You are on page 1of 13

Available online at www.sciencedirect.

com

Engineering Failure Analysis 16 (2009) 2638 www.elsevier.com/locate/engfailanal

Failure analysis of storage tank


imc k, J. Bocko a *, F. S a F. Trebun
ice, 042 00 Kos ice, Slovakia Department of Applied Mechanics and Mechatronics, Technical University of Kos Received 13 December 2007; accepted 27 December 2007 Available online 20 January 2008

Abstract This paper has studied failure of a hot water storage tank. During operation, one of the two tanks has been damaged by collapse of its roof. Consequently, analytical, numerical and experimental analysis of possible failure reasons was performed. Extreme stresses in the structure during its operation were determined taking into account shape and geometry imperfections as well as corrosion inuence. These values were veried numerically by the nite element method. Extensive experiments performed by strain gauge measurements on the second tank of the same design helped us determine timedependent stresses in extremely loaded locations during the chosen regimes of operation. The results of analysis allowed us to assess the failure reason and to express the recommendations for further analysis of the non-damaged tank. 2008 Elsevier Ltd. All rights reserved.
Keywords: Tank failures; Finite element analysis; Stress analysis

1. Introduction Two vertical non-pressure tanks of the same structure (approximately 30 years old) originally designed for heavy oil storage were used as storage tanks for hot water (temperature 6595 C). The tanks were reconstructed, equipped with the necessary technological devices and brought into operation (Fig. 1). After two weeks of operation the roof structure of the storage tank collapsed (Fig. 2). The aim of solution is to analyze the loading of the storage tank during its operation, to judge the reasons for its failure and to propose the corrective actions for safely operation of the second tank, which was designed for the same purposes and operation as the rst one. 2. Basic facts about the tank and its operation As mentioned above, the aim of the paper is to investigate the vertical cylindrical tank with the spherical roof and basic dimensions as given in Fig. 3:
*

Corresponding author. Tel.: +421 55 6022462; fax: +421 55 6334738. a). E-mail address: frantisek.trebuna@tuke.sk (F. Trebun

1350-6307/$ - see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.engfailanal.2007.12.005

 et al. / Engineering Failure Analysis 16 (2009) 2638 F. Trebuna

27

Fig. 1. View of the tanks.

Inside diameter Height of cylindrical part Roof height Roof radius (spherical) Thickness of cylindrical shell Thickness of roof shell Volume Material S235

D = 14 289 mm Hs = 9 208 mm h = 1100 mm R = 24 500 mm d1 = 6 to 8 mm d = 4 mm 1475 m3 for water surface height H = 9108 mm ReH min = 235 MPa, Rm min = 340 MPa

The roof of the tank consists of a steel shell with thickness of d = 4 mm, which is welded to the steel framework of the roof by intermittent chain, welds (Fig. 4). The framework of the roof is constructed from outer and internal rings, polygonal rings and radial ribs. The bottom of the tank (Fig. 5) consists of a ring plate of the thickness d2 = 16 mm to which the rst steel band cylindrical wall (shell) of the thickness d1 = 8 mm and the bottom plate of tank are welded. The operational regimes of the storage tank are the following:  Filling up the tank with hot water.  Operation of the storage tank based on regulation of water temperature by lling the tank with cold water (discharging) or with water which is hotter than that contained in the tank (charging). During these processes, the height of the water level can change at the velocity determined by the technological equipment of the tank.  Discharging of water from the tank (maintenance, unavailability time, etc.). Corresponding control equipment ensures all the above-mentioned functions.

28

 et al. / Engineering Failure Analysis 16 (2009) 2638 F. Trebuna

Fig. 2. Collapsed roof of the tank.

Fig. 3. Basic dimensions of the tank.

