PERFORMANCE OF WINE STORAGE TANKS: LESSONS FROM THE EARTHQUAKES NEAR MARLBOROUGH James Rosewitz1, Christopher Kahanek2 ABSTRACT: Stainless steel wine storage tanks suffered significant and varied damage as a result of the recent MW6.6 earthquakes near Marlborough, the largest wine producing region in New Zealand. The wine storage tanks are a critical production component at the wineries, providing for the storage of each season's stock prior to transfer and bottling. The response of the tanks to the earthquakes, and the associated damage, varied greatly depending on proximity to the epicentres, volume of product within the tank, and the tank construction. This paper discusses observations from damage assessment of the wine tanks at 20 wineries throughout the Marlborough region. Different types of observed damage are discussed along with an evaluation of the causes and recommendations for future design improvements. 4,000 stainless steel tanks, to assess damage on behalf of insurers. While seismic damage to buildings at the wineries was minor or non-existent, almost half of the wine tanks had visible deformations or other forms of minor distress and more than 70 tanks ruptured, resulting in loss of contents. Many of the catwalks directly connected to the tanks also suffered damage. As of April 2014, damage investigations continue at several of the wineries. The intent of this paper is to provide a timely summary of the extent, magnitude, and types of damage to the observed wine tanks in Marlborough. Much can be learned from the damage caused by the recent seismic events because they were large enough to cause deformations indicating the early modes of failure, but small enough to avoid catastrophic failures that complicate or prevent forensic investigation. Conclusions in this paper are preliminary in nature, provided as data for the on-going discourse about the design and repair of wine tanks in seismic regions. 2 EARTHQUAKE OVERVIEWS A series of significant earthquake events occurred in the Cook Strait in July and August of 2013. The Cook Strait separates the North and South Islands of New Zealand, and the Marlborough region is located near its southwest shore on the South Island. The sequence began on 19 July 2013 (NZT) with a MW5.5 foreshock 30 km east of Seddon, New Zealand. This was followed by two significant events, the Cook Strait Earthquake on 21 July and the Lake Grasmere Earthquake on 16 August 2013. KEYWORDS: Stainless steel liquid storage tanks, wine tanks, Lake Grassmere Earthquake, seismic design 1 INTRODUCTION New Zealand wine exports were valued at $1.2 Billion (NZD) in 2012, and the industry continues to grow. Marlborough is the largest wine producing region in New Zealand, producing approximately 70% of the industry’s wine by weight. Furthermore, wine production is approximately 20% of the Marlborough region’s economy [1]. The industry is relatively new to the region; most of the exponential growth and associated construction has occurred in the last two decades. The Marlborough region experienced seismic events on 21 July 2013 (MW6.6 Cook Strait Earthquake) and 26 August 2013 (MW6.6 Lake Grassmere Earthquake). Subsequent to the events, the authors visited over 20 wineries in the region, which collectively own more than 1 James Rosewitz, P.E. (USA), Project Engineer - Thornton Tomasetti, Christchurch, NZ Email: [email protected] Figure 1: Earthquake epicentres and Marlborough region seismograph locations (in yellow) The Cook Strait Earthquake measured 6.6 on the moment magnitude scale (MW) and was centred 20 km east of Seddon. The earthquake struck at 5:09 pm NZT on 21 July 2013 at a depth of 13 km. The Lake Grassmere Earthquake measured MW6.6 and was centred 10 km southeast of Seddon. It occurred at 2:31 pm NZT on 16 August 2013 at a depth of 8 km. The earthquakes occurred on a previously unknown offshore extension of the London Hills Fault. The epicentres of these events are shown in Figure 1. 2 Christopher Kahanek, CPEng, IntPE(NZ), MIPENZ, Senior Associate - Thornton Tomasetti, Christchurch, NZ Email: [email protected] Damage in the Marlborough region from both of these events was limited. Notable damage occurred in the region surrounding Seddon, including partial collapses Figure 2: Leg Supported Tank Diagram and land damage. Blenheim experienced less severe ground motions and less damage. Commercial and residential structures in Blenheim generally performed well [2]. 3 MARLBOROUGH WINERIES The wineries in Marlborough vary in configuration, but in general they are light industrial campuses ranging in size from 1 to 10 hectares located on relatively level land. They consist of a number of low-rise buildings and rows of wine tanks commonly referred to as “tank farms.” The buildings accommodate a variety of uses including wine tasting rooms, restaurants, administrative offices, equipment storage, and wine barrel storage. Wine tanks are generally located outside, but in some cases, tanks are inside buildings to provide better temperature control. The tank farms are usually arranged in parallel rows with similarly sized tanks placed adjacent to each other. Since the tanks need to be accessed from above, catwalks are usually installed in a double-loaded corridor fashion to permit efficient access to the top of every tank. Piping, usually containing water and refrigerants, also runs throughout the tank farms, and is often installed directly under the catwalks. The tank farms, both indoor and outdoor, are paved with a continuous concrete slab sloped to drain to underground stormwater pipes. The site plan and functional layout of a winery can evolve over time. Wine tanks are portable structures, and tank farms are often rearranged to meet operational needs. Some of the large wineries have multiple locations, and they occasionally move tanks and large quantities of wine between sites. 