This article presents a review of nuclear power plant operators who must design many of their systems and structures to withstand potential natural disasters such as earthquakes, floods or tornadoes. Tornadoes are a part of the design considerations at the Arkansas Nuclear One facility in Russellville, Ark., a state that averages 19 twisters a year. Arkansas Nuclear One is one of four nuclear energy sites run by Entergy Operations Inc., based in Jackson, MI. Entergy engineers saw an opportunity to avoid bringing out the air cannon again by replicating the 13-year-old laboratory test on a computer. The engineers at Entergy used Algor’s Accupak/VE Mechanical Event Simulation software to recreate events involving motion and its consequences, including inertial effects, impact, permanent deformation, and residual stresses. Bill Hovis, P.E., a senior design engineer with Entergy’s mechanical, civil, and structural engineering department, performed the computer analyses.
A lot can change in a dozen years. A nuclear power plant, for instance, may find that it needs to hold less water in its reserve tank. It also may discover that there are faster and cheaper alternatives to expensive laboratory testing for safety.
Nuclear power plant operators must design many of their systems and structures to withstand potential natural disasters such as earthquakes or floods. Tornadoes are a part of the design considerations at the Arkansas Nuclear One facility in Russellville, Ark., a state that averages 19 twisters a year. Arkansas Nuclear One is one of four nuclear energy sites run by Entergy Operations Inc., based in Jackson, Miss.
In 1985, Entergy Operations was required to assess the structural integrity of a new water storage tank in tornado conditions. At the time, the tests required expensive specialized services and laboratory facilities. One test had engineers using an air cannon to fire a scale model of a car at a scaled-down tank.
The emergency condensate storage tank is one of the safety features at Arkansas Nuclear One. At 30 feet high and 42 feet in diameter, the tank can hold up to 321,000 gallons of demineralized water. The stainless steel tank is the primary source for the emergency feedwater system, which is ready if the main feedwater supply is interrupted. The emergency feedwater system ultimately uses the water to cool the nuclear reactor if the main feedwater system fails.
If the condensate storage tank was unavailable to supply water for the emergency feedwater system, plant operators would use water from another source such as a nearby lake. But contaminants in lake water could reduce the life of the steam generator. Entergy Operations needed to test the design of the tank and determine if a tornado could cause it to rupture and lose its water supply.
Using Hand Calculations
When the tank was designed in 1985, engineers based their assessments on hand calculations and laboratory testing. First, they performed calculations addressing the potential damage caused by relatively small, flying debris from a tornado. Objects with low mass move at high velocity in tornado winds. Engineers considered the effects of a steel pipe and a piece of timber. Their investigation established that these objects could rupture the tank under severe tornado conditions, so a 5-foot-tall, 18-inch- thick concrete barrier was constructed around the base to protect it. A hole above the 5-foot barrier would leave the tank with enough water to cool the steam generator while plant operators prepared to use water from an alternate source.
The company also needed to determine the possible damage caused by large flying objects, which move slower than small ones in tornado winds, but still could threaten the tank because of their large mass. Engineers considered the impact of a 4,000-lb. automobile tossed by a tornado.
Given the difficulties of throwing a full-scale car at a full-size tank, Entergy used an air cannon to fire a scaled-down model of a car at a scaled-down tank filled with water.
Entergy referred to its safety analysis requirements to determine that the propelled object could travel as fast as 50 mph and hit the tank as high as 25 feet above ground in a hypothetical tornado. In this laboratory test, Entergy found that the car model caused permanent deformation at the point of impact, but would not rupture the tank.
Last year, more than a dozen years after the original laboratory tests, Entergy came up with a new idea concerning its storage tank and needed to run tests to see if the idea was safe.
Entergy had determined that lowering the water level from the tank’s full 30 feet to a depth of 11 feet would give plant operators greater flexibility during maintenance and inspection, yet would still provide enough water, 118,000 gallons, to effectively operate the emergency feedwater system. No one was sure how the lower water level might affect the tank’s structural integrity, or if the partly empty tank would be safe during a tornado. The tornado tests would have to be duplicated.
No more Air Cannons
But this time around, Entergy engineers saw an opportunity to avoid bringing out the air cannon again by replicating the 13-year-old laboratory test on a computer.
The engineers at Entergy used Algor’s Accupak/VE Mechanical Event Simulation software to recreate events involving motion and its consequences, including inertial effects, impact, permanent deformation, and residual stresses.
Bill Hovis, P.E., a senior design engineer with Entergy’s mechanical, civil, and structural engineering department, performed the computer analyses.
Hovis decided to first try replicating the laboratory tornado test involving a car model striking a full tank of water, to see if the software’s analysis correlated with the 1985 laboratory test results. Comparing the results of the old lab experiment and the computer simulation would let him verify the accuracy of his analysis setup.
Working on a PC running Windows NT 4.0, Hovis designed a 3-D model of the scaled-down condensate storage tank. The tank was drawn to the same scale, 1:12.4, as the 1985 laboratory tank model.
Hovis generated a finite- element mesh made up of plate elements to represent the tank’s geometry. For the model of the scaled-down tank, he specified the thickness as that of 28- gauge steel, or 0.0149 inch. The original tank is about 3/16-inch thick at its top tier and 5/16-inch at its lowest tier. Although the real tank is tiered to thicken at its base, Hovis applied the real tank’s thinnest measurement (scaled 1:12.4) to the entire model, as researchers did in the 1985 laboratory test. The decision simplified analysis by setting conservative conditions. Hovis used solid brick elements to represent stiffening rings in the area where the tank’s cylindrical body meets its dome ceiling.
