1. Introduction
The liquid-filled cylindrical shell structure is a classic engineering structure with excellent performance, widely used in fields such as underwater vehicles and marine pipelines [
1,
2,
3,
4,
5,
6,
7]. Influenced by internal fluid impact, corrosion, and fatigue, cracks can easily develop on the surface or inside of cylindrical shells, leading to failure. Additionally, once fluid-filled cylindrical shells like high-pressure pipelines are damaged by cracks, the cost of repair or replacement is high [
8,
9,
10,
11,
12,
13]. Therefore, studying the crack propagation and failure processes of damaged fluid-filled cylindrical shells under high-pressure conditions is crucial in optimizing design and enhancing safety.
The immersed particle method (IPM) is a method often used to simulate structures that fracture under impact loading [
14,
15,
16,
17,
18]. It uses a meshless particle method for fluids and structures and is suitable for dealing with complex FSI problems including crack extension and penetration [
19,
20]. Meanwhile, many researchers have used the crack particles method (CPM) to simulate dynamic fracture of metals [
21]. CPM does not require an explicit representation of the crack topology, which makes it suitable for dynamic fracture and fragmentation problems [
22]. However, the particle approach requires a large amount of computational resources and is not favorable for dealing with the multivariate crack path problem. In addition, the dual-horizon peridynamics method (DH-PDM) has been gaining attention in recent years. It can be used to analyze the material fracture problem by introducing two independent field-of-view concepts that naturally include the changing field-of-view dimensions, and completely solving the “ghost force” problem in the traditional peridynamics [
23]. However, DH-PDM, as a relatively new technique, lacks sufficient experimental data to support its applicability in material nonlinearity. In contrast, the extended finite element method (XFEM) has an advantage in computational efficiency by adapting to crack extension through local remeshing. Additionally, it captures the singularities at the crack tip through local enrichment functions, which is beneficial for modeling nonlinear patterns of complex cracks [
24]. XFEM does not require additional mesh encryption at the crack tip. Although XFEM has been widely used to solve discontinuous displacement field problems, its performance in fluid-filled cylindrical shells has not been extensively studied [
25,
26,
27,
28,
29].
Material damage criteria, including crack initiation criteria and damage evolution criteria, are used to study damage formation and evolution in fluid-filled cylindrical shell models [
30]. The damage initiation criteria suitable for XFEM include the maximum principal stress (Maxps) or strain (Maxpe), maximum nominal stress or strain, sub-nominal stress or strain, and user-defined damage initiation criteria [
31].
Selecting appropriate damage parameters is crucial for improving model simulation accuracy [
32]. Liu [
33] studied the toughness crack propagation of X80 pipeline steel and found the maximum principal strain criterion more suitable for numerical analysis. Motamedi and others [
34] used XFEM to analyze the fracture performance of polysulfone and unidirectional glass fiber composites, successfully predicting crack propagation paths under different loads. Okodi [
35] used XFEM with maximum principal strain and fracture energy as damage parameters to analyze crack propagation in pipelines. Compared to the actual material strength, the maximum principal stress resulted in higher predictions of crack propagation. Ameli and others [
36] studied the critical maximum principal stress values of X42 pipelines, which aligned well with experimental data. However, the crack initiation stress was about 4.3 times the yield stress of X42, suggesting that a strain-based damage criterion might be more suitable for predicting toughness fractures. Under high-pressure conditions, high-strength materials in fluid-filled cylindrical shells often undergo ductile rather than brittle fracture. Caleyron [
37] used the SPH method to establish a fluid-structure interaction model for a damaged fluid-filled cylindrical thin-shell structure, predicting the failure modes of fluid-filled tank-like containers under impact loads and analyzing the failure forms of the damaged fluid-filled cylindrical thin-shell structure. Ozdemir [
38] used a meshless method to study the buckling behavior of cylindrical shells with penetrating cracks, examining the effects of through-cracks on the buckling coefficients and modal shapes of cylindrical shells, and modeling fluid-structure interaction for cylindrical shells with through-cracks to compare results, validating the effectiveness of the meshless method in this research.