3.1 Although stainless steel has been used to construct wine storage tanks for more than half a century, tank design procedures and construction methods can vary widely by manufacturer and region. The basic components of a tank are the cone, barrel (or shell), base, skirt, and support (either plinth or leg supported). The components are welded together with either a full penetration butt weld or a fillet weld, depending on the tank’s configuration. Different thicknesses of stainless steel are used for different components. Figure 3: Plinth Supported Tank Diagram Tank supports in the Marlborough region typically consist of either steel legs that support a steel framework (Figure 2), or a concrete “plinth” that directly supports the base of the tank (Figure 3). Plinths are used to support tanks of all sizes. Leg supports are commonly used for the tanks that hold 60,000 L or less. Although it is possible, tanks larger than 60,000 L on legs were not observed. WINE STORAGE TANK CONSTRUCTION Wine storage tanks in Marlborough are constructed typically of Type 304 stainless steel sheet metal. Stainless steel is ideal for storing wine because hygienic and does not corrode when correctly maintained. The smallest tanks observed hold 1,200 L, and the largest tanks hold up to 580,000 L. They can be as tall as 20 m and up to 10 m in diameter, although 6 m is a common maximum diameter due to road transportation restrictions. 3.1.1 Cone The tank cone is the top cover of the tank barrel. It is shaped to allow for exterior drainage and precise filling of the tank. It is usually 2.5 - 3 mm thick. Tank cones in the region are either asymmetric with the turret offset to one side as shown in Figure 2, or symmetric with the turret at the centre. 3.1.2 Barrel (Shell) The tank barrel is constructed with multiple “strakes” of sheet metal, up to 1,500 mm wide, butt welded together either with a continuous helical or parallel horizontal pattern. Barrel thicknesses vary based on the tank size. The bottom strake in a larger tank can be up to 5 mm thick and decrease up the height of the tank to 2 mm thick. Smaller tanks (45,000 L or less) commonly have a uniform barrel thickness of 2 mm. The barrels are typically surrounded by refrigerant lines. Older tanks are surrounded by a refrigerant layer created by a continuous, often dimpled, second skin of sheet metal. Newer tanks are surrounded by a helix of stainless steel parallel flange channels welded continuously to the side of the barrel walls (Figure 2). The barrels are also often surrounded by a layer of insulation (polystyrene) and a “second skin” of stainless steel. These tanks can be identified because the refrigerant lines are hidden behind the insulation layer. However, refrigerant lines are apparent from the tank interior. drainage. The plinth is typically partially visible under the tank skirt. In most cases, the plinths are not designed by the tank manufacturers, but instead are regarded as site work separate from the tanks. The plinth diameter is usually 50 mm less than the diameter of the tank to accommodate placement tolerances. A gap of varying width exists between the bottom of the skirt and the ground slab. Materials such as polystyrene or asphalt are usually sandwiched between the plinth and the tank base. 3.1.6 Legs and Frame Leg frame structures have a variety of configurations. The horizontal frame that supports the tank is composed of square or rectangular hollow tube sections arranged as spokes with thinner rectangular hollow tube sections around the exterior perimeter. The frame supports a polystyrene or rigid chipboard layer, upon which the tank is placed. Cylindrical legs, typically made of structural pipe, are welded to the frame at the centre and ends of the spokes. Newer leg-supported tanks have a centre leg for each spoke, but older tanks commonly only have a single leg at the spoke intersection. The legs have shims or telescoping bases to adjust to the sloped floors. A series of welded tabs connect the perimeter tube of the frame to the bottom strake of the barrel (Figure 4). 3.1.3 Base The typical tank base consists of rows of sheet metal laid flat and sloping toward the drain. Base sheet metal thickness is uniform and typically varies between 2 mm and 3 mm based on tank size. Many bases have a downward indention at the drain described as a “gullet” to allow for complete removal of the contents. The base is typically connected to the barrel wall by cold rolling its edge upward and then either butt welding or fillet welding it to the bottom strake of the barrel. This important detail, known as the “knuckle,” has a constant radius that varies based on the thickness of the base sheet metal. 3.1.4 Skirt Most tanks have a stainless steel “skirt” that creates a level ring around the base. Its top edge is fillet welded to the side of the bottom strake of the barrel. The skirt can be used as a structural component where anchor rod brackets are attached (Figure 3), or as a non-structural component used to cover framing under the tank (Figure 2). When the skirt is structural, it can be up to 6 mm thick, and is seldom thicker than 1.5 mm thick when non-structural. 3.1.5 Plinth Concrete plinths are used to directly support the base of larger tanks (capacity larger than 60,000 L). They are constructed with typically unreinforced concrete cast directly on top of the slab-on-grade. In some cases, the plinths are tied into the slab below. They are sloped to accommodate the slope of the tank base to allow Figure 4: Typical welded tab connection Most leg support structures consist of relatively short legs welded to a tube steel frame sloped to match the slope of the tank base (Figure 2). Taller leg support structures, which can be up to 1 m tall, typically have cross-bracing between the legs. 3.1.7 Anchor Rods Most plinth-supported tanks are connected to the slabon-grade or their foundations with anchor rods. All observed anchors were post-installed with epoxy. In some cases, necked anchor rods are used to induce ductile behaviour in a controlled region. The anchor rods are engaged by anchor chairs welded to the tank skirt. Anchor rods are typically designed to resist uplift (tension) forces only, but in some cases also resist compressive forces. 