VNIIGAZ Offshore Hazards: Assessing the Impact of Icebergs on Offshore Production Platforms
Dmitry A. Onishchenko (Gazprom VNIIGAZ LLC) Part 2
Introduction
Resolving the challenge of developing Russia’s offshore oil and gas resources, those on Arctic shelf being first and foremost, requires much scientific research, both theoretical and practical. One of the main challenges is attempting to calculate the ice loads that can hit offshore platforms which is still critical for developing most fields in Russia’s Arctic shelf. The correct estimation of the ice loads that are likely to be encountered is key – overestimation of the loads will result in higher capital expenditure, while underestimation will increase the risk of damage or even destruction during operation. As with other loads caused by natural factors, ice loads can vary enormously, and this is why calculating ice load values must be done within the probabilistic framework. One of the current challenges facing the industry is the construction of offshore platforms for a number of promising fields based in a number of Russia’s northern seas, including the Barents and Kara Seas.
Using current data, the study proposes a physical and mathematical model which allows various iceberg hazard indicators to be analyzed. These include the probability of an iceberg/platform collision during a given timeframe, the probability distribution of the iceberg’s kinetic energy on collision, etc.
Requirements are established for the initial data required for corresponding calculations. The paper discusses the results obtained and the possibility of their use in practical design work.
1 Factors behind the probabilistic nature of an ice load
The ice load on offshore platforms arise when the platform is affected by moving ice cover, generally of spatially inhomogeneous nature and consisting of various types of ice formations (e.g. level ice, rafted ice, ridged ice, icebergs etc) – the full list can be found in World meteorological organization’s sea ice nomenclature [1]). Ice breaks as it advances and comes in contact with the platform and, in turn, creates pressure on the platform’s hull. Therefore, the ice load on an offshore platform is governed by two separate destruction processes – local ones that take place during ice/platform contact, and the global processes which complement the destruction of the ice cover or its individual elements forming part of that cover on the whole (generally expressed as ice cracking). We should note that in almost all cases, an iceberg-to-platform collision will result in the local destruction of ice.
The total ice load on the entire platform (it is often referred to as global load) is a function of time which can, generally, vary significantly. Because of this variation, it is important to examine ice loads as random processes. There are a large number of parameters to take into account, including those that determine the ice regime in the area near the platform, while others do “manage” the interaction of ice cover with the platform.
Listed below are some of the parameters which determine the nature of an ice load. These parameters in particular determine the variability of the ice cover:
» large number of ice formations (first year and multi-year level ice, ridges, icebergs etc);
» thickness, morphometric composition, spatial boundaries of ice formations;
» ice drifting at various velocities;
» velocity and temperature during the contact with the structure;
» frequency of certain ice formations shapes;
These parameters in particular determine the destruction of the ice cover:
» local ice strength;
» large number of destruction patterns (shearing, crushing, bending, stability loss, cracking, piling-up and etc.);
» few interaction scenarios (impact, pile-up, freeze-up and etc.).
The above parameters are random values, based either on observations or mathematical modeling. For many of the listed parameters, however, the observation ranges are short, and may contain errors (as compared against “true” distributions which are apparently, unknown). Moreover, due to the internal heterogeneous properties of ice, its load, even for the same formation (e.g. level ice) with spatially invariable “external” parameters is of course random.
2 “Probabilistic” and “deterministic” design method
The random nature of ice loads is nothing new – it is normal for all loads caused by natural factors such as wind, wave, snow etc. Engineers have long mastered random loads: so called representative and design load values are used for design purposes along with representative and design values of the strength of structural material and soil. Design criteria for limit state methods are generally expressed as
Qd ≤ Rd, (1)
where Qd is a design value of the force factor such as force, bending moment, stress in a given element of the structure under design (or “action effect” using the new terminology [2], which is calculated for a given combination of applied loads;
Rd is a design value of bearing capacity for an element, usually calculated through strength properties of the soil or a material. (We should note that in regulatory documentation equations such as (1) are usually seen in modified form with additional multipliers. This doesn’t affect subsequent analysis in any meaningful manner, thus for the purpose of simplicity we shall use this imprecise equation (1) for the design criteria.
In turn, design values are determined based on representative values Q0,R0:
Qd = γf Q0, Rd = R0/γm, (2)
where γf and γm are so called partial safety factors (for load and for material, correspondingly).
