Ultimate Compressive Strength of Steel Stiffened-Plate Structures Triggered by Brittle Fracture under Cryogenic Conditions

This chapter presents a practical method to investigate the effects of brittle fracture on the ultimate compressive strength of steel stiffened-plate structures under cryogenic conditions. Computational models are developed to analyse the ultimate compressive strength of steel stiffened-plate structures, triggered by brittle fracture, under cryogenic condition. A phenomenological form of the material model for the high-strength steel at cryogenic condition is proposed, that takes into account the Bauschinger effect, and implemented into a nonlinear finite element solver (LS-DYNA). Comparison between computational predictions and experimental measurements is made for the ultimate compressive strength response of a full-scale steel stiffened-plate structure, showing a good agreement between them.


Introduction
Steel stiffened panels are used in naval, offshore, mechanical and civil engineering structures as their primary strength sub-structures. Occasionally, they are exposed to cryogenic conditions, e.g., due to the unwanted release of liquefied gas such as LNG (liquefied natural gas) or liquefied hydrogen as discussed in Paik et al. [1].
The ultimate strength is a primary criterion for the design of steel stiffened-plate structures [2][3][4][5][6], and it is essential to characterise the effects of cryogenic condition on the ultimate strength of such structures. The authors of this chapter have previously conducted collapse tests on full-scale steel stiffened-plate structures under axial-compressive loading at room temperature [1], at cryogenic condition [7] and at -80°C [8]. This chapter is part of a sequel to investigate the brittle fracture of steel stiffened-plate structures under cryogenic conditions. The tested structures were designed from a reference plate panel in an as-built containership carrying 1,900 TEU (twenty-foot equivalent units). They were fabricated at a shipyard using exactly the same welding technology as used in today's shipbuilding industry [9].
Structural fracture modes are classified into three groups [3,4]: rupture, ductile fracture and brittle fracture. Material rupture occurs when failure occurs by cracking associated with necking localisation during large plastic flow. If the strain at which a material fractures is small, with very little ductility, it is a brittle fracture. An intermediate fracture mode between rupture and brittle fracture is called ductile fracture with partial ductility. Fracture behaviour of ductile materials such as carbon steels is quite different from that of inherently brittle materials. Ductile materials generally exhibit slow stable crack growth during crack extension, but they can show a similar behaviour to brittle materials at specific environments such as very low temperatures or lower than the ductile-to-brittle fracture transition temperatures (DBTT) and/or impact loading. It is also recognised that the Bauschinger effect of materials cannot be neglected at sub-zero temperatures and cryogenic condition as the material behaviour in compression is distinct from that in tension [3,10,11]. Figure 1 illustrates the transition of the ductile-to-brittle behaviour for structural steel.
To compute the failure behaviour of structures at cold (sub-zero) temperatures (or higher than the ductile-to-brittle fracture transition temperatures), constitutive equations of materials have been proposed in the literature [12][13][14][15]. It is recognised that most of previous studies are associated with predominantly ductile behaviour or at least with partial ductility, but studies applicable to entirely brittle fracture at cryogenic condition are lacking. It is also recognised that the approaches using the constitutive equations are not always practical for the ultimate strength analysis of supersized structures because they are too complex to apply for the problem.
This chapter presents a method for computing the ultimate compressive strength of steel stiffened-plate structures by nonlinear finite element method (NLFEM) using the multi-physics software package LS-DYNA implicit code. Mechanical properties of high-strength steel with grade AH32 used for fabricating the tested structure were obtained from tension and compression tests at low temperatures and cryogenic condition [7,8], and a phenomenological relation of engineering stress versus engineering strain of the material was formulated. The material model is implemented into the LS-DYNA implicit code. To demonstrate the validity of the computational model, the NLFEM is compared to experimental results from a full-scale physical test.