3.2 CATWALKS The catwalks are typically constructed around the tanks or after they are in place. They are lightweight galvanised or painted steel structures meant to support light foot traffic. They are entirely self-supporting at some wineries, and at others they rely partially or entirely on the tanks for support. The tank-supported catwalks use a variety of details to allow for differential movement. Some horizontal plate connections allow for sliding movement, whereas certain tab connections allow for essentially no differential movement. Figure 2 shows a condition where the catwalk is connected to the sides of the tanks with tabs that allow very little horizontal movement. 4 HISTORY OF TANK DESIGN GUIDANCE IN NEW ZEALAND Significant damage to storage tanks was documented in earthquakes as early as the 1964 Alaska earthquake [3]. The seismic behaviour of tanks has been the subject of much research in the last five decades. Tanks observed in the Marlborough region were constructed as early as the 1960s, and as recently as 2013. The most common tanks were constructed between 2001 and 2013. Definitive standards for the seismic design of wine storage tanks in New Zealand (i.e., directly referenced by the New Zealand Building Code) do not exist [4]. The seismic loading standard for structures (NZS1170.5) specifically excludes liquid storage tanks from its scope [5]. Different guidelines have been used for tank design in New Zealand over the last several decades, and all of these standards were likely used for tanks in the region [4]. A brief history of pertinent tank design guidance documents that were likely used for design in the region is presented below. 4.1 THEORY OF TANK SEISMIC LOADING The basis for seismic design of liquid storage tanks is the Housner multi-component spring/mass analogy [6]. The Housner procedure separates the lateral forces produced by seismic shaking into impulsive and convective forces on the tank walls. Impulsive forces include the inertial movement of the liquid, and it is primarily a function of tank wall stiffness. Convective forces are a function of fluid hydrodynamic movement (i.e., sloshing). The natural periods for these components vary. Impulsive periods for common wine storage tank geometries are generally around 0.4 seconds or less, and convective periods may be as long as several seconds. When tanks are full and sealed, the convective mode is minimized due to the liquid being confined, and the impulsive mode governs response [2]. 4.2 API 650 – APPENDIX E The American Petroleum Institute (API) first published a specification for welded steel storage tanks in July 1936 entitled API Standard 12C, Welded Oil Storage Tanks. This specification was replaced by API 650, Welded Steel Tanks for Oil Storage in 1961. API 650 is currently in its Eleventh Edition. Seismic loading was introduced to API 650 in the early 1980s with the addition of Appendix E – Seismic Design of Storage Tanks. API650-E utilizes the Housner method with modifications by Wozniak and Mitchell [7] that account for tank flexibility [3]. This document was used to design petroleum storage tanks in New Zealand; however, its seismicity recommendations were for the United States, and its direct application to New Zealand was limited [6]. 4.3 NZSEE “RED BOOK” (1986) NZSEE first published Recommendations for Seismic Design of Storage Tanks in 1986 (the “Red Book”) [8]. Since its publication, the document has been widely used for storage tank design in New Zealand, and was likely used in the design of 70 – 80% of the tanks in the region. The design results from this document tend to be more conservative than API 650-E [9]. The performance criteria for this guidelines states that a tank should be serviceable following a design earthquake (i.e. able to store and move contents). The tank should also be designed with a hierarchy of failure which protects the contents in events larger than the design earthquake. This guideline generally does not allow yielding of tank elements, or a reduction of design forces for ductility [9], except for elevated tanks. Design details are mentioned briefly only to state that they must comply with the performance objectives, but no specific guidance is given. 4.4 STANDARD SEISMIC RESISTANT DETAILS FOR INDUSTRIAL TANKS AND SILOS (1990) The New Zealand Earthquake Research Foundation was commissioned by the Earthquake and War Damage Commission (EQC) to investigate and provide standard details for seismic resistance for industrial tanks and silos [4]. The report reviewed relevant guidelines for tank seismic design including the Red Book and API 650. Recommendations were made for several aspects of tank design, including appropriate ductility factors for different anchorage types and details to resist seismic overturning. Ductility recommendations were based on the allowance for ductility for elevated tanks in the Red Book and only applied to the anchorage, not the tank itself. The details provided by this document were commonly observed throughout the region, even in the most recent tank designs. 4.5 NZSEE “BLUE BOOK” (2009) The NZSEE guidelines were updated in 2009 (the “Blue Book”) [10], however the same general performance criteria were maintained from the Red Book. The guidelines were updated to be consistent with the current seismic design philosophy used by the seismic loading standard NZS1170.5. This included implementation of updated local seismicity, soil conditions, seismic risk factors, and damping. The update also allows for the limited use of ductility for design of the tank components, as well as hold down anchors [4]. Approximately 10-15% of the tanks observed in the region were designed using this guideline. 