the inspection programs. This, together with small NDE detection targets, complicated qualification of NDE, and – together with high assumed bounding CGRs for the materials and the environment concerned – resulted in the imposition of short inspection intervals. With respect to reactor safety as well as plant economics, it was concluded that replacement was the best solution. As a result it was decided in 2002 to replace the vessel heads in units 3 and 4.

Claude Amzallag of EDF presented a life prediction model for the initiation time of SCC in Alloy 600 components. The model is based on Monte Carlo simulations that take into account the dispersion in the input SCC parameters; operating temperature, activation energy, stress, and a material susceptibility index. Inputs for the various parameters were thermohydraulic analyses for the operating temperature, laboratory data and literature data for the activation energy, FE analysis for stresses (with direct measurements on mock-ups to support and validate the FE analyses), and laboratory data, as well as field experience for the material susceptibility indices. The model was applied to predict cracking in VHPs, where random sampling of the input parameters was repeated 100 times for each nozzle, taking into account their angle with the head (which affects residual stress), operating temperature, SCC material susceptibility index, and activation energy. The results showed that the model was able to predict the evolution of cracking in VHPs, and that the predictions at 5% probability of occurrence yielded the best correlation with field data. The model was also applied to bottom mounted instrument (BMI) nozzles for which there have so far been no incidents of cracking in France. For peripheral BMI nozzles, the model predicts no cracking during 40 years of operation, while a few cases of cracking were predicted for 60 years of operation. Regarding straightened peripheral nozzles, the model predicts a higher probability for cracking towards the end of 40 years of operation.

Steve Fyfitch of Areva NP, Inc. presented a U.S. perspective on PWR materials aging life prediction. He summarized field experience, laboratory data and available material and fabrication data before describing various life prediction approaches employed in the U.S. Over the years, several practical (mainly empirical) tools have been developed based on operating experience to assess the risk of PWSCC in Alloy 600 components. These approaches include ranking models, Weibull-based approaches, risk-based decision planning, and Monte Carlo simulations. The EPRI MRP risk assessment methodology predicts the probability of leakage using field experience and a Weibull model, as well as using a probabilistic fracture mechanics (PFM) model for assessing the probability of nozzle ejection. Plants at various head temperatures were analyzed and the components were categorized in risk categories, based on probability of net section collapse (NSC) and cumulative leakage probability. Examples of cumulative probabilities of NSC and leakage were presented, where the probabilities were given for various temperatures. Vessel heads with less than 10 effective damage years (EDY, based on time and temperature) were placed in the low-risk category, while heads with more than 18 EDYs were categorized as high-risk. When entering the high-risk category, bare metal visual inspections can reduce the probability of NSC to acceptable levels indefinitely, while NDE every 4 EDYs significantly reduces the probability of NSC. The presentation ended with a summary of the utility actions and the current U.S: aging management approach. Most vessel heads with a high risk for PWSCC have been replaced, and since 2004 no new indications have been found in BMI nozzles.

IGSCC and IASCC of austenitic stainless steels in BWRs

The session on the case study “IGSCC and IASCC of austenitic stainless steels in BWRs” was co chaired by Shunichi Suzuki of TEPCO and Peter Ford. Three presentations were made;

The current experience of SCC/IASCC in Japanese BWRs (Suzuki)
Mitigation Effectiveness of HWC &NMCA in the Field (Pathania)
Approaches to the Prediction of IGSCC (Shoji)

In the first of these presentations, Shunichi Suzuki of TEPCO addressed the discrepancy that sometimes exists between actual plant behavior and experimental observations. This obviously impacts the relevance of any life prediction methodology based solely on laboratory data. Such discrepancies arise from various factors, such as:

Definition of SCC initiation & propagation
Evaluation of SCC initiation time
Evaluation of SCC propagation
Crack propagation behavior into a dissimilar metal
Intermittent crack propagation in the components
A lack of understanding of the real conditions in actual power plants

He elaborated on some of these factors including, for example, the change of the oxide film formation with plant ageing and the effect that this has on crack initiation. Further examples included the evaluation of crack arrest at the interface to a dissimilar metal and its effect on overall crack propagation as well as the evaluation of in-core crevice conditions and their effect on definition of the real “metal/environment system”.

