Professional Documents
Culture Documents
1 Introduction research report used SAFEROOF software to show the results of fail-
ure pressure on the roof-to-shell connection with the axisymmetric
Vertical vaulted storage tanks can be damaged by increases in
element [9]. To obtain the failure pressure in the roof-to-shell con-
internal pressure during normal use. Failure usually occurs at the
nection, some researchers also considered the influence of edge
roof-to-shell connection or the bottom-to-shell connection of the
load [10–12] or used the finite element method and experimental
tank. The bottom-to-shell connection of the tank is mainly failed
method to modify the calculation formula of the stress distribution
by the liftoff of the bottom of the tank under the inner pressure.
of the tanks [13–16]. However, most of the calculations and analy-
The large circumferential stress at the bottom-to-shell connection
ses focused on predicting the local yield buckling under internal
will cause plastic hinge and local instability and lead to failure of
and external pressure [17–19]. There is currently a lack of
the tank [1,2]. In all the failure modes, the roof-to-shell connec-
research regarding calculating the strength failure in the roof-to-
tion of the tank is expected to be the failure point of storage tanks
shell of the tank. Therefore, we attempt to establish the spatial
during normal use [3]. Under the same internal pressure, when the
finite element model of the storage tank using three kinds of stor-
roof-to-shell connection of the tank is destroyed earlier than the
age tanks that were used in the API650-2005 report [9]. Also, the
bottom-to-shell connection, the tank is called a weak roof struc-
failure criterion in the roof-to-shell connection of the tank has
ture. In the case of an overpressure failure, for two fixed roof
been established and the strength failure pressure in the roof-to-
tanks with the same specifications, in a weak roof tank, the roof
shell connection of the tank has been calculated according to the
will be separated completely or partially from the shell, and the
JB4732 standard [20]. Finally, the strength failure criterion has
tank pressure is released while the content below the shell remains
been verified by the tank failure test. This study could improve the
intact. Although the residual liquid in the tank will be on fire, the
prediction of failure modes and failure pressure in the roof-to-
risk of damage to the staff around the tank will be limited. How-
shell connection of the tank and provide more accurate theoretical
ever, a nonweak roof tank fails like a launching rocket, where the
results on the actual tank destruction process.
tank is blasted off with a flame and a storm, leading to possible
human injury or fatality. Eventually, the tank will collapse leading
to large economic loss [4,5]. Thus, it is necessary to calculate the
2 Analysis of Different Failure Modes
failure pressure and construction of the damage criteria in the
roof-to-shell of tanks. Many researchers had calculated the failure in the Roof-to-Shell Connection of the Internal
pressure in the roof-to-shell of tanks [6–8]. The API650-2005 Overpressure Vaulted Storage Tank
Two kinds of failure modes generally occur in the weak roof
tank caused by overpressure in the vaulted storage tank: compres-
Contributed by the Pressure Vessel and Piping Division of ASME for publication
in the JOURNAL OF PRESSURE VESSEL TECHNOLOGY. Manuscript received June 10, 2016;
sion ring section yield buckling failure and strength failure in the
final manuscript received January 31, 2017; published online March 10, 2017. roof-to-shell connection. As shown in Fig. 1, compression ring
Assoc. Editor: Akira Maekawa. section yield buckling failure in the roof-to-shell connection
causes localized buckling in the roof-to-shell connection, which bottom plate, angular plate, and the connection welds between
leads to tank structure buckling failure and irregular deformation. adjacent components. The extra radial dimension of the edge plate
It is very difficult to repair this damage, and the relevant mainte- out of the perimeter of the tank is 75 mm. Table 1 shows the geo-
nance cost is very high. metric parameters of the three tanks of different volumes. The
As shown in Fig. 2, when strength failure occurs in the roof-to- tanks are made of ASTM A36 steel with a yield strength of
shell connection, the position of the weld seam is split by the tank 250 MPa, a modulus of 2.07 105 MPa, and a Poisson’s ratio
internal pressure, which leads to strength failure in the roof-to- of 0.25.
shell connection. After the strength failure, the tank internal pres- The diameter-to-thickness ratio of the tanks is very large; as a
sure can be quickly released because the geometry of the damaged result, the tanks are qualified as thin-wall cylindrical shells. In the
part is more regular. It could be restored by mechanical deforma- API650-2005 research report [9], an axisymmetric model was
tion more easily and at much lower cost compared with trying to built for the tanks of exactly the same volumes (tank nos. 1, 2, and
restore from the damage caused by a compression ring section 3). Here, a similar axisymmetric model is built by ANSYS with
yield buckling failure [21]. the two-node axisymmetric element shell51, as shown in Fig. 3.