5 SHAKING MAGNITUDE/SEISMIC INTENSITY AT THE INVESTIGATION SITES The Cook Strait and Lake Grassmere earthquakes were both strongly felt in the Marlborough region. Measured peak ground accelerations (PGAs) exceeded 0.2g in the Cook Strait Earthquake, and 0.6g in the Lake Grassmere Earthquake. Table 1 presents an estimated breakdown of PGAs experienced at the sites that were visited. PGA calculations at each site are based on data provided by GNS Science (www.geonet.org.nz) and linearly interpolated between recording stations. Table 1: Estimated Peak Ground Accelerations at Marlborough sites 21 July 2013 PGA (g) Quantity < 0.10 18 0.10 – 0.20 0 0.20 – 0.30 2 0.30 – 0.40 0 > 0.40 0 26 August 2013 PGA (g) Quantity < 0.10 7 0.10 – 0.20 4 0.20 – 0.30 7 0.30 – 0.40 0 > 0.40 2 Wine tanks are most often either full or empty. When they are full, the convective modes with longer periods typically do not participate in their response, and their response is governed by the impulsive mode. Figure 5 shows the spectral response from different seismographs (also obtained through GNS Science) in the Marlborough region, with the range of expected tank impulsive mode periods shaded in grey. The seismograph locations are shown in Figure 1. Note that RCS1 and RCS2 were installed after the Cook Strait Earthquake, and only recordings from the Lake Grassmere Earthquake are available for these two seismographs. Two design elastic response spectra from the NZSEE Blue Book for Blenheim (Z = 0.33, 5% damping) with two different importance levels (IL) are also shown in Figure 5. The solid line is the response spectrum for wine tanks (IL1 and a 50-year design life correlating with a return period of 1 in 100 years). The dashed line is the response spectrum for a typical building (IL2 and a 50-year design life correlating with a return period of 1 in 500 years). The Cook Strait Earthquake was relatively minor with the exception of one recording station near the epicentre that exceeded the wine tank (100-year) design spectrum. The earthquake did not exceed the 500-year design spectrum. This is consistent with reported damage from this earthquake. Figure 5: Spectral Accelerations at Various Seismographs vs. Typical Design Spectrum The Lake Grassmere Earthquake was more significant. Two recording stations (RCS2 and WDFS) for the short period range exceeded both the 100- and 500-year design spectrums. These recording stations were closest to the epicentre in the nearby towns of Seddon and Ward. Only a few of the observed wineries were located near these stations. Stations farther from the epicentre recorded similar spectral accelerations to the Cook Strait Earthquake. MGCS was closest to several wineries, and was well below both design spectra in the impulsive period range. Comparing the recorded ground motions to the design spectra, both events were relatively minor (well below the design spectra) for the majority of the observed sites. The event was significant at a limited number of sites near the Lake Grassmere Earthquake epicentre. These results are dependent on an assumption of similar soil types, and the results may vary with more detailed investigations. 6 EVALUATION OF OBSERVATIONS The authors observed wine tanks at 20 sites throughout the Marlborough region. The first site visit occurred between the Cook Strait and Lake Grassmere Earthquakes. Additional site visits began within two days of the Lake Grassmere Earthquake and continue through the first half of 2014. At some sites, observation was limited to a thorough walk-through to provide a high-level estimate of damage. To date, 1900 tanks have been individually assessed for damage. Of these, 1200 were investigated in detail at specific sites including component measurements, exact quantification of damage, and entering the tank for interior inspections. Percentages discussed in the observations below are based on the 1200 tanks observed in detail, which consist of approximately 500 plinth-supported and 700 legsupported tanks. Most of the observed earthquake-related damage to the tanks can be organized into typical patterns. The most convenient way to separate the tanks is by their support system (i.e. plinth or leg supported). Tanks with differing support conditions may have similar patterns of damage, but their response and hierarchy of failure (components that failed prior to or instead of other components) varies significantly. The observations described in this section are the most current understanding of the conditions. Precise quantification of the patterns of damage (i.e., exact numbers of tanks that experienced different types of damage) is in progress as of April 2014. 6.1 Shell buckling was not common at most sites, but both types were observed at tanks closer to the epicentre. In total, only 35 tanks experienced some form of shell buckling. Shell buckling was observed most commonly in the bottom strake of the tank, but in several cases elephant foot buckling was observed in the second and third strakes between glycol channels (Figure 7), corresponding with a change in the steel sheet thickness. This behaviour was limited to a specific group of 100,000 L tanks at one site. PLINTH-SUPPORTED TANKS 6.1.1 Barrel and Cone Cone creasing is a common form of damage caused by suction within the tank, usually by displaced contents, or sloshing of the tank contents (Figure 6A). In extreme cases, tanks collapse inward as contents are rapidly emptied after a rupture in the tank. Collapses due to suction did not occur during the recent events; however, cone creasing was observed in approximately 30% of the plinth-supported tanks. Cone creasing can occur during the tank’s normal service life when contents are emptied without opening the turret causing internal suction. Determining whether cone damage was caused in this manner or during an earthquake is difficult. Cone creasing related to earthquakes typically occurs in conjuncture with other significant damage or in cases when the tank was partially full and sloshing occurred. Figure 7: Observed Elephant Foot Buckling Between Refrigerant Channels 6.1.2 Barrel-to-base knuckle connection Downward deformation of the knuckle connection between the edge of the plinth and the barrel wall was a common condition observed throughout the region, occurring at about 55% of all plinth-supported tanks (Figure 8). The deformation occurred when the tank barrel was not directly supported by the ground via the skirt, and barrel wall compression forces were transferred to the plinth support through the knuckle. As described previously, the plinth diameter is typically