Less Analysis Time
To reduce analysis time, Hovis modeled a 90-degree cross-section of the tank because the laboratory test showed that the buckled region around the point of impact included 54 degrees of the tank’s cylindrical body. He added 18 degrees on either side in anticipation of additional displacement due to the lower water level. He constrained the bottom portion of the cylinder that connects to the floor with translational boundary conditions.
For the tank, Hovis specified stainless steel material properties, which he obtained from the American Society of Metals source book on stainless steel. He used a nonlinear material model based on the von Mises yield criterion with simplified average values used to represent the strain hardening portion on the stress-strain curve. He also selected the Total Lagrangian analysis formulation because the von Mises material model was based on the steel’s engineering stress-strain curve. The Total Lagrangian option is also ideal for analyses involving large strain, which Hovis expected based on the large deformation found in the laboratory-test tank at the point of impact.
Hovis created a 3-D solid brick element model that was scaled to the same proportions as the car model used in the laboratory test. He used contact elements between the car and the tank models to simulate their interaction, including the transfer of inertia from the car to the tank.
Next-Generation Tornado Forecasting
TESTING HAS BEGUN on the next generation of tornado forecasting technology. The system, designed to guide forecasters in predicting hazardous weather patterns, helped save lives in the devastating series of tornadoes that hit Oklahoma last May. Although 41 lives were lost, the National Oceanic and Atmospheric Administration believes countless more were spared because of the earlier warnings of approaching danger. Some warnings went out as much as an hour before the storm. According to researchers at the Georgia Tech Research Institute, the system could increase warning time by as much as 50 percent in the state.
Researchers are working on the National Severe Storms Laboratory’s Next Generation Warning Decision Support System. The warning system was installed at the National Weather Service’s Peachtree City office earlier in May, and two more systems were scheduled to be in Georgia Tech laboratories by the end of June. One of the labs is on Georgia Tech’s main campus in Atlanta; the other is northwest of the city.
According to Gene Greneker, a research scientist who heads the institute’s new Severe Storms Research Center, the system can be tuned to perform more efficiently in Georgia. “Tornadoes in Georgia and elsewhere in the South-east are often short-lived events,” Greneker said. “They can come and go in 10 minutes, as opposed to an hour in Kansas. As a result, the radar signal processing may need to be set slightly differently from those that were developed for the Great Plains states."
Researchers will collect storm data and determine if changing parameters in the system’s algorithms will make it work better in Georgia.
The system includes advanced image processing, artificial intelligence, neural network, and other algorithms that use Doppler radar data. The information is integrated with other weather sensor data to help forecast tornadoes, severe thunderstorms, and flash floods.
The Severe Storms Lab has tested the Warning Decision Support System successfully around the country since 1994. An earlier version of the system went to Peachtree City in preparation for the Olympic Games, when it became part of the temporary Olympics weather support service.
Funding for the test systems in Georgia comes from the Georgia Emergency Management Agency, the Federal Emergency Management Agency, and the Georgia General Assembly. Bell South Business Systems is providing funds to pay for high-speed data transmission lines.
In his first analysis iterations, the car model passed through or stuck to the tank wall. Hovis saw it on the screen and halted the analyses to adjust the setup. He found that, to obtain an accurate contact scenario, in addition to adjusting the time step increment, he had to adjust the contact elements’ stiffness, cross-section area, and distance at which contact begins, the software’s three contact element settings.
Hovis applied an acceleration load curve for gravity and adjusted the car model’s originating position to account for its drop due to gravity, so that it would strike the tank at the same height as it did in the laboratory test. This was the maximum height at which a car would travel in a hypothetical tornado. The car’s velocity was set to 51.4 mph, the same speed measured for the laboratory test car model at the point of impact.
Hovis set the duration of the simulation to 0.1 second. To correspond to the high-speed camera that recorded 298 frames per second in the 1985 laboratory test, he initially set the timestep increment, the number of moments at which analysis information is recorded, to one every 0.0034 second. This setting was too slow and did not display the necessary details around the time of impact. Hovis shortened the timestep to 0.00056 second, or about 1,800 steps a second. He also reduced the duration of the event to 0.084 second to save analysis time.
The software analysis found large deflection at the point of impact that measured within a quarter-inch of the deflection recorded in the laboratory test.
After successfully replicating the laboratory test involving the full water tank, Hovis conducted a second scenario. This time, he applied hydrostatic pressure along the tank wall representing the lower water level. Again, the analysis results found large deflection at the point of impact, but no permanent deformation near the tank’s base. The software analysis results also revealed that stress did not exceed the steel’s yield point near the tank’s base. These results proved that the condensate storage tank could withstand this tornado impact load while storing only 118,000 gallons of water.
“I created a visual replay of the event that enabled colleagues to see a complete re-enactment of the car flying toward the tank, striking it, causing permanent deformation, and bouncing away toward the ground,” Hovis said. The replay could be reviewed with any application capable of viewing an .avi file, such as the Windows NT, 95, and 98 utility media player.
Hovis also decided to analyze the effects of severe tornado winds and pressure on the tank with the lower water level.
First, he applied internal and external pressure load cases to the tank’s cylindrical shell that represented an atmospheric pressure drop. He then combined these with pressures to the dome ceiling that represented uplift caused by the tornado’s wind speed. Each analysis used a time step of one-quarter second and increased the pressure load over the first few seconds. The analysis results found high, but acceptable stress values where the dome ceiling connects to the tank’s cylindrical walls. These results affirmed that the condensate storage tank could withstand this type of tornado loading while storing only 118,000 gallons of water.