Operators of nuclear power plants have two goals: safety and reliability. Unexpected events can reduce confidence in safety cases and result in costly down time. Access to a suite of tools for materials research helps plant operators respond quickly to any such incidents by generating better understandings of underlying causes.
On January 16, 1997, the Point Lepreau Generating Station (PLGS) was shut down for repairs due to discovery of a heavy-water leak. This unexpected outage cost New Brunswick Power (NB Power) about $400,000 per day to replace the power generation, which represented about one-third of the provinces’ electricity. Although the leak did not pose an immediate safety threat, NB Power urgently needed to provide assurance to the regulator that the issue was understood and could be managed effectively.
The leak was found in a crack in one of the hundreds of steel pipes, known as “feeders,” that circulate hot water from the reactor core to the steam generators and back. All feeders contain several bends, and the crack was in one of the bends.
NB Power and Atomic Energy of Canada Ltd. (AECL) immediately began a failure analysis, which included tests on both the failed feeder bend and an archived feeder bend that was manufactured in the same way. One of the first critical steps in the failure analysis was non-destructive stress measurement with neutron beams.
The Canadian Neutron Beam Centre (CNBC) granted AECL immediate access to a beamline. Within about 16 hours of receiving the feeder, the CNBC had established that the stress in the archived feeder bend was very high, nearly at the point of yielding, at the location corresponding to where the crack appeared in the failed feeder; in fact, it was a worst-case scenario that could result from the manufacture of the bend, short of cracking. The feeder pipes in PLGS had been fabricated from carbon steel followed by cold draw bending of the pipe, assisted by local flame heating.[1] The high residual stresses that resulted from this manufacturing process would increase the crack growth rate and flow-accelerated corrosion. Thus, ‘stress-corrosion cracking’ was determined to be a major root cause of the failure.
Meanwhile, NB Power completed a weld repair to the damaged feeder and conducted extensive examinations of the reactor. NB Power found that the cracked feeder was not configured as per the design,[2] leading it to propose that the crack might have initiated by mechanical means—potentially a low-probability, unique event in this case—after which the high stress would accelerate the crack growth. NB Power’s primary risk-management strategy was to demonstrate that it presented a low safety risk,[3] and further leaks might be prevented by improved configuration control. This strategy was sufficient to assure the Atomic Energy Control Board (AECB) to allow the restart.[4]
This unexpected outage ultimately lasted about 60 days, costing about $24 million in electricity, and another $7 million in repairs, inspection and related work.[5],[6]
Following restart, inspections and research on the issue continued to meet commitments to the AECB, and AECL accessed the CNBC again that year for more stress measurements to further clarify the underlying mechanism by examining the effect of flow-accelerated corrosion on the stress.
On March 7, 2001, another heavy water leak through a cracked feeder bend caused a second unplanned outage. The Canadian Nuclear Safety Commission (CNSC), the successor to the AECB, required NB Power to thoroughly investigate and fix all problems before it would grant authorization to restart.[7] NB Power performed inspections and found two more partial cracks, demonstrating that feeder failure was more widespread than previously supposed.
Once again, AECL accessed the CNBC in the failure analysis, to examine the spatial distributions of the stress, and found, just as in the 1997 incident, that the crack location was in a region of high stress, providing more evidence that the manufacturing method used to create the bends was the key issue.
The cracked feeder bends were cut out and new bends were welded into their place. The CNSC subsequently made the restart conditional on a more aggressive inspection program to avoid further surprises, and approved the restart for one year. The repairs cost an estimated $4 million, in addition to about $650,000 per day for replacement electricity for 39 days until it ultimately restarted on April 16.[8] The outage also cost Maritime Electric, the PEI utility, up to $100,000 per day in higher cost of replacement electricity.[9]
The immediate impacts of these failure analyses was to provide strong evidence quickly that the problem was sufficiently understood and that the station could be restarted safely, thus avoiding further outage costs in these incidents, which already totaled over $50 Million. AECL could not have obtained the stress data non-destructively or in a timely manner by any other means. Neutrons are the only probe able to non-destructively measure the stress deeply in the walls of the bends, which can be up to about 7 mm thick of steel. Other non-destructive techniques usually measure at most 1.5 mm depth. Neutron measurements of stress generally start at 1 mm and go up to 25 mm in ferritic steels.
The two events at PLGS had repercussions throughout the industry and posed a challenge for PLGS to managing the risk of cracking in the years ahead. The CNBC make important contributions to the industry’s responses to these events, which are explored in the next articles in this series.
Next: Part 2: Managing Continued Risk of Feeder Cracking at Point Lepreau
This research story was republished with the permission of the Canadian Institute for Neutron Scattering.