In the current research on damaged liquid-filled cylindrical shell structures, the maximum principal strain criterion used does not consider the material’s nature and microstructure, and it only evaluates damage based on local strain, which cannot accurately reflect the true fracture behavior of the material. Therefore, this research studied the crack propagation and failure process of damaged liquid-filled cylindrical shells. It employed the XFEM method with the maximum principal stress criterion (Maxps) as the damage parameter to analyze crack propagation and failure of liquid-filled cylindrical shells under high-pressure conditions, and investigated the impact of crack parameters on structural failure. Additionally, based on fluid-structure interaction theory, a dynamic model for damaged liquid-filled cylindrical shells was established to analyze the crack propagation process under the influence of fluid, as well as the changes in fluid pressure and flow state.
2. Mathematical Model
For the purpose of theoretical modeling, the following assumptions need to be made: the shell material is isotropic; the overall deformation is linear-elastic; the effects of rotational inertia, in-plane inertia, and transverse shear deformation of thin-walled cylindrical shells are neglected; and the internal fluid is incompressible.
2.1. Element Decomposition Method
The element decomposition method divides the overall solution domain into several independent discrete units, processes the approximation function of each unit, and then integrates all the discrete units into an accurate overall shape function [
39,
40]. Each discrete unit in the solution domain contains a unit decomposition function, and
of all discrete units in the solution domain needs to satisfy [
41]:
In the above equation, i is each discrete unit in the solution domain; is the unit decomposition function of each discrete unit. The unit decomposition method requires that the sum of the unit decomposition function of each discrete unit is always 1.
The unit decomposition function
in the above equation has different functional forms. The form function used by the finite element method is one of the discrete unit decomposition functions. Through the form function, the finite element method is able to calculate the displacement at any position in the discrete unit. The displacement expression of the finite element method is shown as follows [
42],
where
u denotes the displacement field inside the discrete cell;
i is the cell node;
is the finite element shape function of each cell node; and
is the displacement of each cell node.
2.2. Extended Finite Element Method Expansion
The extended finite element method is an improved method of finite element method for discontinuous problems [
33]. Equation (2) is the displacement solution equation of the traditional finite element method. The extended finite element method introduces an expansion term
on this basis, which is specifically used to describe the displacement field of discontinuous surfaces. The expression for solving the discontinuous displacement field
at any place in the solution domain based on the extended finite element method is [
43]:
where
is the finite element shape function of each unit node;
is the displacement of each unit node, and
is an expansion function added to describe the intermittent discontinuity field. Further, a cell decomposition of the expansion term
, Equation (3), can be expressed as [
44]:
Equation (4) is the generalized displacement field description form of the extended finite element method; the former term is the displacement field solution of the traditional finite element, and the latter term is the expansion term of the extended finite element for describing the discontinuous field. Where j denotes the nodal cells of the discontinuous field, denotes the total number of cells added to the interrupted discontinuous field; denotes the enrichment function that corrects the displacement field of the discontinuous cell; and denotes the displacement field of the discontinuous cell.
The unit classification of the crack-containing damage model based on the extended finite element method is shown in
Figure 1. The black solid curve represents the expanding crack; the white area is the ordinary continuous unit; the orange area is the crack unit; and the red area is the crack tip unit. Based on this, the displacement field expression of the extended finite element method for the crack extension problem can be further rewritten [
45]:
In the above equation, is the conventional finite element method displacement field function, i.e., the ordinary continuous cell displacement field; denotes the cracked cell displacement field; and denotes the cracked tip cell displacement field.
Further, the displacement field anywhere in the solution domain of the crack extension problem can be expressed as [
46]:
In Equation (6),
represent the ordinary continuous unit, crack unit, and crack tip unit, respectively;
represent the total number of ordinary continuous units, crack units, and crack tip units, respectively;
represent the shape function of ordinary continuous unit
, crack unit
, and crack tip unit
, respectively;
is a step function for the characterization of the intermittency of the cracks, and the specific form is shown in Equation (7).
is a crack tip enhancement function, and the specific form is shown in Equation (8).
In Equation (7), is the integration point; is the closest point of the crack to the integration point; is the shortest distance of the integration point from the crack. In Equation (8), expressing the crack tip enhancement function, and are the polar coordinates of the crack tip.