up to 50 mm smaller than the barrel diameter, and the overhang had high bending stresses that resulted in the permanent deformation. When viewed from the exterior only, the deformed knuckle creates a condition that could be misinterpreted as stretched anchor rods or a plinth that has settled relative to the anchor rods (Figure 9B). Figure 6: Typical plinth-supported tank cone and barrel damage Damage to the tank barrels includes shell buckling, indentations related to pounding, and deformations at catwalk tab connections (Figure 6B). Shell buckling includes both diamond buckling and elephant foot buckling. Diamond buckling is membrane compression buckling (Figure 6C), and elephant foot buckling is elastic-plastic collapse of the tank wall (Figure 6D). Deformation of the knuckle can change the response of the tank in several ways during an earthquake. (1) Rotation of the knuckle can cause the corner of the plinth to spall, lengthening the cantilever of the base, thereby exacerbating the condition. (2) The anchor rods meant to resist tension forces are disengaged as the tank settles downward causing the tank to rock, and consequently drift further than anticipated. (3) In extreme cases, the knuckle may deform enough to allow the skirt to rest on the ground increasing compressive stresses in the barrel wall. Based on our observations, this increases the potential for shell buckling. (4) When the tank is full, the settlement can reduce the capacity of the tank and cause the contents to push on the cone and swell upwards. deformation, then they often experienced less anchorage damage because the anchors were no longer engaged. Anchor rod failures observed included rupture (Figure 9A), concrete pull-out (Figure 9B), epoxy pull-out, anchor rod buckling (Figure 9C), and anchor rod thread shearing. In some instances, anchor bolts were installed horizontally into the concrete plinth. Anchor bolts tended to shear off when they were installed in this configuration (Figure 9D). Figure 9: Typical anchorage failures 6.2 Figure 8: Knuckle Deformation in a Plinth Supported Tank Leaks caused solely by knuckle deformation were rare in spite of its pervasiveness. Only one site had tanks with leaks caused by this condition, and fewer than 5% of the tanks with knuckle deformation had leaks. No leaks resulted from cone swelling. 6.1.3 Skirt Damage isolated to the skirt was relatively rare for plinth-supported tanks, but in some cases, elephant foot buckling of the tank barrel also caused the skirt to buckle. This occurred rarely, likely because the skirt increased buckling resistance when acting with the barrel wall. As a result, in many cases, elephant foot buckling occurred in the first strake immediately above the skirt. Skirt buckling also occurred in cases where the tank anchorage was detailed to resist compressive forces. In most cases where the anchorage resisted compression, the anchor would buckle prior to the skirt. However, in some rare cases involving retrofit anchorage on older tanks, the anchor bolts were encased in concrete, causing them to be significantly stiffer and resulting in failure of the skirt. 6.1.4 Anchorage A wide variety of anchorage configurations and failures were observed throughout the region. Anchor rod failure was common, occurring at approximately 40% of all plinth-supported tanks and up to 50% at some specific sites. If plinth-supported tanks settled due to knuckle LEG-SUPPORTED TANKS 6.2.1 Barrel and Cone Barrel and cone damage was rare for leg-supported tanks in the region. Damage to the barrel was typically limited by the capacity of the legs, which failed first. These tanks also have a smaller height-to-diameter ratio than plinth-supported tanks, and therefore they typically had lower overturning demands relative to the plinthsupported tanks. Less than 1% of the leg-supported tanks observed in the region sustained cone or barrel damage that was not related to an adjacent tank or structure colliding with the tank. One exception was isolated elephant foot buckling that occurred directly over the legs at the bottom section of the barrel in specific groups of tanks. The observed buckling was different from that in the plinth-supported tanks because it was in a small concentrated region rather than extending around a large section of the tank circumference. These failures occurred due to a concentration of compression force through the tank’s barrel directly over the legs. This may have been coupled with deformation of the tab connecting the tank barrel to the leg framing. This mode of failure is discussed further in Section 6.2.3. 6.2.2 Legs and Frames Approximately 30% of leg-supported tanks had damage to the legs or frame. Damage was more common to tanks 30,000 L and larger. Damage to the legs included buckling of the main leg section (Figure 10A), buckling of the telescopic lower leg section (Figure 10B), and tilting of the entire leg (Figure 10C). The hollow tube frame section attached to the leg often deformed inward when the leg tilted (Figure 10D). Damage to the tab connections is not obvious when inspecting a tank because they are usually hidden by the skirt. The tank skirt must be inspected for subtle evidence in the form of a square-shaped punched section or a small localised buckle (Figure 11D). Since the underside of the tank base cannot be observed because it is covered by insulation or fibreboard, the tank must be emptied to assess the extent of base deformation and rupture. The skirt must be removed in order to directly view and repair the tabs. There is an inverse correlation between the tab connection failures and the leg failures. Tanks with damaged legs are less likely to have tab damage. This suggests that either the tab connection or the legs act as an unintentional fuse that dissipates energy. Figure 10: Typical leg-supported tank leg damage Leg-supported tanks were both anchored and unanchored throughout the region. Most sites were either entirely anchored or unanchored. No correlation between the amount of observed damage and tank anchorage was immediately obvious. However, a site closer to the epicentre with all tanks anchored experienced less leg damage than other sites that had lower accelerations. This may be due to a number of factors other than anchorage, including tank age and dimensions. Red wine tanks are different (and often smaller) than the typical leg-supported tanks in the region, but they are also higher off the ground. They have taller legs that are typically diagonally braced by tube sections. Leg damage was less common in red wine tanks, but in many cases the diagonal brace buckled, and in rare cases connections between the brace and the leg ruptured. 