Figure 1: Reactor head sitting on the reactor vessel.
On November 15, 2011, operators began restarting the Unit 2 reactor at the Brunswick nuclear plant located south of Wilmington, North Carolina. The reactor had been shut down 11 days earlier to find and remove a damaged fuel assembly from the reactor core.
About 17 hours into the startup, the operators observed indications of unusually high leakage of reactor cooling water into the containment building. With all the valves and piping, a small leakage rate is normal. But instruments inside containment detected unusually high leakage. A maintenance crew was sent into containment to repair a valve that had been leaking a little before the outage.
The workers completed maintenance on the leaking valve, but the leak rate continued increasing. At 2:12 am on November 16, the indicated leak rate reached 5.88 gallons per minute. This leak rate exceeded the maximum limit established by the plant’s operating license. Workers had a short time to find and fix the leak or the reactor would have to be shut down.
Workers re-entered the containment building and observed water dripping from equipment and running down the containment walls. The operators in the control room noted that the temperature in the upper part of the containment was 240°F, or 40°F above the normal temperature in this area with the reactor at 100 percent power.
The leak rate continued to increase. By 2:53 am, it reached 10.11 gallons per minute. The head of operations ordered workers out of the containment building. At 3:01 am, the operators declared an Unusual Event—the least serious of the NRC’s four emergency classifications—due to the high rate of leakage inside the containment. The operators manually scrammed the reactor (i.e., depressed two pushbuttons to cause all the control rods to rapidly insert into the reactor core and interrupt the nuclear chain reaction) at 3:09 am. As the temperature and pressure inside the reactor vessel decreased following the reactor’s shut down, the leak rate also decreased. It dropped to 0.13 gallons per minute by 2:38 am on November 16.
The following day, workers on the refueling floor were able to turn several of the retaining nuts for the reactor vessel head by hand. This should not have been possible. As shown in Figure 1 above, the reactor vessel head is the domed-shaped part on top of the cylindrical reactor vessel. It is fastened to the reactor vessel by numerous heavy-duty bolts and nuts around the dome as shown in Figure 2.
When workers had reassembled the reactor vessel after the repairs earlier in November, they used a stud tensioning rig to tighten the nuts onto the bolts. Figure 2 shows four stud tensioners (in yellow) suspended over four nuts about to be tightened. Procedures guided the workers on the proper sequence to install the nuts and directed them to tighten the nuts to 13,000 pounds per square inch force. The stud tensioner had a digital read-out. The workers tightened each nut until the digital read-out said 1,300. They believed that the value in the read-out window was ten times the pressure. In reality, it displayed the pressure. They had tightened the nuts to one-tenth of the proper pressure.
The procedure used a second method of assuring proper tensioning. As the nuts were tightened onto the bolts, the applied pressure caused the bolts to elongate. The procedure had the workers measure the bolt lengths before and after tensioning and ensure each bolt elongated by 0.041 to 0.049 inches. As workers measured the bolt lengths after tensioning, they noted results ranging from -0.001 to 0.004 inches. They assumed that these values were not actual lengths but rather the deviation from the target elongation of 0.045 inches. Based on the measurement results, they thought all nuts had been tensioned within the acceptable range. But once again, they misunderstood what the instruments were telling them. The read-outs actually showed the elongation, or lack thereof. Because the nuts had not been properly tensioned, the bolts had not properly elongated.
A worker and a quality control inspector signed off each bolt and nut as being properly tensioned when in fact, none were.
The NRC discovered that formal training on reactor vessel disassembly and reassembly had not been conducted at Brunswick since 2000. Only 4 of the 13 workers who reassembled the reactor vessel head in mid November 2011 had been formally qualified to do the tasks.
Because the reactor vessel head had not been properly reassembled, increasing pressure inside the reactor vessel during the reactor startup lifted the head enough to squirt water out past the flanges. The water was hot, increasing the temperature in the top portion of the containment. And the water drained down into the basement of the containment where instrumentation recorded the increasing leakage rate.
The direct safety implications of this event are minor. Instrumentation inside containment detected the leakage and the plant’s operating license limited the leak rate low enough that a shut down was required before reactor cooling was jeopardized.
However, the indirect safety implications of this event are troubling. Administrative measures such as personnel qualifications, training, and second party verifications are intended to yield quality outcomes. Here, few of the workers were qualified or trained for the assigned tasks. And none of the parties involved understood what they were doing. In other words, none of the administrative measures succeeded.
In this case, the breakdown in administrative measures became self-evident—the reactor vessel started spraying down the containment with increasing amounts of hot water. But what if unqualified and untrained workers with little understanding of the task at hand performed maintenance on the emergency diesel generators, or emergency core cooling pumps, or standby gas treatment system, and such? Would those failures also conveniently reveal themselves? Or might they remain undetected until called upon to mitigate an accident? The time to discover that a car’s air bag as a hole in it is not as that car wraps itself around a tree.
The NRC discovered that the company stopped training workers on reactor vessel disassembly and reassembly procedures in 2000. The lack of training contributed to this event. Was other training also eliminated or reduced as a cost saving measure? If so, are there other latent safety problems lingering at Brunswick due to workers not having been properly trained? Time will tell.
“Fission Stories” is a weekly feature by Dave Lochbaum. For more information on nuclear power safety, see the nuclear safety section of UCS’s website and our interactive map, the Nuclear Power Information Tracker.