Suzuki gave several examples of the conservatism of the JSME S NA2002 disposition curves when applied to 300A PLR pipe and nickel–based alloys, where the crack propagation rates have been retarded or even arrested at dissimilar metal weldments. With regard to disposition relationships in low alloy steels, where a crack may have initiated and first propagated through an adjacent austenitic alloy to the low alloy steel interface, it was suggested that the current disposition rates quoted in, for instance the EPRI BWRVIP-60A report, may be over-conservative because of crack-arrest arguments.

The particular question of crack propagation into a low alloy steel pressure vessel after active SCC in an austenitic weld attachment (or cladding) material was discussed by Suzuki in more detail via reference to studies conducted between TEPCO and GE Global Research Center. This has been a topic of some concern for BWRs on NWC given the possibility that significant propagation into the low alloy steel may occur under specific conditions of loading (e.g., ripple loading or high stress intensity with limited plastic constraint), oxidizing conditions and anionic concentration (especially of chloride). Suzuki and his co-workers concluded that:

No crack propagation occurred under constant K conditions up to 50 MPa√m in water containing 0 to 30 ppb sulfate.
Applying periodical unloading/loading cycles, or adding 5 ppb or more chloride in tests at constant K, promoted crack propagation into the LAS.
A small amount of crack growth into LAS was observed at a constant K of 65 MPa√m in pure water but the crack appeared to arrest based on DCPD measurements. It should be noted that testing a 1T C(T) specimen at 65 MPa√m is invalid according to ASTM E399.
Crack growth rate testing of homogeneous A533B base metal implied that the resistance of the LAS to SCC propagation becomes higher when a larger specimen with a higher degree of mechanical constraint is used.

Raj Pathania, Jonathan Kubiak and Bob Carter of EPRI reported on a project whose objective was to evaluate the effectiveness of IGSCC mitigation in BWR internals by Hydrogen Water Chemistry (HWC) and Noble Metal Chemical Addition (NMCA) and, thereby, to provide the technical basis for inspection relief. Life prediction algorithms, or more accurately, disposition relationships for crack propagation exist for these operating conditions, but there are remarkably few data from operating plant to evaluate their accuracy. The approach used was to evaluate re-inspection data for core shrouds and stub tubes which had operated for some time under either Normal Water Chemistry (NWC), HWC or NMCA. The crack depth and length data from 10 plants were evaluated and translated into average crack propagation rates (CGR) via the crack dimension increment during the inspection interval, which varied from 1 to 3 years for NWC and 4-10 years for mitigated plants. These CGRs were analysed via cumulative frequency distributions and a comparison made between the average rates and the bounding values for NWC and HWC used in BWRVIP-14-A

For the case of the core shrouds, the lower CGRs for the HWC/NMCA mitigated plants compared to the NWC plants strongly support the effectiveness of mitigation technologies deployed by BWRs. Moreover the data reflect a decrease in CGR with deepening cracks for HWC/NMCA when compared to NWC. Furthermore, the current disposition relationships for both crack length and depth are conservative in 86% and 99% of the examined cases, respectively. This conservatism may be related to the complex residual stress profiles in, for instance core shrouds which will also diminish over time due to irradiation–induced relaxation.

With regard to stub tube cracking, two plants were analyzed where inspection records were examined before and after the application of a hydrogen-related mitigation action; NMCA in one case and HWC in the other. One plant had 129 CRD penetrations that had material conditions that are highly susceptible to IGSCC (furnace sensitized 304 SS). Prior to HWC and Noble Metals implementation (2000), approximately 2 to 3 new leaking CRD penetrations were identified during the ASME leakage test performed at the end of each refueling outage. Since HWC and Noble Metals implementation in 2000, no new leaking penetrations were identified in three successive refueling outages (2003, 2005 and 2007), or in forced outages. One new location was, however, identified in the 2001 refueling outage approximately 6 months after NMCA applied; however this was likely a pre-existing location. Preliminary analysis of the stub tube cracking in the second plant indicates a significant impact of hydrogen injection which was applied in 1996; the crack growth rate was reduced by a factor of ten.