Literature survey on structural behaviour at cold (sub-zero) temperatures
A number of studies in modelling of material behaviour for structural steels at cold (sub-zero) temperatures are available in the literature. Most of the studies dealt Transition of the ductile-to-brittle behaviour for structural steel [5].
It is recognised that structural steel behaviour is predominantly ductile at temperatures higher than the temperature of the ductile-to-brittle fracture transition, as shown in Figure 1. As the temperature decreases approaching cryogenic condition, the material behaves predominantly in a brittle manner with partial or no ductility [33][34][35][36][37][38][39][40]. Majzoobi et al. [41] observed that the ductile-to-brittle fracture transition of carbon steel occurs at about -80°C, and the material behaviour becomes entirely brittle at -196°C.
Although there are considerable uncertainties associated with the ductile-tobrittle fracture transition temperature (DBTT), a number of evidences for brittle fracture behaviour of steel structures at cryogenic condition have been seen in the literature, depending on the type of materials and loading conditions (e.g., quasistatic or impact), among other factors. Crushing testing of steel tubes under quasistatic loads at -60°C [42,43] showed ductile fracture, as shown in Figure 2. Dropped-object impact testing of steel stiffened plate panels at -60°C [44] showed brittle fracture, as shown in Figure 3. Full-scale collapse testing of a steel stiffenedplate structure under axial-compressive loading showed that the ultimate strength was reached by a trigger of brittle fracture [1], as shown in Figures 4 and 5. At room temperature, the structures reached the ultimate strength by flexuraltorsional buckling [1], but brittle fracture triggered the global failure at cryogenic conditions [7].
Here, an attempt is made to develop new fracture criteria based on the hypothesis that crack initiates if an equivalent stress exceeds a critical value, to model the fracture phenomenon of high-strength steel (AH32) under cryogenic conditions. Existing material models for the fracture analysis is first reviewed.

Maximum principal stress based fracture criterion
The maximum principal stress-based fracture criterion is the simplest among all fracture criteria. It is useful to predict fracture behaviour of brittle materials under predominantly tensile loads. In this criterion, brittle fracture is expected to occur when the largest principal normal stress reaches the ultimate tensile strength (σ T ) of the material, which is usually obtained from tension tests of coupon specimens. The maximum principal stress-based fracture criterion is expressed as follows: Ductile fracture of a square tube under quasi-static crushing loads at -60°C [42].

Coulomb-Mohr fracture criterion
The Coulomb-Mohr fracture criterion gives reasonably accurate predictions of fracture in brittle materials for which the compressive strength far exceeds the tensile strength, e.g., concrete or cast iron [45]. It is presumed that fracture occurs in a certain stress plane of material when a critical combination of normal stress and shear stress acts on the plane. The linear relation of the combination of critical stresses is given by:   [1,7,8].
where τ is the shear stress, σ is the normal stress, μ and τ i are constants for a given material.

Johnson-Holmquist fracture criterion
The Johnson-Holmquist fracture criterion [46] is useful for modelling brittle materials, e.g., ceramic and glass, over a range of strain rates. It is one of the most widely used models in dealing with the ballistic impact on ceramics, which is expressible as follows: where σ * i is the uniaxial failure strength of intact material, see Eq. (4), σ * f is the uniaxial failure strength of completely fractured material, see Eq. (5), and D is a damage accumulation variable, see Eq. (6).
where A, B, C, m, n are material constants, t is time, ε p is the inelastic strain, and ε f is the plastic strain to fracture. The asterisk indicates a normalised quantity, where the quantities of each variable are defined as follows: where σ * and p * are the stresses normalised by the stress at the Hugoniot elastic limit, and T * is the tensile hydrostatic pressure normalised by the pressure at the Hugoniot elastic limit.
The Johnson-Holmquist damage model was modified by Deshpande and Evans [47] and Bhat et al. [48], where it is considered that the propagation of an initial crack is a function of the stress state, the fracture toughness and the flaw characteristics.