We should note that all listed values are deterministic, thus for conventional design work, the random factor is effectively excluded: it only appears in the determination of representative (or, directly design) strength and load values. From now on, we only discuss action effects and corresponding loads. The representative value of ice load on the structure (which is random as we have stated) is generally accepted at a value with predetermined recurrence period T. Recurrence period values vary for different load types. Thus, present day regulatory norms [3], recommend T = 100 years for principal load combinations and T ~ 103…104 years for extraordinary combinations.
Thus, according to its definition, the representative load value expressed as qα, is determined as a value that can be exceeded during a given (arbitrary chosen) year with a probability of α=1⁄T, which can be expressed in the following equation:
Pr{Q>qα at least once a year}=α, (3)
where Pr{A} stands for the probability of a random event A. Value α is called accumulated probability or occurrence; it is often expressed as a percentage value. We should note that α = 0,01, or 1 %, for recurrence period T = 100 years. Sometimes condition (3) is better expressed as
Pr{Qmax>qα }=α, (4)
where Qmax is annual maximum of ice load.
Considering that ice load may occur under the influence of various ice formations (level ice, rafted ice, ridges, icebergs etc), thus equations (3) and (4) are normally written separately for different types of ice formations. That is why a few representative ice load values are used in design work corresponding to various design situations – action from level ice, rifted ice, ridges, icebergs etc.
Now let’s introduce cumulative distribution function for Qmax value and denote it as FQ (x), so the equation (4) can be turned into:
FQ (qα )=1-α. (5)
On the front, equation (5) seems very simple. A closer look, however, reveals that finding function FQ (x) is not at all a trivial task: it depends (and often in a very complicated way) on distribution functions of all values influencing the load as well as on geometrical properties of the structure under design. To illustrate the associated challenges we should mention that a different approach than equation (4) is used to determine representative load values for other natural load types such as wave loads or current loads, specifically:
Q0=Q(ω1,ω2, …), (6)
where ω1,ω2, … is a set of design parameters values for a design situation, and Q(x1,x2, …) is the so called load formula. In cases of wave load for example, the set of parameters includes wave height of certain occurrence along with associated period and average wave length value. In cases of current, governing parameter is the current velocity of certain occurrence. It could also be noted that Russian codes up until recent times used a similar approach for estimating events of level ice impacts; governing parameters for this are representative ice thickness (at 1% occurrence) and the design ice strength. Note that Q(x1,x2, …) are regular deterministic functions explicitly stated in corresponding structural codes.
Thus, design value for ice load in case with level ice, uses the following equation
Q0=mkb kV Rc Dhd, (7)
where hd is the design ice thickness at the platform location, Rc is the design compression resistance (strength) for ice, D – width of the structure affected by ice, kb, kV and m are some constants that only depend on structure geometry and hd.
It is crucially important that representative value Q0 has no (at least no explicitly stated) occurrence value assigned (this is why we use distinctly different symbols for related values of Q0 and qα), while occurrence requirements are applied to the parameters of “impacting” natural objects: wave height, current velocity, level ice thickness etc. This approach makes it possible to make a clear distinction between the stage of design preparation when design parameters of “impacting” natural objects are defined (this is traditionally done by specialized engineering research organizations) and the design stage of a project itself, when the engineer’s task is to ensure the accuracy of inequalities such as (1) with consideration of (2) and (7) by proper selection of suitable structural design and materials.
When we look at a “probabilistic approach to design work”, illustrated by equation (4), this effectively creates a catch twenty two situation: the design (development of construction solutions) can’t be performed until probability distributions FQ(x) are known for all estimated events, while these distributions can’t be calculated until structural design is available. Moreover, this challenge requires the surveyors to build probability distribution functions for all variables and factors affecting the loads (a large but incomplete number of which are listed above), whereas the conventional approach only envisages the determination of corresponding representative values at the survey stage. The former requires a significantly larger volume of observational data. This raises the following question – in this catch 22 situation, who is responsible for the adequacy of required probability distributions which must be known even to their tail values, including those for α reaching
10-3…10-4, and even up to 10-5?
