Conference Agenda

This presentation summarizes the investigations into the root causes of cracking in NEK’s safety injection line SI-53 that led to RCS leakage and forced shutdown in October 2023.

Non-destructive analysis identified the leakage crack in the region of weld BW-9, between the 4-inch pipe and 6”x4” reducer, shortly after the RPV nozzle.

Destructive testing and analyses determined corrosion fatigue to be the initial cracking mechanism, followed by low cycle fatigue interrupted by zones of ductile tearing.Measurements of the oxide thickness on the crack surface suggest that the leakage crack developed over a period of more than 10 years.

Observations during pipe replacement activities determined that a significant unexpected cold spring was present in the line, that may to some extent have been caused or exacerbated by misalignment or welding abnormalities during the original installation.

Computational fluid dynamics (CFD) analyses determined that turbulent penetration and stratification could be present in the line, leading to significant unanalyzed stresses in the line. Temperature monitoring confirmed the presence of thermal penetration and stratification during normal operating conditions, with the addition of lower amplitude temperature cycling that correlated with displacement and strain gauge measurements. Further CFD analysis demonstrated that during a low-flow safety injection transient thermal stratification could develop to an extent that could cause plastic deformation in the SI-53 pipe section, which may also have contributed or caused the observed cold spring. The measured amount of cold spring was determined to lead to a very high level of tensile stress that is believed to have been a significant contributor to the onset of corrosion fatigue. During the pipe replacement particular attention has been paid to ensure careful alignment and well controlled welding in order to avoid introducing cold spring.

Operators at Nuklearna Elektrarna Krško (NEK) identified a leak in the Reactor Coolant System (RCS) on October 4th, 2023, and initiated the appropriate abnormal condition procedures. Two days later, on October 6th, the decision was made to shut down the plant in order to determine the source of the leakage. The investigation, completed by October 8th, revealed the leakage on the SI-53 Safety Injection line to the reactor vessel. As a result, both SI-53 and its sister line, SI-52, along with the associated reducers and elbows, were removed and replaced.

The failure analysis process involved close collaboration between Westinghouse Electric Company (WEC), NEK, and Institute of Metals and Technology (IMT). A section of the SI-53 piping was subjected to detailed destructive examination at Westinghouse Churchill Site (WEC) to investigate the Direct Cause Analysis (DCA) for the crack initiation and its propagation. The crack was identified as circumferential crack, which initiated in the heat-affected zone (HAZ) on the Inner Diameter (ID), very close to the fusion line with further transgranular propagation through the weldment. Very limited or no branching of crack was observed. Two different crack propagation mechanisms were identified on the fractured surface. Chemical analysis, tensile tests and hardness measurements were also performed on the investigated sections of the pipe, all of which did not reveal any irregularities.

The findings of the investigation into initiation and propagation of this crack were the basis for the Root Couse Analysis (RCA).

The computational fluid dynamics (CFD) has been used to elucidate the thermal-hydraulic conditions in the direct in-vessel safety injection (SI) lines of the NPP Krško (NEK), which have been replaced after the leak of the SI-53 line in October 2023. The primary goal of conducted CFD studies in the frame of the root cause analysis of the event in NPP Krško has been to provide insights into the thermal-hydraulic behaviour of horizontal dead-end pipe configuration, with focus on potential flow unsteadiness that can lead to temperature fluctuations, potentially causing thermal fatigue and through-wall cracks in stainless-steel pipes. Indeed, fatigue striations have been observed and acknowledged by the direct cause analysis as the crack propagation mechanism in the second part of material as the crack propagated. These striations were (after other reasonable hypotheses excluded) finally attributed to thermal fatigue, as also confirmed by in-situ measurements of temperature, displacement and strains on the replaced pipe.

We have simulated part of the reactor vessel downcomer and around 4 m long horizontal dead end section of the SI-53 pipe. Conducted study, presented here, predicted turbulent swirl in around 0.5 m long section of the SI pipe near the reactor vessel and laminar-to-turbulent natural convection loop in the remaining 3.5 m long horizontal section. Combination of the swirl flow and natural circulation loop is highly sensitive to settings configuration of the CFD setup (e.g. applied boundary conditions, mesh resolution, turbulence modelling and numerical methods). Focusing on unsteady features of the flow, if any, different transient models were tested. Inherent flow instability with spontaneous movements of the frontline between the penetrating swirl flow and the natural convection loop, which accounted in significant unaccounted-for stresses, has been captured only with the use of specific turbulence models belonging to the group of scale-resolving simulations, while the conventional unsteady RANS approach predicted rather steady flow behaviour with only small-amplitude turbulent fluctuations at inner pipe wall. Uncertainties of CFD simulations, associated with applied cooling conditions, geometrical features of direct vessel injection (DVI) nozzle and choice of turbulence model in addition to other numerical aspects, have been quantified with the steady-state RANS simulations, while the longer transients were simulated only with the limited set of parameters.

The second part of this paper summarizes the results of simplified transient CFD simulations that were performed to support the hypothesis of low inflow (Froude number < 1) existence at check valve during the past inadvertent SI actuations. It has been confirmed by CFD that slow flooding of the horizontal section of SI-53 pipe with cold water could strengthen already existing flow/temperature stratification in the SI-53 pipe, and eventually contribute to the pipe bending due to increased vertical temperature gradients.