Proposed fracture criteria
The yield and ultimate tensile strengths of structural steels tend to increase with a decrease in the temperature [49], and subsequently the maximum load-carrying capacity (ultimate strength) of steel structures at cold temperatures is greater than that at room temperature [43,50]. Figure 6 shows a schematic of ductile and brittle failure behaviour, where the brittle fracture-induced ultimate strength P u2 at cryogenic condition is greater than the ductile collapse-induced P u1 at room temperature. However, the post-ultimate strength behaviour becomes very unstable if brittle fracture triggers the structural collapse at cryogenic condition. In this case, the strain energy absorption capability of structures can be more useful than the ultimate strength itself in terms of the structural safety assessment as it is obtained by integrating the area below the load-displacement curve until or after the ultimate strength is reached. The absorbed energy E 3 at cryogenic condition can be smaller than E 1 or E 2 at room temperature or a temperature higher than the DBTT (ductile-to-brittle fracture transition temperature). For this purpose, the entire behaviour of structural collapse involving brittle fracture at cryogenic condition must be quantified efficiently and accurately.

Fracture criteria
A practical model is proposed for carbon steels which can be used for the ultimate strength analysis triggered by brittle fracture at cryogenic condition or in the region of ductile-to-brittle fracture transition. An elastic-perfectly plastic material model without the strain-hardening effect is used similar to a typical application at room temperature (20°C). However, the material behaviour in compression is different at low (sub-zero) temperatures or cryogenic condition from that in tension as the Bauschinger effect plays a role. However, the Bauschinger effect is usually neglected at room temperature with In the present model, it is hypothesised that brittle fracture occurs if the equivalent stress (σ eq ) reaches a fracture stress which is defined as the yield strength of material at the corresponding temperature, which can be expressed as follows: (1) In tension: (2) In compression: where σ eq is the equivalent stress, σ FT and σ YT are the fracture or yield stresses in tension at cryogenic condition or in the region of ductile-to-brittle fracture transition (which depends on types of materials), σ FC and σ YC are the fracture or yield stresses in compression at cryogenic condition or in the region of ductile-to-brittle fracture transition, γ t and γ c are test constants for a given steel in tension or compression, which may depend on various sources of parameters including chemical composition (grade), temperature and strain rate.
In Eqs. (6) and (7), σ eq can be calculated as a function of principal stresses by the von Mises stress [51] as follows: For plane stress state, σ eq can be simplified as follows:

Formulation of the engineering stress-engineering strain relations
The relations of the engineering stress versus engineering strain can be formulated following the fracture criterion defined in Section 3.1. If the steel temperature, T s , is above the ductile-to-brittle fracture transition temperature (DBTT), i.e., T s > DBTT, the material behaves according to the ductile region. In this case, the stressstrain relation in tension is expressed as follows: where σ is the engineering stress, ε is the engineering strain, E is the elastic modulus, ε Yt is the yield strain in tension, and ε ft is the fracture strain in compression. Figure 7 shows a schematic view of the proposed model of the engineering stress-engineering stress relation with full or partial ductility at room temperature or low temperatures.
In compression, the following equation similar to Eq. (13) is obtained.
where ε Yt is the yield strain in tension and ε fc is the fracture strain in compression.
If T s ≤ DBTT or the steel temperature is lower than DBTT, the material behaves according to the brittle region. In this case, the stress-strain relation in tension is expressed as follows: In compression, a similar equation to Eq. (14) is obtained as follows: In the region of entire brittle fracture, i.e., with T s ≤ DBTT, Figure 7 can be redrawn as shown in Figure 8 when elongation can be neglected after reaching the yield stress in tension or compression as the material exhibits brittle fracture immediately.