2.3. Maximum Principal Stress Criterion (Maxps)
The material is subjected to different stresses at different locations, and if the stress exceeds the compressive strength of the material at a certain location, the structure will be damaged, and this location is the crack damage initiation location [
47]. Usually, the maximum principal stress method uses the compressive strength as the critical maximum principal stress. The expression for the maximum principal stress law to determine the damage initiation is shown as follows [
48],
where
denotes the critical maximum principal stress;
is a relation of the value of the maximum principal stress of the material
, the value of which is related to the direction of the maximum stress.
is shown as follows [
49].
Equation (10) indicates that the maximum stress value
is valid when the maximum stress in the material is positive, otherwise it is 0. This indicates that the crack initiation location will not start to expand when only stressed.
Cracks in damaged fluid-filled cylindrical shells satisfy the maximum principal stress criterion and begin to expand when the result of the calculation in Equation (9), , satisfies Equation (11). is a correction factor, and in general it is sufficient to take 0.05.
The displacement-based crack damage evolution criterion predicts the crack damage evolution by observing the amount of material displacement. It is categorized into two types: linear and exponential evolution; Equations (12) and (13) show the expressions for linear and exponential evolution, respectively.
where
um is the displacement of the damaged fluid-filled cylindrical shell unit in the current state;
u0 is the initial displacement of the unit when the crack in the damaged fluid-filled cylindrical shell starts to expand;
umax is the maximum displacement of the damaged fluid-filled cylindrical shell in the process of crack expansion, and
a denotes a dimensionless parameter.
2.4. The Equation of Fluid Motion within a Crack in a Damaged Fluid-Filled Cylindrical Shell
The fluid motion within a crack in a damaged fluid-filled cylindrical shell is depicted in
Figure 2. As the fluid within the cylindrical shell moves radially toward the crack, the direction of the fluid motion is parallel to the plane of the cylindrical shell’s crack, constituting tangential flow. When the fluid enters the crack, that is, into the damage-enriched elements, the flow transitions to a normal direction perpendicular to the crack plane. The tangential flow within the crack of the damaged fluid-filled cylindrical shell is described by the Newtonian flow formula, as shown in Equation 14; the formulas for calculating the normal flow within the crack are presented in Equations (16) and (17).
In the above expression,
denotes the volumetric flow rate of the fluid within the enriched element;
represents the pressure gradient across the crack plane of the cylindrical shell;
signifies the opening displacement of the cylindrical shell’s crack plane;
indicates the permeability coefficient, which is defined in the Newtonian flow formula as shown as follows [
50].
In the equation,
represents the fluid viscosity.
In the equation, and represent the volumetric flow rates of the fluid at the upper and lower surfaces of the crack plane after entering the crack in the cylindrical shell, respectively; and denote the filtration coefficients at the upper and lower surfaces of the crack plane, respectively; and indicate the pore pressures at the upper and lower surfaces, respectively; signifies the fluid pressure within the crack of the cylindrical shell.
2.5. The Fluid-Structure Interaction Control Equation for a Damaged Fluid-Filled Cylindrical Shell
The core concept of a dynamic model that encompasses both fluid and solid domains is the conservation of relevant physical parameters at the interface where the fluid and solid meet. For a damaged fluid-filled cylindrical shell, conservation of stress, displacement, temperature, and heat flux at the interface between the cylindrical shell region and the fluid region on the inner wall of the shell enables the coupling of the fluid and solid domains [
51,
52].
In Equation (16), and respectively represent the stresses at the interface between the fluid domain and the solid domain; and respectively denote the temperatures at the interface between the fluid domain and the solid domain; and respectively indicate the displacements at the interface between the fluid domain and the solid domain; and respectively signify the heat fluxes at the interface between the fluid domain and the solid domain.
2.6. The Solution Method for Fluid-Structure Interaction in a Damaged Fluid-Filled Cylindrical Shell
The computational methods for fluid-structure interaction (FSI) solutions are divided into direct and indirect FSI approaches. The direct FSI approach provides an intuitive and accurate method for solving FSI problems by solving the fluid and solid domains simultaneously within a single solver. The control equations for the direct FSI approach are presented as follows [
53].