Tetsuo Shoji, Yoichi Takeda and Kazuhiko Sakaguchi concluded the Case Study section of the Special Session with a presentation entitled “Approaches to the prediction of IGSCC- surface integrity and structural integrity”. This presentation served as a useful bridge between the short-term ( i.e. ≤5 year) developments so far discussed and longer term developments (i.e.>10 year) that would require significant advances in mechanistic understanding of many of the phenomena relevant to SCC initiation and propagation.

One motivation for such long-term developments is the continued observation of SCC of 316L(N) stainless steel piping and core shrouds in BWRs. This unexpected cracking in supposedly resistant alloys has been associated with, (a) severely machined surfaces with highly strained/hardened surface and roughness, (b) local coarse slip bands with high, internal, elastic dislocation stress and, ( c ) highly localized residual stress as well as macroscopic residual stress from welding. The microscopic factors described above are not part of current life prediction methodologies and there is, therefore, a need to define the role of surface integrity in structural integrity and for better predictive capability of IGSCC. This applies not only to cracking in oxidizing environments (where the slip oxidation mechanism of cracking probably dominates), but also to cracking in reducing environments (where the internal oxidation or hydrogen induced cracking mechanisms are reasonable hypotheses).

Shoji went on to elaborate on these ideas with reference to the localized oxidation that occurs in both oxidizing and reducing environments, and which may govern crack initiation. As illustration he referred to laboratory studies (using Raman spectroscopy) and to observations from plant components such as the PLR piping and core shroud at Onagawa unit 1 BWR and the steam generator inlet nozzle from Mihama unit 2 PWR. He concluded that although uniform oxidation is predictable, localized oxidation is not. Factors associated with this uncertainty concerning the trigger that promotes localized oxidation and SCC initiation include the autocatalytic effect of a localized stress at a growing oxide/metal interface, which can be an important factor for continued localized oxidation. Furthermore, hydrogen and oxygen generated by water molecule dissociation on the metal or oxide surface may play a critical role in localized oxidation, as well as in accelerated oxidation. Other factors that may play an important role in localized oxidation include the local stress field generated by dislocation morphology .

Based on these observations, Shoji recommended that there should be more study of the oxidation kinetics of structural materials in LWR environments with special emphasis on the role of stress (both the macroscopic stress caused by things such as welding, and the microscopic stress caused by local machining, dislocation structure, microstructural inhomogenity and oxidation), role of hydrogen, role of radiolysis products, neutron irradiation, as well as their interaction. A further recommendation was that an analysis of the oxide/metal interface profile in terms of the Abbot bearing curve may provide useful information about the process of localized oxidation and dynamic analysis (time dependence of the curve). Such developments would be essential to follow the interface dynamics and resultant SCC initiation and propagation.

General discussion and future prospects

The final discussion session was co-chaired by Ford & Staehle, beginning with an overview of longer-term research needs by Roger Staehle. These opinions were given in a presentation entitled “Predicting the Initiation and Propagation of Stress Corrosion Cracking by Identifying and Quantifying Unit Micro-Nano Processes”. Roger made the point that no fully mechanistic predictive models exist presently. Current phenomenological models are not able to predict future events quantitatively in an a priori sense and his belief was that a model stemming from atomic-level, basic processes is required. The aim of such a model should be to predict cracking from the initial condition of a material surface through precursor stages, initiation and on to the steady-state propagation stage. The precursor and initiation stages were emphasised particularly, based on the aim of predicting low frequency cracking incidents, occurring only after many years into the extended design life of new plants.

