Structural safety assessment for FLNG-LNGC system during offloading operation scenario

The crashworthiness of the cargo containment systems (CCSs) of a floating liquid natural gas (FLNG) and the side structures in side-by-side offloading operations scenario are studied in this paper. An FLNG vessel is exposed to potential threats from collisions with a liquid natural gas carrier (LNGC) during the offloading operations, which has been confirmed by a model test of FLNG-LNGC side-by-side offloading operations. A nonlinear finite element code LS-DYNA is used to simulate the collision scenarios during the offloading operations. Finite element models of an FLNG vessel and an LNGC are established for the purpose of this study, including a detailed LNG cargo containment system in the FLNG side model. Based on the parameters obtained from the model test and potential dangerous accidents, typical collision scenarios are defined to conduct a comprehensive study. To evaluate the safety of the FLNG vessel, a limit state is proposed based on the structural responses of the LNG CCS. The different characteristics of the structural responses for the primary structural components, energy dissipation and collision forces are obtained for various scenarios. Deformation of the inner hull is found to have a great effect on the responses of the LNG CCS, with approximately 160 mm deformation corresponding to the limit state. Densely arranged web frames can absorb over 35% of the collision energy and be proved to greatly enhance the crashworthiness of the FLNG side structures.


Introduction
The increasing global demand for natural gas has motivated the exploitation of offshore gas fields. However, the use of traditional pipeline systems is not considered as a technically or economically feasible option for many stranded and marginal gas fields. Floating liquid natural gas (FLNG), which is a new concept for offshore facilities to extract and liquefy natural gas, has been recently proposed. To explore stranded offshore gas reserves, FLNG vessels can be stationed at the target gas fields for a long time, and offload the liquid natural gas via an LNG carrier (LNGC). For the offloading operation for FLNG and LNGC, side-byside transfer system has been considered as a feasible configuration. In the side-by-side configuration, the FLNG vessel is closely connected to the LNGC by many hawsers and fenders, and the two vessels are exposed to the complex ocean states during the whole offloading operation period. Loads from the connecting system and the hydrodynamic interactions between the FLNG vessel and the LNGC could increase the risk of contact accidents. There exists a high probability of collision between the FLNG vessel and the LNGC during offloading operations. Severe collisions could lead to cargo tank damage and result in an uncontrolled release of the liquid products. Therefore, guaranteeing the structural crashworthiness of FLNG structures and CCSs is a critical safety issue, which should be recognized as one of the most important challenges for the progress of the FLNG technology.
As a new type of floating offshore structure, the previous research on FLNG has primarily focused on the hydrodynamic performance and sloshing effects (Zhao et al., 2012a(Zhao et al., , 2012b(Zhao et al., , 2014Kim et al., 2010), while little effort has been spent on the crashworthiness and safety assessment of FLNG vessels under LNGC collisions during offloading operations. Because a similar LNG CCS is employed on both FLNG vessels and LNGCs, previous studies on the crashworthiness of the LNGC structures and the CCS used in LNG cargo tanks can provide a good guidance for the present research. Paik et al. (2002) analyzed the resistance of spherical-type LNGC structures to bow collisions. Based on energy dissipation capacity, two limit states for LNGC structures are introduced and a CRI (collision resistance index) coefficient is defined to evaluate the safety of LNGC structures under bow collision scenario. Wang et al. (2008a) analyzed the responses of the membrane structure of the spherical LNG carriers under ice loading effect. In addition, Wang et al. (2008b) concluded that the ice load applied on an LNGC can be considered a quasi-static load, and the acceptance criteria and a complete procedure for the strength evaluation of LNGC structures under ice loads were developed. To evaluate the fatigue strength of the membrane cargo containment systems, Shaposhnikov et al. (2014) performed a finite element analysis by taking both the ice effects and the local structural vibrations into consideration. Shaposhnikov et al. (2014) also studied the emergency situations for LNG transport, such as the accidental collisions of vessels and grounding on rocks. Through numerical simulations and experimental methods, Kim et al. (2011) performed a comparative safety assessment for independenttype and membrane-type LNG carriers under ice collision scenario. It was observed that the vibratory accelerations induced by ice impacts were intense and transient. The mastic ropes in the membrane CCS were highlighted in another research of Kim et al. (2012), and in this research, the static and dynamic material properties of the mastic were examined via low and high speed tensile tests and thermal tests, on purpose of evaluating the safety of the mastic ropes upon repeated ice impacts. Park et al. (2014) examined the safety of FLNG vessels in collision with icebergs in the Arctic area, and the effect of temperature on the structural performance of steel panels was emphasized in the research. Considering seawater effects, Lee and Nguyen (2011) used the fluid-structure interaction (FSI) analysis technique to estimate a more realistic and reliable ice-LNGC collision response. Zhang (2006Zhang ( , 2007 studied the ship collisions with liquid tanks taking fluid-structure interaction into account by the arbitrary Lagrangian-Eulerian finite element method. In summary, the above studies on the response of LNGC in collision scenarios provided a good guidance for the study on the structural response of FLNG in collision scenario. Nevertheless, the FLNG vessels have different displacement, constructional features, mooring deployment and operation modes compared with the LNGC, therefore, the responses and characteristics of FLNG-LNGC collisions during the offloading operations are different from those of ship-LNGC collisions or ice-LNGC collisions. Thus, re-search on the crashworthiness of FLNG structures is necessary and has a profound engineering significance.
In the present work, a crashworthiness study of an FLNG cargo containment system and side structures under LNGC collisions during side-by-side offloading operation is conducted to evaluate the safety of the offloading operation. Three sets of finite element models are used to perform the numerical simulations. The LNG containment system is modeled in detail on the left bulkhead of the LNG cargo tank in the FLNG side model. The major parameters of the collision scenarios are defined by a model test conducted at Deepwater Offshore Basin in Shanghai Jiao Tong University. Based on the results of the model test and potential accidents, two typical collision scenarios are designed. A limit state for FLNG structures under collisions is determined based on the failure of the CCS. To assess the safety of the FLNG vessel under LNGC collisions, great attention is paid to the structural responses of CCS and the performance of the primary structures resisting collision forces. The energy dissipation and collision forces are also obtained and an in-depth analysis is made. Finally, the crashworthiness of FLNG side structures is evaluated, with helpful insights for the structural design of the FLNG vessel.