3. Visual inspection of the collapsed tank On the basis of the visual inspection it was found that: the steel framework of the roof was extensively deformed (Fig. 6); the welds between radial ribs [11] and outer rings were damaged (Fig. 7), supporting elements of the roof steel structure were signicantly corroded (Fig. 8); maximum corrosion depths of the proles are given in Table 1. Roof shell thickness was mostly in the range 3.43.7 mm, with the local minimum 2.83.1 mm.

a et al. / Engineering Failure Analysis 16 (2009) 2638 F. Trebun

29

Fig. 4. Steel framework of the roof. (1) radial ribs U 100, (2) polygonal rings L 70 70 8, (3) polygonal rings L 65 65 6, (4) internal ring U 140, (5) outer ring U 180.

Fig. 5. Bottom of the tank.

Fig. 6. Deformation of the roof frame.

30

 et al. / Engineering Failure Analysis 16 (2009) 2638 F. Trebuna

Fig. 7. Fractured weld between the radial rib and the outer ring.

Fig. 8. Corroded elements of the supporting structure.

Table 1 Decrease of the thicknesses of carrying elements in the roof Prole L65 x 65 x 6 L70 x 79 x 8 U 100 U 140 U 180 Thickness decrease 6.5% 1.5% 6.5% 11.5% 7.5%

4. Loading of the storage tank The storage tank is loaded during operation by:  self-weight of the structure,

a et al. / Engineering Failure Analysis 16 (2009) 2638 F. Trebun

31

 hydrostatic pressure of the contained liquid and by pressure (or under pressure) above the liquid surface in accordance with working conditions dened by operating manuals,  climatic loading (snow, wind). At this point, it is necessary to note that during operation of the storage tank the pressure of 1.5 kPa produced by steam above the liquid surface was ensured by the control system of the tank. 5. Analytical computation and numerical verication of the tank The storage tank can be considered a non-pressure tank loaded mostly by hydrostatic pressure of the contained liquid, over- or under pressure of 0.51.5 kPa, respectively. During analytical solution check calculations were carried out in accordance with the standard procedures [16]. Comparison of the computational results with the real structure shows the following anomalies in the tank:  incorrectly chosen dimensions of the ring plate at the bottom of the tank (Fig. 5) that increase negative inuences at the junction of the vertical wall and the bottom of the tank (according to experience, the recommended dimensions are d2 = 6mm, b = min. 500 mm, [3,7,12]),  considerable imperfections in the geometry of the rst steel band of the cylindrical wall (Fig. 9) mismatch in the thickness direction, considerable deviation from the vertical direction (up to 100 mm), non-constant radius of curvature,  according to empirical dependencies the stieners of the roof should consist of ve polygonal rings and the cross-sections should have bigger dimensions. Numerical procedures were applied to assess the strength of the cylindrical shell and the tank bottom taking into account the conditions of continuity between the cylindrical shell and the tank bottom. The computation methodology does not consider the elasticity of the foundation in the rst approximation, the fact that has to be considered for a real structure. The bending moment for such hard boundary conditions would induce high stress levels with a considerable degree of plastic deformation. The high level of shear force probably led the designer to the proposal of the bottom ring plate 16 mm thick in order to ensure the stress level at the bottom of the tank at 170 MPa. With such values of internal force quantities, the radial displacement of the cylindrical shell would reach

Fig. 9. Imperfections in the geometry of the rst steel band of the cylindrical wall.