In the equation, represents the system matrix for the fluid domain; represents the system matrix for the solid domain; and denote the coupling matrices between the fluid and solid domains; indicates the time step for analysis; and respectively represent the solution values for the fluid and solid domains; and respectively denote the external forces acting on the fluid and solid domains. As the solutions for the fluid and solid domains are solved within the same control equation, their solution processes are synchronized, eliminating computational errors that might arise from asynchronous processing.
4. Results and Discussion
4.1. Effect of Crack Initial Angle on the Paths of Crack Propagation in Damaged Fluid-Filled Cylindrical Shells
The initial crack was a planar straight edge crack with a crack width of 10 mm and a depth of 1.5 mm. The crack propagation simulations of the damaged fluid-filled cylindrical shell were performed by respectively setting the initial crack angles to 0°, 15°, 30°, 45°, 60°, and 75°.
Figure 8 shows the simulation results of crack propagation at each initial crack angle. It can be observed from the figure that, regardless of changes in the initial crack angle, after a period of propagation, the crack consistently extended along the axis of the damaged fluid-filled cylindrical shell.
The results of crack propagation on the inner wall and the final morphology of the crack surfaces at different initial crack angles are shown in
Figure 9 and
Figure 10. Observation of the crack propagation on the inner wall under different initial crack angles and the final results of crack propagation on the outer wall shows that they were basically similar. The crack surfaces were generally flat, although in some cases the propagation speed of the inner wall cracks was slightly faster than that of the outer wall. However, the overall trend remained consistent.
As indicated by the above figures, the initial crack angle of crack propagation influenced the direction of cracks in the damaged fluid-filled cylindrical shell during the initial stages of propagation. The direction of crack growth gradually shifted from the initial crack angle towards the axis of the cylindrical shell. However, after a brief period, the crack invariably extended outward along the axis of the cylinder. During the phase before the crack penetrated the outer wall of the cylindrical shell, the crack propagation speed was relatively slow. Once the crack breached the outer wall, the outer wall crack propagated faster, following a similar path to the inner wall crack, and then continued at a consistent speed. Since the inner wall was directly subjected to pressure, the crack propagation path may have advanced a few grid units faster than that on the outer wall. At the end of the propagation, the crack patterns on the inner and outer walls of the damaged fluid-filled cylindrical shell were consistent, and the crack surfaces were smooth.
4.2. Effect of Initial Crack Angle on the Failure Process of Damaged Fluid-Filled Cylindrical Shells
The crack depth in the damaged cylindrical shell was set to be 2.00 mm, and the crack width was set to be 10 mm. The effects of different initial crack angles on the crack penetration velocity in the damaged fluid-filled cylindrical shell were investigated by adjusting the initial crack angles to be 0°, 15°, 30°, 45°, 60°, and 75°.
Figure 11 shows the time-displacement curves of the crack penetration unit in the outer wall of the damaged fluid-filled cylindrical shell model with different initial crack angles.
The time-displacement curves from the damaged fluid-filled cylindrical shell models at different initial crack angles indicate that as the initial crack angle increased, the time of abrupt change in the time-displacement curve of the outer wall units was delayed, and the slope after the abrupt change also gradually decreased. This preliminary observation suggested that the crack penetration speed in the damaged fluid-filled cylindrical shell decreased as the initial crack angle increased. By analyzing the data from the time-displacement curves of crack penetration units at various initial angles, we recorded the times of abrupt changes, the onset of plastic deformation
, and the occurrence of crack penetration
. These findings are plotted and presented in
Table 3 and
Figure 12.
Figure 12 shows that as the initial crack angle increased, the rate of plastic deformation and crack penetration in the damaged fluid-filled cylindrical shell model gradually decreased. Additionally, the difference between the time of plastic deformation of the outer wall units and the time of crack penetration remained consistently around 0.02 s when the internal crack began to expand. When the initial crack angle was 0°, the crack penetrated the cylindrical shell model the fastest. Between 15° and 60°, the slope was smaller and the increase in time was slower; whereas at the extremes of 0° and 75°, the slope was steeper, and the crack penetration time increased rapidly. Due to the internal fluid pressure, cracks always extended toward both ends along the axis of the cylindrical shell. Therefore, when the initial crack angle was 0°, it aligned with the path of crack propagation in the damaged cylindrical shell, resulting in the fastest crack penetration. Conversely, when the initial crack angle was 75°, the direction of the crack differed significantly from the propagation direction, resulting in slower crack penetration.