The results are compared with the API650-2005 research report
[9] to validate the effectiveness of the commercial finite element
3 Numerical Method for Calculating the Failure (FE) code.
Pressure in the Roof-to-Shell Connection To consider the weld geometry, the spatial, nonlinear FE model
is established for the entire tank structure by employing various
3.1 Establishing a Finite Element Model of the Tank. The element types available in ANSYS, as shown in Fig. 4. The model
640 m3 (tank no. 1), 3200 m3 (tank no. 2), and 6800 m3 (tank no. mainly considers the axisymmetry of the storage tank with the
3) vertical storage tanks in the API650-2005 research report [9] roof, shell, bottom plate, tank weight, and internal pressure. To
were selected to calculate the failure pressure in the roof-to-shell balance the computational effort and accuracy of the numerical
connection of the tank. The tank consisted of a roof plate, shell, model, the shell93 element and the solid45 element are used to
Tank Volume Diameter Height Down shell Up shell Bottom Roof thickness Angle
number (m3) (m) (m) thickness (mm) thickness (mm) thickness (mm) (mm) (mm)
Table 2 Various element types used for the two tank models
Fig. 5 Von Mises stress distribution for tank no. 1 using the axisymmetric model in the roof-
to-shell (loading condition 1): (a) yield of the compression ring section (criterion 1) and (b)
strength failure (criterion 3)
Fig. 6 Von Mises stress distribution for tank no. 1 using the spatial model in the roof-to-shell
(loading condition 1): (a) yield of the compression ring section (criterion 1) and (b) strength
failure (criterion 2)
Table 3 Failure pressure values of three tanks of different volumes calculated by different models and different evaluation
methods
Tank number Model Criterion 1 pressure (kPa) Criterion 2 pressure (kPa) Criterion 3 pressure (kPa)
pR 6lM E
r2h ¼ 6 2 þ 3 ð b2 M þ QÞ
t t 2b2 RD2
pR 6M (8)
r2z ¼ 6 2
2t t
r2/ ¼ 0
Fig. 8 Equivalent stress distribution at the dangerous paths
When taking the basic parameters of the tank shell and the internal (sections) in the roof-to-shell of the tanks
pressure P as known quantities, the three-direction stress at the
edge of tank shell can be obtained. connection based on the analytical solution. According to the
In practical engineering applications, the tank roof and tank strength failure criterion, the average stress in the dangerous sec-
shell are connected as a whole structure. Therefore, displacement tion will reach the yield strength at failure, which can then be sub-
and stress are continuous in the connection between the tank roof stituted into the analytical solution to determine the failure
and shell, so the edge stress for the tank roof and tank shell can be pressure. In order to verify the strength failure criterion, a small-
combined. Provided that the stress proportion at the edge of tank scale experiment of the storage tank has been conducted.
roof is k1, the proportion of tank shell will be 1 k1, and the
expression of stresses at the connection can be written as
5.1 Design of Experimental Tank Model. According to the
r1 ¼ k1 r1h þ ð1 k1 Þr2h characteristics of the tank structure, and considering the unity of
data for the experimental and the real tank structural, the experi-
r2 ¼ k1 r1z þ ð1 k1 Þr2hz (9)
mental tank structure is designed based on the principle of struc-
r3 ¼ 0 tural geometry similarity, stiffness similarity, and mass similarity.
Figure 8 shows the spatial distribution of the equivalent stress
This obtains the three-direction stress of the connection along the dangerous paths (sections) of storage tanks of three dif-
between the tank roof and shell. To determine the value of k1, a ferent volumes as well as the experimental tank. The variation
comparison analysis is carried out between the results from finite trend of the equivalent stress at the dangerous section of the roof-
element analysis and theoretical calculation. According to the to-shell of the experimental storage tanks is similar to those in the
numerical results, the coefficient after the integrated solution is actual tank. Therefore, the experimental model can be used to
k1 ¼ 0.879 [28]. indirectly verify the failure evaluation criterion in the roof-to-
Substituting the coefficient k1 into Eq. (9), we obtain shell of an actual tank.