32

 et al. / Engineering Failure Analysis 16 (2009) 2638 F. Trebuna

the value exceeding 5 mm. Stress in the circumferential direction would reach the value of approximately 160 MPa. Taking into consideration the elastic foundation and relations between the shell and the bottom of the tank that corresponds to sucient protrusion of the ring plate of approximately 50 mm to the outside, the bending moment decreases to 1450 Nmm mm1 and the stress is 144 MPa. Stress at the bottom part does not exceed the value of MPa units. Radial deformation of the shell in the junction with the bottom of the tank is as a rule a few tenths of a mm. Stress in the shell in its junction to the bottom of the tank is approximately 145 MPa with the equivalent stress not exceeding 135 MPa. The stresses in circumferential direction are approximately 81 MPa so that the structure can operate safely. Analysis of welds and imperfections in the welded joints clearly shows critical locations and unsuitability of some types of welds in the manufacture of storage tanks. Analysis of the eects of imperfections pointed out a highly considerable increase in the stress level. In transition parts between the elements of the roof structure and the shell there were internal force quantities determined from the conditions of deformation continuity. Shell deformations (of the roof and cylindrical wall) as well as deformations of the roof rings (internal and outer) were taken into account during computations. Based on the assessed internal force quantities it was possible to determine stresses both in the shells and rings. In this computation, apart from the self-weight of the structure and the weight of snow the pressure (or under pressure) above the liquid surface was taken into consideration. The computations conrmed that with under pressure 1.5 kPa the strength conditions for the nominal values of the geometry of the carrying elements were fullled. In combination with snow loading pressure, stresses already reached critical values. The stability limits were reached also with the nominal geometry and the above mentioned under pressure in combination with the snow. The decreased thickness of the carrying elements along with the above mentioned combination of the loading could cause plastic deformation or damage of the structure. As there were found certain serious breaches of the conception of the roof design, special attention was given to determination of the critical loading of the roof not as an isotropic shell, but as a shell stiened by radial ribs and polygonal rings [4,8,13]. Strength computation clearly documented that the critical location was between the second and third polygonal ring (Fig. 6) where the value of the critical under pressure was the lowest. Nominal dimensions of the roof structure with under pressure 1.5 kPa as well as the snow loading resulted in the loss of stability due to the failure of several radial ribs. As the character of the failure indicated that it was caused by under pressure above the surface of the liquid, the eect of the decrease of the liquid level on the pressure above the surface was investigated. It was found that the decrease of the liquid level even by 10 mm (with maximum height of the level) induced under pressure 1.5 kPa. This statement applies on the assumption that the volume above the liquid surface is without access of air. Graphical illustration of the change of the pressure Dp above the surface with the decrease of the level DH for individual volumes of air above the surface and for the initial pressure above the surface p0 = 101.5 kPa is shown in Fig. 10. The levels of stresses and deformations in the shell were veried by the nite element method. Fig. 11 gives, as an example, a eld of von Mises stresses in the junction of the roof to the cylindrical shell with under pressure 1.5 kPa. The results of the numerical analysis conrmed the conclusions of the analytical computation. 6. Experimental stress analysis of the cylindrical shell and the roof of the tank Experimental stress analysis of the cylindrical shell and of the roof of the tank was performed on the second non-damaged tank that had an identical geometry and dimensions as the original one. The methods of strain gauges were used for the measurement. Positions of extreme loadings were chosen as locations for verication the junction of the stiener ribs to the outer ring of the roof from the lower side, location in the lower part of the roof shell between the stiener ribs, outer and rst polygonal rings and locations on the cylindrical shell. In the locations of the measurement on the cylindrical shell the following imperfections could be found slight overhang of the bottom ring plate,

 et al. / Engineering Failure Analysis 16 (2009) 2638 F. Trebuna

33

Fig. 10. Pressure changes caused by decrease of the level for individual air volumes above the surface and initial pressure above the surface p0 = 101.5 kPa.

Fig. 11. Distribution of von Mises stresses with under pressure 1.5 kPa.

conical ness of the rst new steel band of the cylindrical wall, shifting between steel bands of the cylindrical wall in the thickness direction, inappropriate positioning of the bottom ring plate, etc. Application of the strain gauges to the empty structure allowed to measure deformation increments during lling of the tank according to the instructions of the operator with respect to possible operational states or maintenance of the tanks and subsequent operation of the tank after connecting tensometric strain gauges to the tensometric apparatus and subsequent balancing of bridges. The proposed procedure is based on preconditions that the regulation of operational parameters is accurate enough and the data given by operator correspond to real values. The location of the applied strain gauges and their labelling is given in Fig. 12. In Fig. 13 are shown applied insulated strain gauges in along one meridian direction in lower part of the tank. Fig. 14 gives the measurement and evaluation chain for determination of stresses from measured values of time-dependent changes of the strains.

