Ultrasonic high-cycle thermal fatigue monitoring
File(s)
Author(s)
Clarkson, Laurence
Type
Thesis
Abstract
High-cycle thermal fatigue (HCTF) failure affects piping within nuclear power plants (NPPs) exposed to temperature fluctuations (0.1–1 Hz) caused by mixing of fluids with different thermo-hydraulic properties. It remains a concern as HCTF damage was identified in NPPs as recently as 2023. Current assessment methods rely on scheduled ultrasonic inspections based on cumulative thermal transient exposure. Existing thermocouple-based monitoring techniques are fundamentally limited for low thermal conductivity materials at frequencies around 1 Hz. A continuous monitoring method that overcomes this limitation would improve data quality and hence extend inspection intervals.
Two ultrasonic methods for HCTF monitoring were investigated: the acoustoelastic effect (AE) and ultrasonic thermometry (UT). AE describes the relationship between ultrasonic
velocity and strain, while UT leverages the thermal sensitivity of ultrasonic velocity. The AE method was limited by its small magnitude (especially in austenitic stainless steel
(SS)) and confounding factors like anisotropy and the small, frequency-dependent volume under thermal stress/strain. UT was shown to be feasible using machine learning, assuming the training data reflects real-world conditions, although the (expected) computational speed benefits over physics-based UT were not found. Experiments showed that inverse thermal modelling (ITM) UT can detect large, rapid thermal transients in 304 SS, although over-prediction was caused by non-1D temperature fields. Relative changes in ultrasonic velocity and attenuation correlated with HCTF damage and crack initiation, respectively. Simulations showed significantly improved fatigue life estimates using ITM-predicted temperatures compared with external temperature measurements. Literature analysis suggested that 1D temperature fields are plausible under best estimates of the Civaux 1 NPP failure conditions. However, a distributed sensor network is required to capture spatio-temporal variations.
The work in this thesis demonstrates promise for ultrasonic monitoring of HCTF in NPPs; however, significant further investigation is needed to prove real-world viability.
Two ultrasonic methods for HCTF monitoring were investigated: the acoustoelastic effect (AE) and ultrasonic thermometry (UT). AE describes the relationship between ultrasonic
velocity and strain, while UT leverages the thermal sensitivity of ultrasonic velocity. The AE method was limited by its small magnitude (especially in austenitic stainless steel
(SS)) and confounding factors like anisotropy and the small, frequency-dependent volume under thermal stress/strain. UT was shown to be feasible using machine learning, assuming the training data reflects real-world conditions, although the (expected) computational speed benefits over physics-based UT were not found. Experiments showed that inverse thermal modelling (ITM) UT can detect large, rapid thermal transients in 304 SS, although over-prediction was caused by non-1D temperature fields. Relative changes in ultrasonic velocity and attenuation correlated with HCTF damage and crack initiation, respectively. Simulations showed significantly improved fatigue life estimates using ITM-predicted temperatures compared with external temperature measurements. Literature analysis suggested that 1D temperature fields are plausible under best estimates of the Civaux 1 NPP failure conditions. However, a distributed sensor network is required to capture spatio-temporal variations.
The work in this thesis demonstrates promise for ultrasonic monitoring of HCTF in NPPs; however, significant further investigation is needed to prove real-world viability.
Version
Open Access
Date Issued
2025-08-18
Date Awarded
2026-04-01
Copyright Statement
Attribution-NonCommercial 4.0 International Licence (CC BY-NC)
License URL
Advisor
Cegla, Frederic
Publisher Department
Department of Mechanical Engineering
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)