Johannes de Wit
The Hague
Brittle fracture of older storage tanks, that can lead to catastrophic spills of hydrocarbon materials, can be reduced by applying preventive measures based on good knowledge of the critical properties of the tanks, and the critical conditions in which the tanks are used.
Some preventive measures include: insulating the tank, not filling the tank to maximum level during lowest temperatures, fully water-testing the tank, repairing serious corrosion, releveling the tank, and avoiding shock loads on the tank shell.
During the last 40 years, there have been only a few brittle fractures of storage tanks, but they resulted in serious spills. Records show that almost all of the brittle fractures were in tanks built in the period 1945-1955, before the phenomenon of brittle fracture was well understood or addressed by international codes and standards, such as BS 2654 of the British Standards Institute, or API 650 of the American Petroleum Institute.
Consequently, at the time these tanks were constructed, the requirements for material selection, welding, and weld inspection were not directed at preventing brittle fracture.
To see how the fundamentals of brittle fracture play a part in the catastrophic failure of older storage tanks, this article will examine and compare four tanks that failed due to brittle fracture.
These tanks are: a 42.7-m dia. by 16.5-m high crude oil tank with a floating roof that collapsed in February 1952 at the Esso refinery in Fawley, U.K., during a water test; a 45.7-m dia. by 14.64-m high gas oil tank with a fixed roof that collapsed in March 1952 at Esso's refinery in Fawley; a 45-m dia. by 12.5-m high fuel oil tank with a fixed roof that collapsed in December 1970 at the Amer power station in The Netherlands (Fig. 1); and a 36.6-m dia. by 14.64-m high fuel oil tank with a fixed roof that collapsed in January 1988 at an Ashland Petroleum Co. terminal near Pittsburgh.
CAUSES OF BRITTLE FRACTURE
Investigations, experience, and test work have shown that a brittle fracture in a tank shell is always caused by a combination of specific factors. Important factors that directly influence brittle fracture include:
- The minimum design metal temperature of the tank shell
- The notch toughness of the shell plate material
- The plate thickness
- The crack-like defects present
- The acting tensile stress on the tank shell
- The overloading effect of a water test.
DESIGN TEMPERATURE
The minimum design metal temperature is important because the toughness of the steel plates will decrease at low temperatures in the 0-25 C. range. Low toughness will increase the risk of brittle fracture.
It should be noted that all brittle fractures of the four tanks examined occurred during a winter period at low temperatures. The temperature of the tank shell is influenced by the ambient temperature at the installation and by the temperature of the liquid stored.
In winter, a product's temperature may be considerably higher than the ambient temperature. The tank shell will be warmed by the liquid stored, so that the temperature of the tank shell will always be somewhat higher than ambient. Wind around the tank may reduce the warming effect of the liquid.
The minimum design metal temperature can be determined from a formula specified by BS 2654: "The design metal temperature shall be the lowest one-day mean ambient temperature (Lodmat) + 10 C." Lodmat = (max.temp. + min.temp.)/2.
The additional 10 C. takes the average warming effect by the stored liquid into account. API 650 uses a similar formula.
TOUGHNESS
Shell plates should have sufficient notch toughness at the minimum design temperature to prevent the possibility of a brittle fracture under the most unfavorable loading conditions. Notch toughness requirements are generally expressed as Charpy-V impact values at a specified temperature, i.e., 20 ft-lb (27 Joules) at - 20 C.
Charpy-V tests are used worldwide as acceptance tests and are made on small test specimens of the tank shell plates. The test method does not simulate the behavior of large shell plates under real loading conditions.
The tests also neglect the influence of some factors, such as the plate thickness, the stress level, the size of the tank shell, loading rate, etc.
This is why the British Standards Institute decided that large-scale tests should be used to simulate the real behavior and loading conditions of the tank shell.
The large-scale test used is the Wells wide-plate test (WWP). The notch toughness requirements in the material selection specified in BS 2654 are thus based on a number of WWP's on carbon and carbon-manganese steel plates carried out at the British Welding Institute, and on all other known WWP test data.