The experimental storage tank is made of Q235, a normal car-
r01 ¼ 0:879r1h þ 0:121r2h bon constructional steel (steady-state deformation of tank material
r02 ¼ r1z r2z (10) with increasing pressure according to 5 mm/min, and the tested
dynamic yield strength of the material is about 250 MPa [29,30]),
r02 ¼ 0 the tank height is 200 mm, the tank diameter is 168 mm, the shell
thickness is 4 mm, the height of the tank roof is 28 mm, the roof
The three-direction stress at the connection between the tank roof radius curvature is 2.4R, the tank roof thickness is 4 mm, and the
and shell can be obtained by Eq. (10), which can be seen as the angle is 25 4 mm. In order to recast the two failure modes in the
stress expression of unknown internal pressure P. Through Eq. actual tank, namely, the strength failure in the tank roof-to-shell
(10), it can be seen that the failure criterion 2 is used as the and the local compression ring section yield buckling failure, the
strength evaluation criterion for the connection of the roof and
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi angle and the shell are welded in two different arrangements: lap
shell. When r1 2 þ r2 2 r1 r2 ¼ rs , strength failure occurred and butt. In lap welding, the roof-to-shell connection will see a
in the roof-to-shell, at which point, the strength failure pressure P higher buckling resistance due to larger cross section area than the
in the roof-to-shell can be obtained. Using the above failure butt welding, as shown in the inserted diagram in Fig. 9. The weld
assessment method, the failure pressure is calculated for all the foot height is 3 mm, and the tank roof starting angle is 28.8 deg. In
three tanks of different volumes; Table 4 shows the results. order to determine roof-to-shell stress, a total of 16 circumferen-
From Table 4, the strength failure pressure in the roof-to-shell tial and axial strain gauges are arranged in pairs, one on the tank
calculated by the theoretical method is slightly smaller than that roof and another on the angle distributed along the circumferential
by numerical simulation; the maximum relative error of the two is direction in 0 deg, 90 deg, 180 deg, and 270 deg, respectively. The
9.7%. This is mainly because the result of the radial stress was strain gauges on the tank roof are located 4 mm from the weld
neglected in the theoretical calculation. seam, one along the generatrix of the tank roof and the other along
the horizontal circle. The strain gauges on the angle are located
8 mm from the weld seam and oriented along the axial and hori-
5 Experimental Verification of the Strength Failure zontal directions, respectively.
Criterion in the Roof-to-Shell
The failure pressure can be determined by the corresponding 5.2 Design of the Experimental Device. The experimental
average stress of the dangerous section in the roof-to-shell setup is mainly composed of three parts, as shown in Fig. 9. The
first part is the main part of the storage tank, mainly composed of observed and recorded by pressure gauge. Figure 10 shows the
the experimental tank model and the bottom support. The bottom experimental device.
support is fixed during the experiment, and the bottom of tank is
not fixed, and therefore, it can be used to accurately simulate the 5.3 Experimental Results. Overall, eight experimental tanks
process where bottom of the tank is lift off as the pressure are tested. They are divided into two groups of four, one with the
increases. The second part is the pressurized part; in order to meet lap welding for the strength failure test in the tank roof-to-shell
the needs of gas space pressure in the failure process, we dynami- and another with the butt welding for local compression ring sec-
cally increase pressure using a reciprocating pump. The third part tion yield buckling failure. Figure 11 shows the strength failure in
is the test part: through the dynamic test system, we test the stress the tank roof-to-shell. Figure 12 shows the local compression ring
in the roof-to-shell via boosting until the tank is damaged. section yield buckling failure of the storage tank.
During the experiment, the liquid in reservoir flows through From Fig. 11, although the strength failure occurs in the roof-
reciprocating pump and then passes through the high-pressure to-shell, the circular shape of the cross section of the tank roof
hose into the experimental tank. The tank gradually deforms under and tank shell is still retained after damage. As can be seen from
high pressure. The dynamic signal testing system collects the Fig. 12, after the tank roof and shell are damaged by the yield
strain in the roof-to-shell and then converts it to a stress signal. buckling of the compression ring section, the section shape of the
The pressure is increased until the connection of roof and shell is tank changes obviously. This verifies the two damage modes types
damaged. At the same time, the tank failure pressure could be for the tank roof in Sec. 2.
Conclusion
(1) In this paper, two failure modes in the tank roof-to-shell
have been proposed: compression ring section yield buck-
ling and strength failure. The finite element model of the
tank has been established by using an axisymmetric and
spatial model. Then, a comparative analysis has been made
Fig. 12 Local compression ring section yield buckling failure on the failure modes and the corresponding failure pressure
in the roof-to-shell of the experimental tank in the roof-to-shell under internal overpressure. Evaluation
methods have also been established for both failure modes.
Table 5 shows the experimental data and finite element calcula- The finite element calculation results show that the yield
tion results for the pressure calculation method considering the buckling failure pressure of the compression ring section is
strength failure in the roof-to-shell of the tank. The failure pres- much larger than the strength failure pressure. The adopted
sure obtained by the assessment method (criterion 2) in this paper spatial model exhibits more advantages than the two-node
Table 5 Experimental results of the failure pressure of the tank compared with numerical results based on strength failure
criterion
Failure pressure
Table 6 The failure pressure values in the roof-to-shell of tanks obtained from different methods
Strength failure in the Compression ring Strength failure in The ratio of yield failure
Tank roof-to-shell connection section yield buckling the roof-to-shell pressure to strength failure
number (numerical simulation spatial model) failure (theory) connection (theory) pressure (theory)