Finite element model
The basic design of the FLNG system in the present research was complemented by MARIC (Marine Design & Research Institute of China) and CNOOC (China National Offshore Oil Corporation). The principal dimensions of the FLNG and the LNGC are shown in Table 1 and Table 2, respectively. To perform the numerical simulations, the following three finite models were established: (a) FLNG side structures with a detailed LNG cargo containment system; (b) LNGC side structures; (c) LNGC bow structures (see Fig. 1). On purpose of reducing the time cost, the FLNG model is built for one hold length and is rigidly fixed at both ends. The length of the LNGC side structures is shorter than that of the FLNG side structure, and the LNGC model is defined with rigid property, assuming the FLNG side struc- ture absorbing all of the striking energy and the conservative simulation results can be obtained. The cross section of the FLNG side structure is shown in Fig. 2. A distinct structural characteristics of the structural design of the FLNG is that the web frames are arranged at every frame. The length of the FLNG side structure numerical model is 37 m, and the frame spacing is 1 m.
The FLNG uses the GTT (Gaz Transport/Technigaz, France) NO.96 CCS, which is shown in Fig. 3a. The GTT system is one of the most widely used membrane contain-ment systems, consisting of two layers of insulation boxes made of plywood. All of the insulation boxes are filled with one kind of powdery insulation called perlite. The boxes have parallel internal bulkheads and the sealing plates made of plywood sheets. Thin sheets of a high nickel alloy called invar are used as the primary and secondary barriers. Mechanical couplers welded on the inner hull to fix the insulation boxes to be the inner hull. On purpose of ensuring the internal surfaces of the insulation boxes are sufficiently flat to support the invar membrane, appropriate mastic ropes are laid under the bottom of the secondary insulation boxes and are bonded to the inner hull. There is a large gap between the adjacent boxes in the longitudinal direction and a small gap in the transverse direction. The large gap of 60 mm is filled with insulation material. According to ABS (2006) and LR (2009), the invar membrane and perlite can be ignored in finite element models. In order to simulate the interactions between the different structural members, the primary boxes, the secondary boxes and the inner hull should be modeled as separate entities, and connected by spring elements with different stiffnesses. For couplers restricting the motion of the insulation boxes, stiff beam elements are used. Both the spring and the beam elements allow relative motions between the different structures, making the numerical simulation more realistic and reliable. The detailed FEA model for the NO.96 CCS on the left tank bulkhead is shown in Fig. 3b.
Both the steel structures and plywood plates are modeled using Hughes-Liu shell finite elements in the nonlinear finite element code, LS-DYNA, and the properties are presented in Table 3. The element size for FLNG side structures is 200 mm, and the element size for plywood plates in NO.96 CCS is 25 mm. The linear elastic-plastic material mild steel is adopted for the FLNG. Because the sizes of the elements used in the steel structures have little differences, a critical failure strain of 0.20 (NORSOK Standard, 2013) is defined for the steel structures. Once this limit is reached, relevant elements will fracture and be removed. The strain ratedependent plastic behavior is considered by the Cowper-Symonds equation: where is the dynamic flow stress, σ 0 is the static flow stress, is the plastic strain rate, C = 40.4 s -1 and p = 5 are the parameters characterizing the strain rate effect. The plywood is one kind of brittle material without plastic behavior, which means the plywood would rupture once its ultimate strength state is reached. So the plywood is defined as the linear elastic material to focus on its maximum stress responses.