6.2.3 Frame-to-Barrel Tabs Damage related to the steel tab that connects the tank barrel and the frame occurred at approximately 40% of observed leg-supported tanks in the region. Damage associated with the tab includes rupture of the tab (Figure 11A), buckling of the tab, which tends to fold under the tank bottom, and subsequent punching of the tab into the base of the tank (Figure 11B). Localised elephant foot buckling, as described previously, appears to be triggered by the tab deforming or punching the tank at the knuckle (Figure 11C). Significant damage (large indents in the tank base) was most common for tanks 30,000 L and larger, although less severe damage was common in smaller tanks. In several instances, tank bases were ruptured by the tab pulling away due to overturning forces. Rupture was most common when only the tip of the tab was welded to the tank base, or when the tab was welded to the knuckle. This mechanism was the most common cause of leaks in the region for both earthquakes. Many leaks were not immediately noticed because sediment at the tank bottom plugged the rupture point. These leaks were not discovered until after the tank had been cleaned, and in some cases refilled with liquid. Figure 11: Leg-supported tank tab related damage 6.2.4 Leg Anchors Many large sites do not anchor the leg-supported tanks. In locations where they were installed, the leg anchors had minimal to no damage. Leg anchor failure was rare because failure of other components such as the legs and the tab connections limited the force transferred to the anchors. This contrasts with the widespread anchor bolt damage at the plinth tanks. 6.3 TRANSPLANTED TANKS Wine storage tanks are unique because they can be transported from one region to another relatively easily. A number of observed tanks were imported from other regions in New Zealand, and some were previously affected by the 2007 Gisborne earthquake. The tanks transplanted from other regions had relatively more significant damage, including diamond buckling and severe skirt deformation. 6.4 TANK FARM LAYOUT Tank farms are arranged to use space as efficiently as possible. The possibility of tanks and catwalks pounding is likely given minimal consideration when designing the site’s layout. Tank pounding was observed at only one site, and there the damage occurred at approximately 20% of the plinth-supported tanks. Although pounding caused dents in the tank walls and cones, it did not result in any leaks. Damage at the pipe-to-tank connections was reported, but it was relatively easy to repair. Many repairs were completed within days. Damage was minimised in most cases by the typical flexible pipe connections used in the region. 6.5 that allowed horizontal movement on one side with a rigid connection similar to the ones describe above (Figure 12C). Another example is catwalk connections with bearing pads on both sides, allowing the catwalk to slide on the supports as the tanks move during an earthquake (Figure 12D). CATWALKS The catwalks have a wide variety of support details, which had varied degrees of performance. Complete loss of support was rare, only occurring at two locations. Collapse of entire catwalks was avoided in some cases due to inadvertent support provided by the piping and the handrails wrapping around the tank turrets. 6.5.1 Self-supported catwalks Self-supported catwalks generally performed very well with no known instances of damage or collapse. In some cases, the configuration allowed pounding to occur between the tank turret and the catwalk. This caused minor localised damage at most. 6.5.2 Tank-supported catwalks Catwalks supported by adjacent tanks experienced a significant amount of damage throughout the region. The connections between the catwalks and tanks were simple side-mounted vertical plates that allowed minimal vertical and horizontal movement in many cases (Figure 12A). They were unable to accommodate the differential movement between the two tanks they connected and frequently ruptured (Figure 12B). This may have been exacerbated by several factors. (1) Similar tanks may have had different content volumes during the earthquakes, resulting in substantially different movement. (2) Different anchorage conditions (especially at larger leg-supported tanks) may have caused excessive movement of the tanks. (3) Tank failures (e.g., knuckle deformation, tab ruptures) changed the expected tank response during the earthquake, causing additional tank movement. 7 DISCUSSION The Cook Strait and Lake Grassmere Earthquakes both caused damage to wine storage tanks in the Marlborough region. For a limited number of sites, they were both notable events, exceeding the typical design response spectrum for wine storage tanks. For most of the region, however, both events were below the design forces from the Blue Book. Damage was generally worse at sites closer to the epicentres; however, in some cases, damage was more severe than expected due to specific tank details. The following discussion interprets the available information to date, and offers considerations to improve tank design in the region. 