Roger noted the paradigm shift since the very “tight” nature of crack tips has been recognized leading to atomic level visualization of the processes involved. The importance of microscopic interactions between dislocations and surface films, vacancy diffusion and hydrogen, etc., were noted. Recent developments in crack tip examinations, using ATEM and other high-resolution material/chemical characterization tools, are expected to provide the tools necessary to make progress. However, it was noted that expertise from beyond the normal pool of EAC research would be needed. The need for a probabilistic approach was also emphasised, particularly in terms of material variability. In Roger’s opinion, the new micro-nano models that could evolve over the next 5-10 years will not be “hydrogen” models or “anodic SCC” models. They will be more detailed and have the capacity to rationalize molecular conditions for SCC as well as the important dependencies.

Peter Ford opened the general discussion by challenging the assertion that current phenomenological models are not able quantitatively to predict future events, noting the successful application of models as presented in the PWR/BWR cases studies in the preceding parts of the special session. He also noted that there is not enough time to produce a mechanistic model, such as Roger suggests, and meet the needs of operating plants, but suggested that the focus should be put on improving existing predictive models. Peter Andresen noted that BWR core shroud SCC was predicted some five years before it was actually observed in plants. Also, current models are very effective in predicting the benefits of mitigation measures. Research should focus on investigating factors which are not perfectly covered by the existing models. However, it was acknowledged by Peter Scott that predicting SCC in a new or different system, as yet unaffected by SCC, is much more problematic. Scott added that, from a utility/vendor perspective, we are a long way from Roger’s vision of a fundamental predictive model. A major barrier is the inability fully to characterize stresses, and even environments, in a real plant system.

Staehle noted again that the ultimate aim of models should be to predict “first of a kind” failures, precisely what is so challenging for current models. He is convinced that new types of failures will occur in the future. Steve Bruemmer suggested that the efficacy of current models could be tested by selecting an emerging issue, like SCC of cold-worked stainless steels in PWRs, predicting future plant behavior, and then comparing those predictions with reality over the coming years.

Ron Horn suggested that attention should be focused on new plants/new components. He suggested that “initial condition” information could be gathered and digital databases be built before operation starts, while the opportunity exists. Roger’s modeling ideas might be used to help identify what data should be gathered. Bruemmer noted that an integrated and focused approach is required, because of the need to involve many disciplines and input from diverse groups. Open communication and discussion is seen to be vital. Steve also noted that expensive data gathering should lead to a useful “product” as soon as possible. The funding realities of today result in a situation where the window is open for only a relatively short time for research on a specific issue.

Gary Was was asked how he might approach the problem from a research perspective. Was noted the ambitious nature of Roger’s vision and commented that there is a significant difference between predicting the first of a kind failure compared to predicting the second or hundredth. That being said, Gary stressed the importance of fundamental understanding to models, which is somewhat lacking in the predictive models of today.

Karen Gott was asked to explain why, as a regulator, mechanistic understanding was considered important. Gott responded that understanding is required to ensure inspections were correctly targeted and to ensure genuine mitigation is achieved.

Claude Amzallag was asked for a Utility perspective. Amzallag emphasized the need for practical and effective models enabling safe and economic operation of plants. The commercial success/effectiveness of the French PWR approach was noted again. However, he did not consider it possible to predict the first occasion of a new failure mode.

Peter Scott noted that one goal for more fundamental science might be to underpin confidence in Alloy 690, or any other alternative, replacement material with high expectations of EAC resistance.

Raj Pathania mentioned the US proactive approach. Attitudes changed after the Davis Besse incident and new work focused on important unknowns has been initiated. The example of investigation of environmental effects on (high temperature) fracture resistance was noted as one result from the proactive degradation matrix work. He also mentioned that the matrix is available at www.epri.com.

Wade Karlsen noted the European “Perform 60” program that has aims in line with Roger Staehle’s vision and encouraged people to joint the end-user-group to follow this project. Roger noted again the need for ongoing discussion and communication between variousgroups, and especially the need to involve key, outside expertise. He cited the examples of the recent EPRI “Cold-Work” and “Crack Initiation” workshops.

Peter Chou (EPRI) proposed that a possible mechanistic model should also include degradation in Generation IV plants, as these reactors are due to be in use by 2035.