Definitions of collision scenarios
A model test of the FLNG vessel during side-by-side offloading operations with an LNGC was completed in Deepwater Offshore Basin at Shanghai Jiao Tong University. The side-by-side configuration is shown in Fig. 4. The details of the experiment could refer to Xu et al. (2015). As shown in Fig. 4, the modeled FLNG vessel and LNGC are connected by eight hawsers and four fenders installed on the LNGC, and the maneuverability of LNGC is not taken into consideration. The collision scenario parameters are determined by the load conditions of the FLNG vessel and LNGC, therefore, three load condition combinations are chosen based on the most probable operating conditions. Primary collision parameters, such as the collision frequency and the relative velocity, can be obtained from experimental data (see Table  4). With the probability of failure of the hawsers and fenders, the connecting systems are ignored in the numerical simulations. Therefore, from a conservative perspective, it is assumed that the impact force is directly applied on the FLNG side structures.
As indicated in Table 4, there is a high collision frequency and a large impact velocity between the FLNG vessel and LNGC. Although the two vessels have consistent yaw motions, the small relative yaw can lead to a substantial effect on the collision area for offshore structures in excess of 200 m. As shown in Fig. 4a, once the LNG carrier turns right to a certain angle and strikes the FLNG vessel, the entire LNGC striking momentum is applied concentrated within a small region of FLNG side structures. Such a collision may cause serious damage to the FLNG side structures.
Given the aforementioned situations, the following two representative collision scenarios are defined: (1) a perfectly parallel collision between the LNGC and FLNG side structures; (2) a collision at the maximum yaw degree between the LNGC bow and FLNG side structures (see Fig. 5).
To ensure more reliable and safer results, conservative collision parameters are selected for the scenario simulations. According to Bond's (2011) and Kenny's (1988) studies, a typical impact velocity for a specified collision design is 2 m/s, which is equal to the maximum velocity in Table 4. Therefore, 2 m/s is used as the striking velocity of the LNGC in all three Scenarios. In Scenario 2, the maximum relative yaw is set to 3.04° according to the data in Table 4. For simplicity, the added mass coefficient for the sway motion is taken as 0.7 (Pedersen and Zhang, 1998). Detailed definitions of the collision scenarios and cases are described as follows, and cases definition can be seen in Table 5.