The RF Government is active in updating its regulatory construction framework, including that for the design of offshore oil and gas facilities. With that, new or updated regulatory documents enforce the probabilistic approach for design criteria as in equation (4) (see e.g. [4]). Unfortunately, the authors of these documents fail to consider that the probabilistic design approach requires a) enormous volumes of initial data (unobtainable through conventional engineering research; especially considering that regulatory norms for design engineering surveys do not include any such new “probabilistic” requirements) and b) subsequent laborious work on constructing probabilistic load distribution functions FQ (x). This makes it hard to expect actual adherence to the requirements of probabilistic criteria such as (4) in design of offshore platforms. There is the hazard that even if these are performed practically, it would only be formal: using some surrogate distributions FQ (x), the reliability of which is next to impossible to substantiate. This, in turn, may negatively impact the reliability of the designed facility.
3 Requirements for iceberg impact design load
The assertions above are not to say that probabilistic methods have no place in design work: they are only to state the necessity a clear understanding of which issues can better be resolved by applying probability theory for developing specific design solutions.
One such issue is the assessment of iceberg hazards for platforms located in offshore areas where icebergs drifts are probable. For the purposes of design, iceberg to platform collision should be considered as a special load. Below we list the examples of constructing a probabilistic model which enables us to form “a hazard” and the estimation of corresponding quantitative indicators. Some studies on this matter use the term “risk”. Formally, it includes assessment of unfavorable consequences along with determination of probability for unfavorable events. Because this paper does not describe the consequences of a possible iceberg collision, we shall not use the term “risk”.
Some iceberg hazard parameters are:
» probability of iceberg to platform collision during a one year period;
» probability of collision during one a year period for an iceberg with dimensions and mass exceeding an established value;
» probability of collision during a one year period for an iceberg with kinetic energy value above the established;
» probability of collision for an iceberg approaching the platform from distance L;
» probability of collision during one year period with global load on the platform below the established value and etc.
We should note that the probability of an iceberg collision (to be more precise, the assessment of probability calculated based on available statistical data using one or another probability model) can’t by itself be considered as a comprehensive data set required to complete a platform design. Actually, if only small icebergs or their bergy bits reach the offshore field, then the corresponding design situation apparently won’t be crucial for the design, while should there exist a probability of an even rare occurrence of large iceberg collisions, it has to be taken into account. Thus, one of possible formalization options envisages the requirement of calculating such parameters is the probability for the platform with known shape and size to be impacted by an iceberg with kinetic energy above an established value for a given time period [5].
The general design criteria require the platform to maintaining its bearing capacity under the influence of certain design loads. Let us examine this issue in relation to the load on a possible iceberg to platform collision. The first thing to mention is that an estimated iceberg collision is rare [6]. This means that the actual structure built to serve for a period of 25-50 years (in most cases) will almost never experience an iceberg collision. Still, the collision probability does not equal zero. To this extent, one known “onshore” analogue is the seismic load.
“Probabilistic” design corresponds to the latter of the listed iceberg hazard factors. The resulting load for an iceberg to platform collision will significantly depend on the local shape of the iceberg surface coming into contact with the
platform hull.
Resolving the task of calculating estimated load qalpha from an iceberg collision (at α~10-4…10-5 ) includes a number of interrelated factors. The necessary input data can provisionally be divided in 3 large units.
Unit I1: statistical data on iceberg observations, their shapes and sizes near the platform location. With that, due to the rare occurrence of such events, statistically solid data would require very long observation data sets; it is apparent that a standard 5-year survey cycle by itself won’t provide the required volumes of information, and therefore archive data analysis is required. For example, corresponding databases were created in Canada for the Grand Banks of Newfoundland and in Russia for the Barents Sea ([7,8]).
Unit I2: The physical and mechanical model of an iceberg to platform collision, which includes ice destruction patterns near the impact area and describes an iceberg’s dynamic behavior during the collision. A number of such models was developed (see e.g. [9-11]). However, a few problems remain somewhat unresolved. In particular, there are challenges related to modeling the dependency of the collision area based on the penetration depth of the collision, which largely determines the intensity of the impact to the platform (an example of resolving such task can be found in [12]), along with consideration of hydrodynamic effects occurring when two massive bodies (iceberg and platform) come close to each other.
Unit I3: iceberg drift model (e.g. rectilinear or chaotic) near the platform location. Main “moving” factors are near-surface currents, winds and, possibly, ice cover. Because the iceberg trajectory observations data are insufficient to obtain reliable statistical conclusions on spatial and time parameters of iceberg trajectories, developing adequate and efficient “atmosphere-ice-ocean” models would assist in finding a solution for this task.
