WIT Press

Probabilistic Methods For Prediction Of Corrosion In A Service Water System


Free (open access)

Paper DOI









470 kb


E. Houston, G. Licina & D. Dedhia


Nuclear plant service water systems are a critical part of the facility’s infrastructure. System integrity and performance are vital for plant reliability and essential to achieving a plant life of 40 years and beyond. The low temperature and pressure service water piping systems are primarily degraded by corrosion in untreated waters. Corrosion allowances, based upon very simplistic considerations of general corrosion in untreated raw water, were a part of the original design. However, long term service in many such systems has shown that localized corrosion phenomena, from microbiologically influenced corrosion, pitting, and underdeposit effects, have compromised system integrity. Because of the complexity and random nature of corrosion processes, it is nearly impossible to develop a mathematically deterministic model (like the typically used corrosion allowances) that accurately predicts pipe wall loss. However, when statistical distributions are used to describe the various corrosion processes, mathematical algorithms that incorporate all of the distributions, iterated a statistically significant number of times, can be used to forecast the most probable number of leaks. This approach was used to predict the condition of service water piping at a US Nuclear Power Plant, comparing results to service experience and inspection results. The results were ultimately used by the plant for targeting inspections and for long term planning of replacements and replacement schedules. Keywords: nuclear power plants, service water systems, pitting, MIC, water treatments, probabilistic methods.


nuclear power plants, service water systems, pitting, MIC, water treatments, probabilistic methods