For fluid-filled cylindrical shell structures, the maximum internal pressure that the structure can withstand is a critical parameter that must be determined to ensure the safe and normal operation of the structural components. The initial crack was set to have a depth of 2.25 mm and a width of 10 mm, represented by a through-thickness straight-edged crack. The maximum bearable internal pressure at the onset of internal crack propagation in the damaged cylindrical shell was defined as
, and the maximum bearable internal pressure at the time of crack penetration through the damaged cylindrical shell was defined as
, and the crack initiation angles were taken to be 0°, 15°, 30°, 45°, 60°, and 75°, respectively. The maximum internal pressures that can be sustained by the damaged fluid-filled cylindrical shells at different initial crack angles are shown in
Table 4 and
Figure 13.
The charts indicate that as the initial crack angle increased, the maximum fluid pressure that the damaged fluid-filled cylindrical shell could withstand also gradually increased. Additionally, the difference between the maximum internal pressure that could be borne at the onset of crack propagation and the maximum internal pressure at the time of crack penetration was approximately 10 MPa. Moreover, the pattern of change in the maximum sustainable internal pressure related to the initial crack angle mirrored the trend observed in the crack penetration speed. When the initial crack angle was in the middle range, the change in maximum bearable pressure was minimal, and the slope of the curve was relatively flat. In contrast, at the extremes, specifically at 0° and 75°, the variation in maximum bearable fluid pressure was more significant. Particularly at an initial crack angle of 0°, the change was most pronounced with a maximum internal pressure decay of about 14.448 MPa, a decay rate of 12.7%, and a decay value of approximately 13.829 MPa, with a decay rate of 13.46%.
Since the path of crack propagation in a damaged fluid-filled cylindrical shell typically extends along the axis of the cylinder towards both ends, cracks are most likely to propagate and cause the shell to rupture when the initial crack direction aligns with the propagation path. Under these conditions, the maximum fluid pressure that the damaged cylindrical shell can withstand reaches its lowest value.
4.3. Analysis of Fluid Flow States in Damaged Fluid-Filled Cylindrical Shells Based on Fluid-Structure Interaction Experiments
Research on fluid domain elements within a dynamic model involves analyzing the experimental fluid in damaged fluid-filled cylindrical shells. This study focused on the maintenance of pressure in the experimental fluid during the crack propagation and failure processes in damaged fluid-filled cylindrical shells, as well as the flow state of the fluid when it bursts through the shell wall upon complete crack penetration.
The process of experimental fluid bursting from a damaged cylindrical shell with axial cracks upon crack penetration is shown in
Figure 14. When the crack penetrated, it first initiated at the center of the crack, causing the experimental fluid to burst out sharply from the penetration point and gradually rise. At this moment, the internal pressure of the experimental fluid within the cylindrical shell was not lost; instead, as the crack path continued to expand, stress concentrated at the crack tip, causing fluid to burst out of the newly formed cracks as shown in
Figure 14c, where the stress in the area near the shell’s crack was significantly higher than in other areas. Subsequently, as the crack continued to expand, a substantial amount of experimental fluid began to burst out, and the damaged fluid-filled cylindrical shell gradually lost its pressure-bearing capability, with the pressure of the experimental fluid gradually dropping to zero. During this process, the initial bursting location of the experimental fluid remained the point of greatest displacement.
Figure 15 shows the displacement of the experimental fluid in the fluid-filled cylindrical shell with axial cracks during the bursting process. It can be seen that the initial burst of experimental fluid occurred at the point of greatest displacement.
The process of experimental fluid bursting from a fluid-filled cylindrical shell with circumferential cracks under an acceleration of 10 m/s
2 is illustrated in
Figure 16. When the crack penetrated, the experimental fluid burst along the initial crack path in a sharp, peak-like manner, as shown in
Figure 16a–c. As the crack continued to expand, vulnerable units appeared on both sides of the circumferential crack. These vulnerable units bent due to the fluid pressure, causing a sudden increase in crack area, which paradoxically led to a decrease in the height of the bursting experimental fluid, as shown in
Figure 16d,e. Subsequently, the cracks formed at the tips of the two circumferential cracks converged into a wider crack due to the fracture of vulnerable units, and the experimental fluid re-emerged in a manner similar to bursting beneath axial cracks, with the center of the crack still showing the greatest displacement of the experimental fluid. After a certain duration of fluid bursting and the formation of a larger crack face in the cylindrical shell, as shown in
Figure 16h,i, the stress on the cylindrical shell was significantly reduced compared to before, resulting in shell failure and loss of pressure-bearing capacity.