Definition of the structural limit state
For special vessels with an LNG cargo containment system, including FLNG and LNGC, researchers have not reached an agreement on the limit state of structural failure. When Paik et al. (2002) studied the resistance of a spherical-type LNGC to bow collision, he defined two accidental limit states: (a) the striking ship's bow penetrates the bound-   Han et al. (2008) deemed this criterion to be too conservative. In Han's structural risk analysis of the NO.96 membrane-type LNGC under ice collision, it was considered that inner hull deflection as the critical factor for the safety of the containment system. In addition Han compared three candidates for the survival limit condition of LNGC and eventually put forward a new limit state, i.e., a 70 cm inner hull deflection. Similarly, a 4 mm per meter deflection limit was used by Kõrgesaar and Ehlers (2015).
To assess the safety of ice-class LNGCs,  analyzed the structural response of an LNGC under ice loads, focusing on the maximum stress in CCS materials. With the relatively low strength of plywood and the function of the CCS, the failure of the insulation boxes is of great importance. Therefore, the failure of the insulation boxes is defined as the limit state for FLNG side structures due to collisions with an LNGC, which is more reasonable and conforms to classification society rules. The failure of the insulation boxes can be divided into three main modes: (tensile/compressive) strength failure, buckling failure and shear failure. ABS (Paik, 2006) and NK (Dobashi and Usami, 2010) undertook a full-scale model test on the NO.96 containment system. Paik (2006) and Lu et al. (2012) studied the structural response of an LNG CCS under impact loads using the FEA method. All of the researchers found that strength or buckling failure initially occurs in the insulation boxes. Therefore, the strength and buckling failures are regarded as the main failure modes for the LNG CCS in this study. The buckling failure mode corresponds to buckling distortions in the vertical bulkheads of the insulation boxes. According to the ABS rule (2006), the ultimate strength of the plywood in the insulation boxes is 40 MPa. Once the maximum stress of the insulation boxes reaches this critical value, the containment system will collapse, and the numerical computation can be terminated.

Results and discussion
The results and discussion comprise three main parts: (1) structural responses of the CCS; (2) structural performance of the FLNG side structures; (3) energy dissipation and collision forces. Different characteristics among the two scenarios are compared and summarized.

Scenario 1
The side segment of the LNGC is located in the middle of the FLNG side model in Scenario 1, with a very large initial contact area for parallel contact. The time histories of the maximum v-m (Von-Mises) stress of the insulation boxes for the three cases in Scenario 1 are shown in Fig. 6. The curves show stress peaks at the very beginning of the contact and fluctuations in the later phases. When Lee et al. (2010) analyzed ice-LNGC collisions, similar phenomena were found in large iceberg collisions. Unlike the stress of the CCS, the deformation of the inner hull increases gradually. Therefore, this dynamic response of CCS in Scenario 1 is excited by the propagation of large shock waves in the structures, instead of deformation. Contact discontinuities play an important role in the production of shock waves. Here, the contact discontinuity corresponds to the sudden contact between the FLNG vessel and the LNGC with a large initial contact area. Such a sharp change in contact is the most prominent characteristics of Scenario 1. At the beginning, the energy is concentrated in one direction normal to the contact surface, so the shock wave propagates in the transverse direction of the FLNG vessel. Then, due to the shear effects on its outline, the wave front gradually changes  into a curved wave shock with radial propagation. This type of diffuse propagation makes the initial energy spreading over a larger non-contact area rather than that only in the transverse direction of the vessel. Therefore, the energy density decreases rapidly. Nevertheless, the collision cases in Scenario 1 have a very large initial area. This feature demonstrates that the non-contact spreading area for the shock waves is limited and the curving of the wave front is relatively slow. Therefore, the energy density remains high when the shock wave reaches the CCS, resulting in abrupt local vibrations and high stress levels.
Another characteristics of the curves in Fig. 6 is the very short duration of the fluctuations, especially before 0.05 s. The high propagation velocity of the shock wave is nearly equivalent to the acoustic speed for the materials (approximately 5300 m/s for mild steel and 2600 m/s for plywood (Hallquist, 2006)). This explains why the fluctuation duration does not exceed several milliseconds in Fig. 6.
The positions where the maximum stress occurs in the insulation boxes are also important concerns for researchers (Wang et al., 2009;Chun et al., 2011). Initially, the maximal stress appears in the vertical bulkheads of the secondary boxes, especially on the side nearest the inner hull, as shown in Fig. 7. This phenomenon is primarily because the impact load excited by the collision is firstly transmitted to the lower side of the bulkhead through the mastic ropes. The deformation of the inner hull gradually increases as the penetration increases. Furthermore, due to the high accelerations, the two layers of the insulation boxes tend to separate from each other and move away from the inner hull. Consequently, there are growing constraints on the insulation boxes by the couplers, and the maximum stress shifts to the seal and bottom plywood of the primary boxes. Nevertheless, for the top covers of the primary boxes where the invar membranes are laid, the max. v-m stress is only 9.49 MPa, which demonstrates that the top covers are sufficiently strong to support the invar membranes.
The structural responses of CCS in Scenario 1 are summarized in Table 6. The LNGC in Case1-3 takes longer time to stop for much larger collision mass, which also causes a larger deformation of inner hull in Case1-3. The deforma-tion of the inner hull is small relative to the size of side structures (3200 mm width). Nevertheless, the peak stress of CCS reaches 37.27 MPa, just a little under the limit state. The phenomenon that the stress does not correspond with inner hull deformation indicates the dominant role of shock wave effect in Scenario 1.