Nonlinear finite element method modelling
The proposed material model is implemented into the LS-DYNA implicit solver with 4-node shell elements (element formulation 16) for the ultimate compressive strength analysis of a structure tested at cryogenic condition as shown in Figures 4 and 5 [7]. Table 1 with Figure 9 shows the dimensions of the tested structure. The nonlinear ultimate compressive strength analysis is simulated in a quasi-static analysis using the LS-DYNA implicit solver. The 4-node shell elements are used to model plating, stiffeners and transverse frames. In order to ensure sufficient resolution in the mesh size, a convergence study was performed by varying the element size following a typical approach as described in Paik [4]. The resulting convergence study provided the element size of 40 mm Â 40 mm which was chosen to obtain sufficiently accurate results while minimising the computational cost. The thermal shrinkage effects for steel at low temperatures were not considered. Figure 10 shows the FE model of the tested structure. Figure 11 shows the loading and boundary conditions which were modelled as much as close to the tested structure, where unloaded edges were kept straight and loaded edges were entirely fixed. Only the middle bay of the tested structure was exposed to the cryogenic condition as shown in Figure 12. Table 2 summarises the measured data of steel temperatures during the collapse testing. For details of Table 2, see Paik et al. [7].
In room temperature, the mechanical properties of steel in compression are typically defined in the same manner as in tension without considering. However,   the Bauschinger effect cannot be neglected at low (sub-zero) temperatures and cryogenic condition [3,10,11]. To define the mechanical properties of AH32 steel at different temperatures (20°C, À80°C, À130°C and À 160°C), material tests in tension and compression were conducted. Details of these test data are presented in separate papers [7,8]. Tables 3 and 4 summarize the test data for the mechanical properties of the AH32 steel. It is found that the yield stress of steel in tension or compression increases as the temperature decreases, while the fracture strain in tension decreases with decrease in the temperature. The elastic modulus of steel Figure 11. The loading and boundary conditions applied to the FE model.

Figure 12.
Middle bay of the structure exposed to cryogenic condition.  Table 3.
Mechanical properties of AH32 steel at room and low temperatures in tension [7,8].
remains unchanged regardless of sub-zero temperatures. This chapter focuses on the ultimate strength of steel structures under monotonically applied compressive loads, but fatigue crack resistance at sub-zero temperatures must be associated with microstructural characteristics which are closely related to low-temperature impact toughness of steel [52,53].
In the present case study, an elastic-perfectly plastic material model was applied without considering the strain-hardening effect. To implement the material model, *MAT_PLASTICITY_COMPRESSION_TENSION, MAT124 in LS-DYNA was used as it is an isotropic elastic-plastic material which can distinguish material behaviour in tension and compression. The von Mises yield criterion was applied using MAT124. Tension or compression was determined by the sign of the mean stress (hydrostatic stress). A positive sign which means a negative pressure is indicative of tension, or a negative sign is indicative of compression. The mean stress, σ mean can be expressed as follows: where σ x , σ y and σ z are the stress components in the x, y and z directions, respectively.
Majzoobi et al. [41] observed that the ductile-brittle transition of steel occurs at À80°C, and the material behaviour of steel is completely brittle at À196°C. With Figure 12 and Table 2, the average steel temperatures of plating and web of stiffeners in the middle bay of the tested structure were À160°C and À130°C, respectively. Therefore, the plating and web of stiffeners in the middle bay of the tested structure were modelled using the engineering stress-engineering stress relation of Eqs. (12) and (13). The rest of structural members in ductile region (above À80°C) were modelled using the engineering stress-engineering stress relation of Eqs. (12) and (13). See Paik et al. [7] for details.
Three types of fabrication-related initial deformations are considered as shown in Figure 13. The measurement data of welding-induced initial deformations for the tested structure [54] as shown in Figure 14 was directly applied to the FE model.
The initial deformations of the tested structure were formulated so as to make easier implementation into the FE model as shown in Figure 15.  Table 4.
Mechanical properties of AH32 steel at room and low temperatures in compression [7,8].

Figure 13.
Three types of welding-induced initial deformations in a stiffened plate structure.

Figure 14.
Measured and idealised deformations of the tested structure due to fabrication by welding.