In general cases, the equation required to find design value qalpha with consideration of (6) can be expanded to
Pr{maxQ (ω(i))>qα }=α, (8)
where ω(i) = (ω1(i), ω2(i) ) is the aggregate of all random parameters influencing the ice load (for an iceberg, that would be its velocity at collision, its mass and inertia moments ice strength and surface shape at the zone of contact and etc. – these parameters appear in units I1 and I2); i = 1,…, N are all iceberg interactions on a platform during one year period (usually N being the number of collisions is random and the case of N = 0 is not excluded; pertaining to unit I3), and
Q = Q(ω) (9)
is the iceberg load formula used to calculate maximum load for a specific single collision event with prescribed values of random parameters ω (unit I2).
Various approaches can be used to model iceberg hazards [3, 5,12-16]. Most of these use the Monte Carlo statistical method. Below is a modification of an approach developed by the pioneering works of Canadian specialists [5,6], which make it possible to obtain estimated correlations in any analytical form. Another example of implementing a modified approach for a population of ice flows affecting the platform can be found in article [17].
End of Part 1
List of literature
1. WMO sea ice nomenclature. WMO/OMM/ВМО – No. 259. Amendment No. 5 (2004). [Digital resource] URL: http://www.aari.ru/gdsidb/XML/wmo_259.php?lang0=1. (as displayed on 15.11.2013)
2. GOST R 54257-2010 Reliability of structures and foundations in construction. Principal provisions and requirements.
3. ISO 19906:2010 Petroleum and natural gas industries – Arctic offshore structures
4. SP 38.13330.2012 Loads and impacts on hydrotechnical facilities (from waves, ice and vessels) (updated version of SNiP 2.06.04-82*).
5. Dunwoody A.B. The design ice island for impact against an offshore structure. Proc. 15th Offshore Technology Conference, Houston, USA, 1983, p. 325–332.
6. S471-04 General requirements, design criteria, the environment, and loads. CSA, 2004.
7. Verbit S., Comfort G., Timco G. Development of a database for iceberg sightings off Canada’s east coast. Proc. 18th Int. Symposium on Ice, IAHR’06, Sapporo, Japan, 2006. Vol. 2, pp. 89–96.
8. Naumov A.K., Zubakin G.K., Gudoshnikov Yu.P., Buzin I.V., Skutin A.A. Ice and icebergs near Shtokman gas condensate field. Works for international conference “Developing Russian sea shelves” (RAO-03), St. Petersburg, Sept. 16–19, 2003, pp. 337–342.
9. Vershinin S.A., Nagrelli V.E., Yermakov S.V., Onishchenko D.A. Impact interaction of iceberg and ice-resistant offshore platform for the Shtockmanovskoye field. Proc. First Int. Conf. on Development of the Russian Arctic Offshore, St.Petersburg, Russia, 1993, pp. 192–196.
10. Matskevitch D.G. Eccentric impact of an ice feature: linearized model. Cold Region Science and Technology, Vol. 25 (1997), pp. 159–171.
11. Matskevitch D.G. Eccentric impact of an ice feature: non-linear model. Cold Region Science and Technology, Vol. 26 (1997), pp. 55–66.
12. Fuglem M., Muggeridge K., Jordaan I. Design load calculations for iceberg impacts. Int. J. Offshore and Polar Engineering, Vol. 9. No. 4 (1999), pp. 298–306.
13. Nevel D. Ice force probability issues. Proc. IAHR Ice Symposium, Banff, Canada, 1992, pp. 1497–1506.
14. Korsnes R., Moe G. Approaches to find iceberg collision risks for fixed offshore platforms. Int. J. Offshore and Polar Engineering, Vol. 4. No. 1 (1994), pp. 48–52.
15. Fuglem M., Jordaan I., Crocker G. Iceberg-structure interaction probabilities for design. Can. J. Civ. Eng., Vol. 23 (1996), pp. 231–241.
16. Naumov A.K. Distribution of icebergs near Shtokman gas condensate field and assessment of iceberg to platform collisions. Lib.: Complex research for ice and hydrometeorological events and processes near the Arctic shelf. Works of AARI, v. 449, St. Petersburg, 2004, pp. 140–152.
17. Onishchenko D.A. Probabilistic modeling as a tool to determine estimated ice loads in arctic shelf environment. Science and technology for the gas industry, №1 (2006), pp. 62–80.