A graph from BS 2654 for the determination of toughness is shown in Fig. 2. Fig. 3 shows a typical example of a wide-plate test.
The test plate has dimensions of approximately 1 m by 1 m. The test plate shown has a weld in the center. Near the weld, the test plate is provided with a substantial notch. The plate thickness and the measured Charpy-V impact value of the plate are known properties for the test.
In a WWP rig, the test plate is loaded with a tensile stress and stressed to failure. The notch toughness requirements of BS 2654 are fulfilled for the test temperature when the test plate does not fracture before about four times the yield point strain is reached at the tip of the notch.
The test plate is provided with a thickness defect through the entire plate. The defect is 1 0 mm and located in an embrittled zone.
It is reasonable to assume that such serious defects would not be produced during construction or would go undetected by proper weld inspection.
CRACK PREVENTION
The philosophy of BS 2654 for material selection is based on prevention of crack initiation. The notch ductility requirements of BS 2654 (Fig. 2) are based on WWP test results, but for material acceptance at steel mills, the Charpy-V values indicated can be used.
The WWP tests show that to have the same crack-initiation resistance to brittle fracture, thicker plates must demonstrate a better notch ductility than thin plates. On the basis of these test results, the toughness requirements of BS 2654 are more stringent for thick plates (2540 mm) than specified by API 650.
For example, BS 2654 requires a 35-mm plate at -5, C. design metal temperature, and a guaranteed Charpy-V impact value (longitudinal) of 20 ft-lb at -34 C.
In the literature, there appears to be some misunderstanding about the possibility to select shell plate materials on the basis of crack-arrest properties. Good crack-arrest properties exist when the material has a toughness high enough to stop a fast-running crack, 500-mm long.
Some have suggested that when a small defect is growing in a material (when the growth is due to corrosion or fatigue) but does not produce a fast-running crack, the material has crack-arresting properties. This is not correct.
The fact that no fast-running crack is produced simply indicates that the defect has not grown enough to initiate a crack. The defect, although growing, is still less than the critical defect size for that material.
The shell materials required by BS 2654 and API 650 certainly do not have such crack-arrest properties. The material selection specifications of those standards is based on the prevention of crack initiation.
That philosophy has proven successful because no brittle fracture of any tank or pressure vessel built to BS 2654 has occurred during the last 30 years. It can, therefore, be concluded that BS 2654 and API 650 are adequate to prevent brittle fracture under specified design conditions, and no changes need to be made in the material selection specifications contained in them.
There is no internationally accepted standard test procedure to demonstrate crack-arrest properties. Some individual research has been done on crack-arrest, mainly for refrigerated, liquified-gas storage tanks, but no requirement of determining crack-arrest properties is contained in any standard, even for these types of tanks.
PLATE THICKNESS
Wide-plate tests clearly show that the thickness of the shell plate has a considerable influence on brittle fracture. The risk of brittle fracture increases as shell plate thickness increases, for the same quality material.
Thick plates are more vulnerable because of the three-dimensional stress pattern around defects in thick plates. Most serious fractures in tank shells started in plates exceeding 20-mm thickness.
This is supported by the fact that the brittle fractures that occurred in the four tanks examined in this article all came from defects in shell plates that were more than 20-mm thick.
DEFECTS
The philosophy for material selection of the British Standards Institute is based on the assumption that defects of limited size will always be present in shell welds. The defects will be less serious than the 10-mm through-thickness defect of the WWP test.
A shell plate in a tank will also never be loaded to the high tensile stress levels reached in the WWP test.
The WWP test does not include the favorable effect of overloading during water testing at higher temperatures than the metal design temperature. During the water test, the crack tips of the defects are blunted by yielding of the material.
These crack tips will be less serious when the tank is loaded during service at low temperature. Therefore, water testing will considerably decrease the risk of a brittle fracture.