Scenario 2
Unlike the parallel collision simulated in Scenario 1, the mass of the entire LNGC initially concentrates over a small region in Scenario 2. Owing to the curvature of the LNGC bow shoulder, the contact area becomes larger gradually as the penetration increases, instead of a sharp change in the contact area. Therefore, the contact discontinuity and shock waves are not present in Scenario 2. The time histories of the maximum v-m stress of the insulation boxes for Case2-1 to Sase2-3 in Scenario 2 are shown in Fig. 8. Clearly, the stress in all three cases increases constantly without initial peaks and obvious fluctuations. The stress of the insulation boxes in Scenario 2 is primarily related to the deformation of the inner hull, which is shown in Fig. 9. The curves in Fig. 8 generally show a similar trend to the curves in Fig. 9. In the inner hull, the region that is first affected locally by the collision has larger deformation. Then, the adjacent insulation boxes have to sustain heavier loads and suffer more restrictions from the couplers. Therefore, the maximum stress increases gradually with the deformation of inner hull increasing.
Without shock waves, the acceleration of the insulation boxes is much smaller in Scenario 2 than that in Scenario 1. Hence, the insulation boxes have a smaller separation tendency, and the coupler has a limited restriction on the primary boxes. Nevertheless, due to the local deformation of the inner hull, both loads from the inner hull and the constraint of the couplers primarily act on the secondary boxes.  Therefore, the danger region in CCS is concentrated in the outer vertical bulkheads and the seal plates of the secondary boxes in Scenario 2. The structural responses of CCS in Scenario 2 are summarized in Table 7. Owing to different relative draughts in various load condition combinations, the initial contact point in Case2-1 is further from the CCS than the other two cases. Therefore, it takes more time for the failure of the CCS in Case2-1. Although the maximum inner hull deformation in Case2-1 is much larger than that in other cases when the CCS falls into failure state, the deformation of the inner hull under the damaged insulation boxes is also about 160 mm like that in the other two cases. In Scenario 2, it could be concluded that deformation of approximately 160 mm in the inner hull corresponds with the limit state of CCS (40 MPa).

Summary of the CCS responses
A summary of the CCS responses is shown in Table 8 based on the aforementioned discussions. Scenario 2 is found to be the most dangerous scenario, and shock waves also induce a pronounced CCS response in Scenario 1. The results indicate that the reliable fender system between the FLNG vessel and the LNGC is of great importance for guarantying the safety of side-by-side offloading operations, especially for frequent collision events like Scenario 2.

Scenario 1
The internal energy of the primary structural members in the FLNG side structures that resist collisions is shown in Fig. 10. The striking energy is primarily absorbed by the platform decks and outer shell plates. The slight fluctuations in the internal energy of the web frames are primarily caused by their buckling behaviors. Every moment when the web frames are buckling, the resistance capability of the web frames is weakened. Correspondingly, the growth rate of the energy dissipation slows down (see the web frames curve in Fig. 10). As a result, the energy dissipated by the outer shell increases more rapidly (see the out shell plates curve in Fig. 10). No structural elements rupture during the collision phase. By the end of the collision, plastic strain leads to a permanent deformation in the outer shell and platform decks. Although there is no plastic strain in the inner hull, the permanent deformation of the other structures makes the inner hull no longer flat, which may bring security risks to the operation of the FLNG system.