34

 et al. / Engineering Failure Analysis 16 (2009) 2638 F. Trebuna

Fig. 12. Location and labelling of strain gauges.

After connection of the strain gauges to the measurement apparatus and setting up zero on bridges, the measurement was carried out during gradual lling of the storage tank with hot water and due to the limited capacity of the chemical water treatment plant as well as water intake this process took several tens of hours.

 et al. / Engineering Failure Analysis 16 (2009) 2638 F. Trebuna

35

Fig. 13. Strain gauges applied along the meridian direction of the bottom part of the tank.

Fig. 14. Measurement and evaluation chain.

In the rst part there was a storage tank lled up to the height 320 mm. After a longer pause the tank was lled up to the required level which was H = 9050 mm. In the further stage, measurements were carried out during the existence of the steam pressure above the water surface and during discharging of water the steam pressure above the water surface was changed. Next, the measurement of the time-dependent changes of the measured strain values during accumulation of heat in the tank was carried out. Cooling of water in the tank with cold water is an inverse process and so it was not necessary to keep a record of it. From the strains, the time-dependent values of the stress changes were determined in the supervised locations of the tank using the linear theory of elasticity. Fig. 15 shows time-dependent values of stress increments taking into account the inuence of the steam charging above the water level from 0 to 1.5 kPa. Fig. 16 shows time-dependent values of the stress

36

 et al. / Engineering Failure Analysis 16 (2009) 2638 F. Trebuna

Fig. 15. Time-dependent values of stress increments for delivery of steam.

Fig. 16. Time-dependent values of stress increments under operating regimes with both, changes of steam extraction and changes of steam delivery.

increments after balancing the measurement system and steam pressure 1.3 kPa. The steam charging was interrupted for 30 s, which with discharging of hot water at 16 t/h caused a signicant decrease of stress in the roof shell. Discharging of hot water in the amount 65 t/h caused signicant decrease of steam pressure resulting in under pressure and only the regulation system of the tank eliminated the danger of the lost stability of the roof. During discharging of water in the amount 26 t/h, steam pressure increased to 1.3 kPa. The graph also shows the stopping of steam charging during discharging of hot water in the amount 100 t/h that again resulted in the decrease of pressure and opening of the vacuum valve. On the basis of experimental analysis of stresses in the cylindrical shell and the roof of the tank during simulation of operation regimes it could be stated that:

 et al. / Engineering Failure Analysis 16 (2009) 2638 F. Trebuna

37

Charging of the tank with water of temperature 65 C and the water level below 320 mm resulted in the stress 45 MPa due to the temperature and bending moment. After temperature alignment the stress did not exceed 14 MPa. Accordingly, the stress dynamics was signicantly inuenced by the temperature. During charging of the tank and increasing of the temperature the stresses in some parts of the structure increased monotonically, but in other locations they varied in both directions and so the change was not monotonic. The last mentioned locations are in places, where there were changes in local bending moments due to the loss of stability or sudden change of the meridian line gradient (shape imperfection). The stress in the roof shell due to the hydrostatic pressure of water column at a height of 9050 mm did not exceed the value 80 MPa. The stress in the lower part of the cylindrical shell caused by bending moment was 170 MPa and circumferential stress in the junction of the new and old wall reached the value 150 MPa. Very interesting is the fact that hydrostatic pressure caused a change in the orientation of the stress increment in the location of breaking of the meridian line (the junction of the rst and second steel band of the cylindrical wall). The stresses measured during lling of the volume above the water surface by steam are shown in Fig. 15. Specied time-dependent values of stress increments clearly documented the inuence of pressure on potential non-monotony in the course of stress increments. Very interesting results of the measured stresses are shown in Fig. 16. These results show that the inuence on stresses was absolutely identical for the interruption of steam charging and for the increased discharging of hot water from the tank. Discharging at a higher level of water caused a higher level of under pressure that can be corrected only by the correct functioning of the overstress- or vacuum valve. This result can be considered to be very important and it arms the conclusion that the failure was caused by under pressure. As the limit level of pressure and under pressure according to the instructions of the producer of the control and protecting system is adjusted during its manufacture, it could be expected that their values would be kept in the full range to avoid tank failures due to exceeding these limits. Time-dependent increments of stresses showed that for the charging of the tank (charging of water with temperature higher than the temperature of discharged water without changing the amount of water in the tank) the highest value of stress increment did not exceed 60 MPa. A more detailed description of the results can be found in [9,10]. 7. Conclusions On the basis of visual inspection, diagnostics, analysis, modelling, stress computations as well as analysis of load cycles modelled during experimental simulation the following facts can be stated:  The ring plate did not have the required width 500 mm that is recommended for all storage tanks with a diameter exceeding 12.5 m.  Although tanks with a diameter up to 15 m can be designed without a stiening structure in the roof, the reviewed tank roof had such a structure. However, in this structure according to conventional rules there should have been one more polygonal ring.  New and old steel bands of the cylindrical wall do not lie on one meridian line (imperfection) and their oset is half of their thicknesses.  Faulty welds between the old and new steel bands of the cylindrical wall were also documented by specimens taken for determination of mechanical properties of the material.  The most important welds are those which connect the rst steel band of the cylindrical wall to the circular plate. Problematic welds are in the junction of the roof shell to the outer ring of the roof (corrosion and leakages), the fact that extremely inuences safe operation of the tank.  Geometry imperfections of the tank caused higher loading of the carrying elements which was also veried by experiment. Further operation of the tank is possible only in the emergency mode and requires periodical inspections and protection against corrosion, especially from the outer side of the tank.

38

 et al. / Engineering Failure Analysis 16 (2009) 2638 F. Trebuna

 Experimental measurements conrmed all the results gained by analytical and numerical analysis. It was shown that during the increased discharging of water from the tank (approximately over 50 t/h) there was an extremely high decrease of steam pressure above the water surface that could cause under pressure above the surface due to not properly functioning control system and thus result in the failure of the tank roof.

Acknowledgments This work was supported by State Project No. 2003 SP 51/028 00 09/028 09 11 as well as by Project VEGA No. 1/2187/05. References
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] Girkmann K. Fla chentragwerke. Wien: Springer-Verlag; 1956. } r die chemische Industrie. Berlin: VEB Verlag; 1955. Kantorowitsch SB. Die Festigkeit der Apparate und Maschinen fu . Brno: PC DIR Nakladatelstv tu ; 1998. upka V, Schneider P. Konstrukce apara Kr inostrojenie; 1976. Lizin VT, Pjatkin VA. Projektirovanije tenkostennych konstrukcij. Moskva: Mas Long B, Gardner B. Guide to storage tanks and equipment. London: John Wiley and Sons; 2004. Myers PE. Above ground storage tanks. New York: McGraw-Hill; 1997. v mas ety na proc nost inostrojenii. Moskva: GNTIML; 1958. Ponomarjev SD. kol.: Rasc ovy . Bratislava: UNM; 1998. ch kons trukci STN 73 1401, Navrhovanie ocel na drz hania, pr c rie a urc cii opatren . a F. A kol.: Akumulac ne za nama iny hava enie zvys kovej z Trebun e, analy ivotnosti po realiza iastkove za very z analy tora. Janua r: TU Kos zy pre realiza ice; 2005. C cii opatren . na drz hania, pr c rie a urc a F. a kol.: Akumulac ne za nama iny hava enie zvys kovej z ivotnosti po realiza Trebun e, analy verec spra va. Ju na l: TU Kos ice; 2005. Za k M. Medzne stavy kovy. Grafotlac a F, Burs a pres ov 2002. Trebun imc prvkov mechanicky k F. Odolnost a F, S a ch su stav. Kos ice: Emilena; 2004. Trebun mir AS. Ustojc uprugich system. Moskva: GIFML; 1963. ivost Vol

You might also like