7.1 THE HIERARCHY OF FAILURE AND REPAIRS Both the NZSEE Red and Blue Books state that a design hierarchy of failure should be established to minimise damage and content loss in a severe earthquake. The guides however do not attempt to minimise the repairs that are required due to the anticipated damage that is established by the hierarchy of failure. Repairs to the less critical elements of wine tanks can be costly and time consuming. The repair for shell buckling, for instance, requires either local or full removal of the strake. Tank downtime can greatly impact a winery, especially if it is near vintage when the winery expects to require its full capacity. The hierarchy of failure in design should account for the ease of repair to minimise potential downtime and loss of business. This is already done in buildings where easy to replace “fuse” elements are installed in lateral force resisting systems. Easy to repair wine tank components that could be used as a fuse include the anchorage and skirt. 7.2 TANK DETAILS The most common damage observed was not the result of overturning or an anticipated yielding mechanism, but instead the result of several key details. Their poor performance resulted in deformations and ruptures that had a disproportionate impact on the functionality and repair of the tanks. The key details are discussed below. Figure 12: Typical catwalk conditions and damage Other connections observed around the region designed to allow differential movement between the tanks performed well. One example is catwalk connections 7.2.1 Non-Uniform Plinth Support The typical plinth-supported tank detail with a gap between the plinth and the skirt, as well as the skirt not directly resting on the ground, causes a condition where a varied length of the tank base is unsupported. The unsupported length is typically up to 50 mm, but unsupported lengths over 100 mm were observed in specific cases. Compressive overturning forces must be transferred through the knuckle into the plinth, causing downward deformation of the unsupported portion. The deformation tends to prevent proper drainage of the tank. Although knuckle deformation rarely led to content leaks, it is difficult to repair, and often requires replacing the tank base. This deformation also modifies the tank behaviour. Future designs of plinth supports should include a welldefined load-path for compressive forces that avoids unsupported sections of the tank base. This can be accomplished in several ways. (1) The gap could be reduced with tighter construction tolerances between the tank and plinth. The base should be designed for the remaining gap to resist the demands without significant deformation, and packing could be used to prevent the tank from shifting and increasing the gap on one side. (2) The anchorage could be designed to resist compressive forces, most likely by extending the anchor chair to the ground. Care should be taken to prevent failure of the skirt due to the concentrated compressive stresses. (3) The skirt could be placed directly on the ground. One method to achieve this is by relying completely on the skirt without a plinth, similar to tanks in the dairy industry. Alternatively, newer tanks in the region are accomplishing this by placing the tanks on the ground without a pre-poured plinth and pumping concrete into the void under the tank. These tanks generally performed well in the recent earthquakes. 7.2.2 Leg-Supported Tank Tabs Specific guidance for design and detailing tanks supported by frame structures is not available, and not provided in any of the mentioned design guides. The frame structure should, however, be considered part of the tank, and should be designed to meet or exceed the performance criteria of the tank. Failure of legs, the frame, and the tab connection of the frame to the tank have all caused significant damage and a disproportionate amount of content loss. Damage to the tab, however, was the most prevalent (and avoidable) in the region. These elements should be protected by a hierarchy of failure. The typical leg-supported tank detail utilises a steel tab to connect the frame to the tank. The unsupported length of the tab between the frame and the tank impacted the performance of the connection. When the tab was relatively long, it tended to buckle or fracture causing minimal damage to the tank itself. When the tab was short and consequently stiffer, it could fold under the tank and punch into the base. The exact location of the weld connecting the tab to the tank also impacted performance. Tabs performed especially poorly when they were only welded at the end of the tab (rather than at the top and along the sides), and when they were welded directly to the knuckle. Tabs welded to the knuckle also resulted in the most tank ruptures and loss of contents in both events. Tabs welded to the side of the tank barrel performed better, but were also prone to buckling or rupture. The tab connection is not inherently flawed, but care must be taken to protect the tank from tab deformation. Tab ruptures were common and could be avoided with wider or longer steel plates. The weld length should be designed to ensure the tab does not rupture the tank. Longer tabs that cover more of the tank circumference may also be beneficial to avoid concentrating loads. The use of capacity design to identify a desirable failure point and appropriate overstrength factors to protect the tab is recommended for all alternative designs. 7.2.3 Frame Leg Connections Legs are welded to the steel hollow tube sections that compose the frame. In some configurations, the width of the leg is less than the width of the hollow tube section, and the leg is not aligned with the wall of the frame section. As a result, during an earthquake, the leg punches into the flange of the section (Figure 10D). This damage was observed throughout the region. This condition is easily avoided by either matching the diameter of the leg to the width of the tube section, or by reinforcing the tube wall where the leg connects with an additional steel plate. 7.2.4 Glycol Channels Elephant foot buckling at the second strake of certain 100,000 L tanks was observed. The elephant foot buckling began near the horizontal strake joint, but followed above the glycol spiral. The design guide assumes a uniform tank wall, and does not account for variations in stiffness that could change the tank’s behaviour. In the case of the 100,000 L tanks, which buckled on the second strake, the increased stiffness from the glycol lines may have altered the tank’s response in a manner not predicted by the design guide, thus causing buckling to occur at a different location. We recommend further assessment into the influence of stiffness changes caused by ostensible non-structural tank components (such as the glycol lines and skirts on leg-supported tanks) to ensure they are not invalidating guideline assumptions and altering the tank’s response. 