After completing the investigation into the changes in the flow state of the experimental fluid within the damaged fluid-filled cylindrical shell as the crack expanded, a study on the changes in fluid pressure inside the shell during the crack expansion process was conducted.
Figure 17 shows the changes in the experimental fluid pressure inside the damaged fluid-filled cylindrical shell over time. In the model of the damaged cylindrical shell, the crack depth was 2.25 mm, and the crack width was 10 mm, and the initial crack angle was set to be 0°. The line graph of the experimental fluid pressure changes in the damaged cylindrical shell shows a gradual increase in internal fluid pressure due to the influx of fluid, as a baffle was placed at the outlet of the damaged cylindrical shell. When the stress at the crack tip on the inner wall of the damaged cylindrical shell exceeded the yield strength, plastic deformation occurred in the units near the crack tip, and the crack began to expand, with the internal experimental fluid pressure at that moment being 91.77 MPa. The experimental fluid pressure continued to increase as the crack expanded to penetrate the outer wall of the cylindrical shell, reaching its maximum value at 102.22 MPa.
When using a dynamics model of a damaged fluid-filled cylindrical shell based on fluid-structure interaction for crack propagation, the fluid pressure inside the shell is extracted from the fluid domain elements, making the calculation of changes in the experimental fluid pressure inside the shell more accurate. The line graph shows that after the fluid pressure inside the shell reached its maximum and the cylindrical shell’s crack penetrated, the fluid did not immediately lose pressure but maintained it for a certain period. Once the crack expanded to a certain extent, the damaged fluid-filled cylindrical shell completely failed, at which point the fluid inside the shell lost pressure.
6. Conclusions
This paper presents a static modeling method for damaged fluid-filled cylindrical shells based on the extended finite element method (XFEM), exploring the impact of different initial crack angles on the crack propagation paths and failure processes of the shells. Additionally, based on fluid-structure interaction theory, a dynamic model of the damaged fluid-filled cylindrical shell was established to capture the crack propagation process under fluid influence. This study also analyzed changes in pressure and flow state during the shell’s failure process and verified these findings experimentally. The main conclusions are summarized as follows:
(1) The initial crack angle influences the direction of crack propagation in the damaged fluid-filled cylindrical shell. Initially, the crack gradually deviates from its original direction towards the axial direction and then extends along the axis. After the crack penetrates the outer wall, it rapidly expands along the path of the inner wall crack and eventually propagates in conjunction with it. The inner wall, subjected to direct pressure, causes the crack path to progress several mesh elements faster than that of the outer wall.
(2) As the initial crack angle increases, the plastic deformation and the speed of crack penetration in the damaged fluid-filled cylindrical shell model gradually decrease. When the initial crack angle is 0°, the crack penetrates the cylindrical shell model the fastest. However, when the initial crack angle is 75°, the difference between the initial and expansion directions of the crack is significant, resulting in a slower penetration speed of the crack through the damaged fluid-filled cylindrical shell.
(3) Cracks typically extend along the axis of the damaged fluid-filled cylindrical shell towards both ends. When the initial direction of the crack aligns with its expansion path, the crack is more likely to extend and penetrate through the shell. In this scenario, the maximum fluid pressure that the damaged fluid-filled cylindrical shell can withstand drops to its lowest value.
(4) Based on fluid-structure interaction theory, a dynamic model of the damaged fluid-filled cylindrical shell was established, studying the changes in internal fluid pressure and the flow state during crack penetration. When the initial crack was an axial crack with a depth of 2.25 mm and a width of 10 mm, the maximum internal fluid pressure in the damaged fluid-filled cylindrical shell was 102.22 MPa. The maximum internal pressure that the shell could withstand, calculated using XFEM under the same conditions, was 99.17 MPa, with an error of 3.07%.