Scenario 2
The energy dissipation of the primary structural members in the FLNG side structures that resist collisions is shown in Fig. 11. The web frames, platform decks and outer shell plates play an important role in energy absorption. The web frames are especially vital because they rapidly increase the energy absorbing capacity. Moderate buckling occurs primarily in the web frames and platform decks, which is shown in Fig. 12. A small number of ruptured elements are concentrated on the web frames and platform decks, while the outer shell has not ruptured until the termination time.

Summary of structural performance of web frames
The arrangement of web frames at every frame is the key feature of FLNG side structures, which requires a special summary, as shown in Table 9. Obviously, the web frames absorb a large portion of striking energy in Scenario 2. A larger penetration and a larger number of affected web frames usually correspond to larger deformation and energy proportion of web frames. So web frames play a dominating role in the energy dissipation, especially in Scenario 2, which may be the most probable collision scenario.

Scenario 1
The energy dissipation-penetration curves and forcepenetration curves for Case 1-2 are shown in Fig. 13. Generally, the energy dissipation-penetration curves increase continuously, but the force-penetration curves exhibit strong nonlinearity. There is a pronounced initial peak of 549 MN in force-penetration curve. In the latter phase, the curves increase and decrease more than once, with each drop corresponding to the buckling of structures, particularly the platform decks and web frames. The collision force largely de-pends on the contact area, in other words, the number of structural components involved in resisting collisions. However, the cases in Scenario 1 have a very large initial contact area and a limited increase in the contact area in the latter process. At the same time, the force is initially affected by the shock waves. The production of shock waves is accompanied by a sharp increase in collision force. The contact discontinuity disappears after their contact, and thus the sudden decrease in the collision force is caused by weakening the shock wave effects. In fact, although the collision force is very large in Scenario 1, the pressure per unit area is small due to the large contact area.

Scenario 2
The energy dissipation-penetration curves and forcepenetration curves for Case2-2 are shown in Fig. 14. All of the curves have a general increasing trend. During the entire collision, some structural components fall into failure, which affects the growth rate of the collision force. Nevertheless, no pronounced decreases in the curves are found. The initial contact area is not large in Scenario 2, and the curvature of the LNGC bow shoulder is very small. Consequently, the contact area can increase rapidly in a short period of time, and many more structures participate in resisting collisions as the penetration increases. Comparatively, there are much fewer fractured elements. Thus, the influence of these elements is nearly invisible in the force-penetration curves.

Conclusions
In this study, FLNG-LNGC collision scenarios are simulated using the LS-DYNA code, with an emphasis on investigating of crashworthiness of the CCS in the FLNG cargo tanks. The structural behavior of the primary structural components in the FLNG side structures is determined from these simulations. Based on the above analysis, the following conclusions are drawn.
(1) The deformation of the insulation boxes is too small to cause buckling in collision scenarios. Therefore, the strength failure of the insulation boxes can be defined as the limit state for FLNG-LNGC collisions in the later study.
(2) In the parallel collision scenario (Scenario 1), a sudden contact with a large initial area can produce large shock waves, which generate peak stress of 37.27 MPa (close to 40 MPa) in CCS and peak collision force 549 MN.
(3) The stress of the insulation boxes is closely related to the deformation of the inner hull deformation, with approximately 160 mm inner hull deformation corresponding to the limit state of 40 MPa.
(4) The web frames can absorb over 35% of the collision energy. By adjusting the inner deformation and separating side tanks, the web fames can significantly enhance the crashworthiness of FLNG side structures and play a protective role for the FLNG vessel. The design of web frame spa-  Fig. 11. Energy dissipation of primary structural members in Case2-2. cing is of crucial importance in the FLNG structural designs.
(5) According to the limit state, the studied FLNG vessel is safe in Scenario 1. LNG leakage for the CCS failure is the main threat for the FLNG in Scenario 2. This finding proves the necessity of reliable fenders between the FLNG vessel and the LNGC and a mechanism for controlling relative motions.
Finally, the research in this paper only addresses single collision conditions. Repeated collisions may cause some other problems worth investigating, which will be analyzed in future researches.