Figure 15.
Welding-induced initial deformations applied to the FE model (with an amplification factor of 100 for plating and column-type, and 20 for sideways initial deformations).
Initial deflection of plating: Column-type initial deformation of stiffener: Sideways initial deformation of stiffener: where z is the coordinate in the direction of stiffener web height, and h w is the stiffener web height.
Biaxial residual stresses developed in the plating of the tested structure between the support members because the welding was conducted in both the longitudinal and the transverse directions to attach the longitudinal stiffeners and the transverse frames. Measurement data of the fabrication-induced residual stresses in the tested structure [55] was also directly applied to the FE model although the biaxial residual stress distributions were idealised as shown in Figure 16 with the measurement data indicated in Table 5.
Stress concentration in structural details or fillet weld toe locations happens due to geometrical discontinuity, and it is a critical factor that must be considered for fatigue limit state analysis [3,56]. Figure 17 shows an example of the effective plastic strain distribution which was obtained from the FE analysis of the ultimate compressive strength of the tested structure. It is obvious from Figure 17 that the effective plastic strain is comparatively large along the weld lines between plating and stiffeners. For the ultimate strength analysis in ductile region, e.g., at room temperature, the stress concentration at the fillet weld toes is usually ignored.
For brittle fracture analysis at sub-zero or cryogenic condition, however, the effects of stress concentration cannot be neglected [4,44]. This is because the weld toes can reach the yield condition earlier, leading to local brittle fracture which can trigger the ultimate limit states at cryogenic condition. Therefore, the nonlinearity at weld toes along the fillet weld lines needs to take into account in the FE modelling.
One of approaches is to model the weld toes directly in the FE model using shell elements with specific properties of weld metal. Figure 18 shows a schematic of  Table 5.
Measured data of the biaxial residual stresses in the plating of the tested structure.  modelling the weld toes using shell elements along the plate-stiffener junction. A similar approach was used to model weld toes by Kim et al. [44] and Nam et al. [35]. The tested structure was fabricated using flux-cored arc welding (FCAW) method and the consumable was CSF-71S, and the mechanical properties of weld metal with the CSF-71S are presented in Table 6. As such, the weld metal was modelled using the engineering stress-engineering strain relation of Eqs. (12) and (13). It is assumed that the yield strength of the weld metal at À160°C increases linearly in the same proportion as the steel ( Table 6).  When only the ductile material model was applied for all structure members without considering brittle fracture, the FE analysis overestimates the ultimate strength significantly. As the yield strength of the material at cold temperature is greater than that at room temperature, the ultimate strength becomes much larger as far as brittle fracture is not allowed to happen. On the other hand, the ultimate strength obtained from the FE analysis with the brittle material model is in good agreement with the test results. Figure 20 compares the ultimate strength behaviour with or without the weld elements along the plate-stiffener junctions. It is seen from Figure 20 that the weld metal model increased the ultimate strength by 4.7%. This is due to the mechanical properties of weld metal which are larger than those of base metal ( Table 9). Figure 21 shows the deformed shape at the ultimate limit state of the tested structure obtained from the FE analysis without brittle fracture model (with only ductile material model). It is seen from Figure 21 that the tested structure reached the ultimate limit state by tripping mode of stiffeners (without brittle fracture) which is similar to the collapse mode at room temperature [1]. However, the brittle fracture model represents brittle fracture behaviour which triggered the ultimate strength as shown in Figure 22, where deformed and fracture shapes of the test structures are compared between physical testing and FE analysis.  Table 9.

Computed results and discussion
Effect of weld metal on the ultimate strength behaviour.

Figure 21.
Deformed shape of the tested structure at the ultimate limit state obtained from FE analysis only with ductile material model.

Concluding remarks
This chapter presents a practical method to compute the ultimate compressive strength of steel stiffened-plate structures at cryogenic condition which is triggered by brittle fracture. Case studies were carried out using the method, and the following conclusions were obtained together with modelling recommendations for NLFEM simulations.
1. A useful material model was formulated to analyse the brittle fracture behaviour of structural steels at cryogenic condition, where the Bauschinger effect was taken into account as the material properties in compression are distinct from those in tension 2. An elastic-perfectly plastic material model was applied without considering strain-hardening effect.
3. The developed material model was implemented into the LS-DYNA implicit code with *MAT_PLASTICITY_COMPRESSION_TENSION, MAT124.
4. Weld elements which are the same type of shell elements but with specific properties of material were introduced to model weld metal (weld toes) at the plate-stiffener junctions where stress concentrations develop. As the mechanical properties of weld metal are typically greater than those of base metal, the ultimate strength usually becomes larger with weld elements at weld toes.

5.
Comparisons were made between the test results and the FE computations on a full-scale steel stiffened-plate structure at cryogenic condition. It is confirmed that the nonlinear FE analysis with the proposed material model, and the weld element model, gives a reasonably good solution of the ultimate compressive strength behaviour for steel stiffened plate structures at cryogenic condition.