7.3 CATWALK DETAILS Tank-supported catwalk connections often failed when the catwalk was rigidly connected to the tanks at both sides. This detail does not allow for relative movement of the tanks, which can happen for a variety of reasons. Damage to the catwalk connections is significant because catwalk collapse can damage the tanks, pipes, and is a threat to life safety. Catwalk connection details that allow some horizontal movement performed better than rigid connections in the region. A slip one-sided connection can appropriately allow for movement between tanks. The amount of movement may be unrealistic when a rocking mechanism is anticipated. Horizontal seating pads (similar to precast stair details) that allow the tank to slide in both directions during an event may be more appropriate depending on the design assumptions. Notably, catwalks that are self-supported performed very well in the earthquakes. Where practical, the best option is to provide self-supported catwalks that are not subject to the relative movement of adjacent tanks. 7.4 TANK DESIGN CRITERIA The NZSEE Blue Book guidelines align tank design demands with the New Zealand seismic loading standard, NZS1170.5. Importance level (IL) determines the return period factor (R) in NZS1170.5 for code-level earthquake loading. Based on conversations with manufacturers, wine tanks in the region are typically designed to IL1 with a 50-year design life, or a 1 in 100 year return period (R = 0.5) [11]. For comparison, typical buildings are designed to IL2 with a 50-year design life, or a 1 in 500 year return period (R = 1.0). Also for comparison, tanks in the dairy industry are often designed for IL2 with a 25-year design life, or a 1 in 250 year return period (R = 0.75). Wine and dairy tanks that support catwalks are not designed with a higher return period factor. The significant amount of irreparable damage to wine tanks in relatively moderate earthquakes has led to discussion about the appropriate earthquake return period for design of these tanks. A higher return period factor would increase the design forces, and result in a more robust tank design. However, it would not prevent the majority of significant damage that was observed throughout the region. Knuckle deformation on plinthsupported tanks and base punching due to the tabs on leg-supported tanks were both caused by detailing rather than the magnitude of demand. This is exemplified by empty tanks that exhibited minor indications of these types of damage, although their demands would have been substantially lower than design. Also, the newer plinth-supported tanks with a fully supported base (designed with R = 0.5) did not exhibit knuckle deformation, while other nearby similar tanks had severe knuckle damage. Careful attention to tank details and consideration of an appropriate hierarchy of failure is recommended over increasing the design earthquake return period to achieve better tank performance in future events. 7.5 RELOCATING TANKS As discussed previously, wine tanks are unique structures because they can be transported large distances relatively easily. However, they are still designed for site-specific earthquake loading. When transporting older tanks to a new site, performance expectations for the tanks must be adjusted to reflect their age, imperfect condition, and the possibility that they were designed for the lower seismic demands of another region. Anchorage installed for these tanks after transport should be carefully designed to account for the capacity of the tank itself rather than the site-specific parameters to prevent overstressing the tank unintentionally. For new design, it may be practical to design tanks to the highest anticipated site-specific parameters (i.e., highest Z factor corresponding to a wine production region in NZS1170.5) to prevent future issues with tanks that have been transported during their lifetime. Cost, robust design, and site flexibility need to be considered with winery owners to determine the appropriate design parameters. REFERENCES [1] Ballingall, J. and Schilling, C., NZIER (New Zealand Institute of Economic Research) report to New Zealand Winegrowers, April, 2009. [2] Morris G.J., Bradley B.A., Walker A., Matuschka, T. Ground Motions and Damage Observations in the Marlborough Region from the 2013 Lake Grassmere Earthquake. Bulletin of the New Zealand National Society for Earthquake Engineering, Vol. 46, No. 4, December 2013. [3] Wozniak R.S., Mitchell, W.W., Basis of Seismic Design Provisions for Welded Steel Oil Storage Tanks. Proceedings of Refining Department, API, Washington D.C., 1978: p.485-501. [4] Crawford, K.N., Standard Seismic Resistant Details for Industrial Tanks and Silos. New Zealand Earthquake Research Foundation. Auckland, New Zealand, October 1990. [5] Standards New Zealand. NZS1170.5, Structural design actions, Part 5: Earthquake Actions – New Zealand. 2004. [6] Priestly M.J.N., Wood J.H., Davidson B.J. .: Seismic Design of Storage Tanks. Bulletin of the New Zealand National Society for Earthquake Engineering, Vol. 19, No. 4, December 1986. [7] Carluccio A.D., Structural Characterisation and Seismic Evaluation of Steel Equipment in Industrial Plants. PhD Thesis in Seismic Risk. University of Naples, 2007. [8] New Zealand National Society for Earthquake Engineering. Seismic design of storage tanks. 1986. [9] Whittaker D., Saunders D. Revised NZSEE Recommendations for Seismic Design of Storage Tanks. 2008 NZSEE Conference Paper No. 4, 2008. [10] New Zealand National Society for Earthquake Engineering. Seismic design of storage tanks. 2009. [11] Crown Sheetmetal Ltd. Personal Communication. August 2013.
© Copyright 2026 Paperzz