TENSILE STRESSES
BS 2654 specifies an allowable design stress of 66% of the yield strength of the plate material during the water test. Because the specific gravity of the liquids stored in the tank are generally between 0.7 and 0.9, the nominal stress in the shell of a full tank in normal service will not exceed approximately 50% of the yield strength of the plate material.
This stress limit is based on a joint efficiency of 1.0 in the stress calculations.
Tanks built before 1965 were designed with a joint efficiency of 0.85, and an allowable design stress of 21,000 psi (144.8 N/sq mm). This results in a maximum nominal stress of 17,850 psi in the tank shells of older tanks. (The nominal stress is the allowable design stress multiplied by the joint efficiency.)
On the basis of extensive laboratory investigations, it is accepted worldwide that brittle fracture will not occur in plates where the nominal stress is less than 10,900 psi, the Robertson plateau.
WATER TESTING
It is a mandatory requirement of BS 2654 that every storage tank be fully water tested before it is put into service. Because the specific gravity of the liquids to be stored are generally between 0.7 and 0.9, the water test will give an overload of 1.1 to 1.4 times the service loading.
The water test should be carried out at a higher temperature than the minimum design metal temperature of the tank shell, which means that the toughness of the shell plates will be higher than the specified minimum values.
During the water test, locations with high stress concentrations (at discontinuities, or defects) will yield so that the stress peaks will be removed and crack tips blunted. Because of the lower density of the stored liquids vs. water, the in-service peak stresses at these stress concentrations will never reach the stress levels reached during the water test (Fig. 4)6
The water test is an important safety measure against brittle fracture of the tank shell, but experience shows that its influence is many times underestimated.
That is why BS 2654 calls for a water test before a tank is put into service. A water test should also be done after modifications to the tank, such as repairs to the shell or releveling of the tank, because modifications can change the pattern of the stresses in the tank shell that create new stress peaks.
Water testing should not be done during periods of low temperature. In addition to the obvious freezing problems, if the tank shell material temperature is too low, toughness will also be low. Yielding of the material in the area of stress concentrations may not be adequate to relieve stress peaks and blunt crack tips.
EMBRITTLEMENT
The factors that influence brittle fracture were involved in the failure of the four tanks listed at the beginning of this article. These tanks were not built to modern standards, and their failures would most likely have been avoided if they had been.
Analyzing the failures of these older tanks with respect to the factors that influence brittle fracture can help determine ways to avoid brittle fracture and failure in other old storage tanks.
Details of each of the four failed tanks, and the location of where the brittle fracture started in the shell, are given in the accompanying box.
All four tanks collapsed at low winter temperatures. The material stored in Tank 3 was changed to waxy fuel oil a few months before the failure.
Wax formed an insulating layer on the inside of the tank shell, preventing the shell from being warmed by the product. The shell temperature was equal to the ambient temperature of - 7 C. at the time of the failure.
Furthermore, the waxy fuel oil also contained thin water layers which evaporated explosively when they entered the outlet heater located inside the tank, causing shock loads on the tank shell.
Tank 4 was put into service without an adequate water test. Therefore, there was no opportunity for the material to yield at a higher temperature, and the crack tip at a defect was not blunted.
All four tanks were filled to more than 75% of the shell height. And the shell thickness of the lowest shell course in each of the tanks exceeded 20 mm.
LOW TOUGHNESS
In all four tanks, the plate material of the lowest shell courses of the tanks was supplied without guaranteed impact values. Impact tests were conducted on the plate material after the tanks failed.
Test results show that the materials of all four tanks had low toughness in the area of the brittle fractures. Fig. 5 shows the transition curves for the shell plates of all four tanks (Charpy-Y. longitudinal).
The transition curves show that the Charpy-Y impact value for the material in each of the failed tanks was below the values specified in Fig. 2.
Generally, an accepted toughness level is 15-20 ft-lb at the minimum design temperature.
Each of the four tanks had Charpy-Y impact values approaching or exceeding 1520 ft-lb at 20 C., but significantly less at the temperatures at which the tanks failed.