This paper summarizes the analysis of temperatures measured on the external walls of the direct in-vessel safety injection pipes SI-53 and SI-52 at the Krško NPP, along with the evaluation of measured displacements and strains on the SI-53 pipe. Following the leak in the original SI-53 pipe and the subsequent replacement of both SI-52 and SI-53 in 2023, measurements were conducted in 2024. These were combined with CFD analyses to assess the presence of thermal fatigue in the SI-53 line.

Due to the difficulty of achieving tight thermal contact between the RTDs and the pipe surfaces, absolute temperature readings—especially near the reactor vessel—could not be fully relied upon. Nevertheless, the data suggest significant thermal stratification in both horizontal pipes, with temperature differences across the pipe diameter (approximately 10 cm) reaching at least 50 K in SI-53 and up to 40 K in SI-52.

Temperature fluctuations point to more intense flow dynamics within the SI-53 pipe compared to SI-52. Cross-correlation of temperature signals reveals strong correlations among the upper RTD measurements on SI-53, a pattern not observed on SI-52. Additionally, correlations were found between SI-53 temperature, displacement, and strain measurements.

Transient events—likely caused by internal flow instabilities—are detected across most SI-53 sensors (temperature, displacement, and strain). These recurring events, lasting about one minute and occurring several times per hour, exhibit consistent characteristics: the upper part of the pipe near the former leakage site heats up by several degrees, while the lower part cools simultaneously. Two displacement sensors show that a section of the pipe first expands, then contracts even more intensely. Although the time profiles of these events are consistent, their amplitudes vary.

While the exact internal behavior during these events remains unclear from the measurements alone, the data, when paired with CFD analysis, prove highly valuable. For the most intense daily events, temperature jumps exceeding 20 K on the outer wall and around 50 K on the inner wall of the pipe cannot be excluded.

The primary water leak that occurred in October 2023 on the safety injection piping SI-53 at the Krško Nuclear Power Plant (NEK) prompted a collaborative investigation to identify both direct and root causes of the event. In addition to WEC (Westinghouse Electric Company) appointed by NEK, a consortium of four independent institutions – UL-FME (University of Ljubljana, Faculty of Mechanical Engineering), JSI (Jožef Stefan Institute), FER (University of Zagreb, Faculty of Electrical Engineering and Computing), and IMT (Institute of Metals and Technology) – undertook a comprehensive investigation addressing the metallographic, computational fluid dynamics and structural mechanical aspects of the leak. This paper focuses on two mechanical analyses carried out during the investigation to support the current hypothesis of the root cause.

The first analysis provides an explanation for the cold spring phenomenon observed during the replacement of the affected piping, which is considered a crucial factor in the root cause of failure. It suggests that during certain safety injection actuations at high Reactor Cooling System (RCS) pressure, the low flow rate of the injected cold water caused it to sink and accumulate at the bottom of the SI-53 pipe, reinforcing the existing stratified flow. This resulted in an enhanced vertical temperature gradient across the pipe cross-section, leading to bending of the 6-inch segment and the induction of plastic strains in the 4-inch segment. Consequently, compressive plastic strains developed at the top of the 4-inch pipe, producing high tensile residual stresses once the injection ended. A geometric imperfection at the weld located at the leakage site likely acted as a stress concentrator, making this location the most probable site for crack initiation. The combined effect of these residual stresses and the operational stresses generated high stress triaxiality (hydrostatic tension) at the weld, creating favourable conditions for corrosion processes. Together with local temperature fluctuations, this mechanism offers a plausible explanation for the initiation of the crack and the corrosion fatigue damage identified in the direct cause analysis.

The second mechanical analysis focuses on thermal-fatigue crack growth driven by the combined effects of static residual stresses and cyclic thermal loading in the pipe wall. These thermal loads result from the interaction between the stratified flow in the 6-inch portion of the pipe and the downcomer flow. The objective is to demonstrate that thermal fatigue is a plausible damage mechanism responsible for propagating the crack. This is achieved by identifying feasible thermal load conditions that reproduce the ~1 µm striation spacings observed on the fracture surface. The analysis follows four consecutive steps: (i) 1D heat conduction in the uncracked 304 stainless steel pipe to generate temperature fields, (ii) 1D linear-elastic stress analysis of the uncracked pipe to compute far-field stresses, (iii) linear-elastic fracture mechanics of the cracked pipe to determine stress intensity factors at the crack front, and (iv) 2D crack growth modelling to calculate crack increments and total crack extension. In step (i), temperature fluctuations of the inner surface of the pipe are simulated using a uniform half-sinusoidal signal with a prescribed period and an unknown amplitude ∆T. In step (iii), a semi-elliptical crack of known size is assumed at the leak location. To account for the possibility that multiple thermal cycles contribute to a single striation, a new variable – CPS (cycles per striation) – is introduced in step (iv). The results demonstrate that the observed crack growth can be explained by incorporating the residual stresses identified in the first mechanical analysis, along with thermal load amplitudes in the range of 10°C < ∆T < 30°C and CPS values between 10 and 100. These findings are consistent with the temperature fluctuations of the inner surface estimated from the outer-surface measurements, thus confirming thermal fatigue as a likely mechanism for the observed crack propagation.