HIGH STRESS LEVELS
Assuming that the four failed tanks were designed for an allowable tensile stress of 21,000 psi with a joint efficiency of 0.85, the nominal stresses in the lowest course of each shell at the time of brittle fracture were: Tank 1, 15,880 psi (51% of yield); Tank 2, 14,510 psi (47% of yield); Tank 3, 12,050 psi (39% of yield); and Tank 4, 13,700 psi (44% of yield).
These stresses are based on a yield strength of 31,000 psi for the plate materials used for the lowest shell courses.
The stress levels in each of the four tanks were above the Robertson plateau.
To reduce the risk of brittle fracture, the minimum Charpy-V impact values for the shell plates should have been: Tank 1, 20 ft-lb at - 5 C.; Tank 2, 20 ft-lb at O C.; Tank 3, 20 ft-lb at 5 C.; and Tank 4, 20 ft-lb at -7 C.
WELD QUALITY LOWER
When the failed tanks were built, weld quality was generally less than that of tanks built today. Weld quality in modern tanks is controlled at the construction site by using spot radiographic inspections of the welds.
Radiography ensures that the weld quality is maintained at that specified by the welding procedure. The welding procedure is tested and approved before actual construction begins.
Between 1945 and 1955, radiographic inspection was not done on tank welds ' Therefore, it is fair to assume that welds of that period include defects, such as porosity, slag inclusions, incomplete fusion and penetration, and small crack-like defects.
For these tanks, it is particularly important that the tanks be fully water-tested to remediate some of these weld defects.
SUGGESTIONS FOR OLD TANKS
Based on the experience gained from the failed tanks, there are several areas that need close attention for tanks not built to current British Standards Institute and API specifications.
Special attention is required for tanks that have shells made of non-impact-tested plate materials. These materials are BS 13, ST.37, A283D, and A283C. For those steels, it should be assumed that the Charpy-V impact value (longitudinal) does not exceed 20 ft-lb at 20 C.
Using this assumption, the maximum allowable plate thickness for the specified minimum design metal temperature can be determined from Fig. 2.
Allowing some relaxation of safety factor, some guidelines on average shell plate thickness can be offered to determine when old storage tanks may need special attention to reduce the risk of brittle fracture.
In the U.K., the guideline is 0 C. minimum design metal temperature for shell plates 21 mm and thicker.
In The Netherlands, the guideline is -5 C. minimum design metal temperature for shell plates 17 mm and thicker.
For the northern parts of the U.S., a - 15 C. minimum design metal temperature for plates 12 mm and thicker is recommended.
For tanks with shells with thicknesses equal to or exceeding these guideline thicknesses, it is recommended that tank shells be carefully inspected, and that each of the factors that influence brittle fracture be thoroughly investigated.
Certain steps can be taken to reduce the possibility of brittle fracture in tanks where the shell plate material has not been impact-tested, where the Charpy-V impact value does not exceed 20 ft-lb at 20 C., and where shell thicknesses exceed the guidelines.
- External insulation of the tank shell will bring the metal temperature of the shell closer to the product temperature. In many cases, the product temperature is 10 C. or above.
- Limiting the fill height of the tank to less than 75% of the shell height during winter conditions will keep stress levels down when material toughness is at its lowest.
- Fully water-testing the tank during warmer periods before placing it into service will remediate defects that can initiate cracks.
The water test should be repeated whenever the tank is modified, repaired, or releveled.
- If serious corrosion is observed in the lower half of the tank shell, particularly the lowest shell course where the thickest plates are located, that corrosion should be repaired before the tank is subjected to low winter temperatures.
- If uneven soil settling around the tank shell is noticed, releveling of the tank may be necessary and the tank water-tested.
- Finally, determine any sources of shock loading of the tank shell, and avoid them if possible.
BIBLIOGRAPHY
- Feely, F.J., and Northrup, M.S., "Why storage tanks fail," OGJ, Feb. 1, 1954.
- "The Tank Failure at the Amer Power Station," (in Dutch), Polytechnisch Tijdschrift, Jan. 5, 1972.
Copyright 1990 Oil & Gas Journal. All Rights Reserved.
