Experimental and Numerical Investigation of Constant-Amplitude Fatigue Performance in Welded Joints of Steel Tubular Flange Connections for Steel Structures

Abstract

Welded joints of tubular flange connections (TFCs) for steel structures are prone to cumulative fatigue breakdown under oscillatory loading regimes. This study investigates the constant-amplitude fatigue performance of these welded connections through combined experimental testing and finite element analysis. Seven tubular flange connection specimens were subjected to constant-amplitude fatigue tests, and the nominal stress range approach was employed to establish S-N curves for the TFC welds, which were then compared with existing design codes. Stress concentration behavior at the weld toe was analyzed using ABAQUS finite element software. Macro- and micro-scale examinations of fatigue fracture surfaces were conducted to elucidate the fatigue crack mechanisms. The results demonstrate an allowable stress range of 82.41 MPa at a 2-million-cycle fatigue strength, exceeding the specifications of current fatigue design codes. The finite element analysis shows that there is a significant stress concentration at the weld toe of the steel tube–flange weld, and the uneven stress distribution in the circumferential direction of the weld makes this position more prone to fatigue failure, which is consistent with the experimental phenomena. The derived fatigue design method for TFCs provides practical guidance for engineering applications.

1. Introduction

Tubular flange connections (TFCs) have been widely adopted in critical engineering applications such as wind turbine towers, petrochemical piping systems, and high-rise steel structures [1,2,3] due to their exceptional load-bearing capacity, detachability, and maintenance convenience [4]. However, during long-term service, these TFCs are subjected to cyclic loading (e.g., periodic vibrations in wind turbines or pressure fluctuations in oil pipelines), which frequently leads to fatigue crack initiation at welded joints and may ultimately result in catastrophic fracture failures. Once such a fracture failure occurs, it will not only cause significant damage to the equipment itself but also pose a great threat to the surrounding environment and the safety of personnel. Hence, the exploration of fatigue characteristics in welded connections of TFCs plays a pivotal role in ensuring structural safety.

Numerous investigations have delved into the fatigue behavior of high-strength bolts in TFCs. Shakeri [3] explored the impact of thread-rolling defects on the high-cycle fatigue life of M30 Grade 10.9 double-end bolts in wind turbine setups. Ni [5] studied the fatigue characteristics of flange-joint high-strength bolts, proposing a fatigue design approach. Yang [6] formulated a time-domain-based analytical framework for assessing the fatigue damage of bolts in monopole communication tower flange connections. Liang [2] centered on high-strength bolts in wind turbine tower flange connections. Specifically, Liang [2]’s work in the wind power sector established a fatigue strength analysis approach considering the effects of external loads on fatigue damage accumulation. Cao [7] employed a multi-scale finite element modeling approach (beam–solid elements) in ABAQUS for an 80 m self-supporting steel chimney, with additional discussion on the impact of bolt preload on fatigue life. Okorn [8] examined how structural irregularities (e.g., 0.03 mm non-parallelism) in a DN40 flange affect the fatigue behavior of pre-tensioned bolts via experimental testing and finite element modeling, finding that even slight geometric deviations—particularly when flange misalignments coincide with eccentric applied loads—significantly amplify fatigue stress on bolts. Kikuchi [9] analyzed a Taikoyama wind farm incident, developing a validated aeroelastic model to predict tower-top bending moments that generate downwind tensile stress due to rotor–nacelle center-of-gravity eccentricity. A detailed FEM further demonstrates that damaged bolts induce threefold higher nonlinear local tensile stress in the tower shell via leverage effects, with predicted tower fatigue life closely matching observed outcomes.

Regarding fatigue performance research on welded joints in steel structures, Hu [10] developed a semi-analytical method in tubular flange connections, with comparative results demonstrating good agreement between the proposed method and conventional approaches in calculating fatigue assessment parameters. Hoang [11] established a testing procedure for bolted flange connections and evaluated the accuracy of existing design methods by comparing their predictions with experimental results. Visentin [12] methodically probed into the fatigue proclivities of stiffened intricate steel pipe–flange junctures under diverse loading modalities, encompassing pure flexure, pure torsion, and conjoint flexure–torsion loading. Oh [13] proposed a simplified fatigue strength prediction methodology that comprehensively considers the effects of residual stresses, assembly stresses, flange flatness tolerances, weld toe angles, and tube diameter-to-thickness ratios.

Despite the widespread application of TFCs, research on their load-bearing capacity under cyclic loading remains insufficient. In TFCs, the steel tube of the flange, in conjunction with high-strength bolts, establishes the joint’s load transfer mechanism—an integrated system where the two elements operate cooperatively and affect each other. Stress concentration is prone to transpire at the weld where the steel pipe conjoins with the flange. When extrinsic loads impinge upon this locale, these zones will sustain elevated stresses, augmenting the peril of fatigue rupture. Imperfections during the welding procedure, such as welding apertures, gas voids, slag inclusions, et cetera, will all give rise to local stress intensification in the weld, thereby diminishing the strength and fatigue longevity of the welded construct. Concurrently, the choice and governance of welding process parameters directly affect the welding quality and fatigue comportment.

To address this gap, the present study systematically investigates the fatigue performance of M12 high-strength bolted flange welds through an integrated experimental and numerical approach: (1) conducting constant-amplitude fatigue tests to determine the 2-million-cycle fatigue strength; (2) comparing experimental results with design codes to propose optimized design recommendations; (3) developing finite element models to analyze stress concentration phenomena and identify fatigue failure mechanisms; and (4) combining macro- and micro-fractographic analysis to elucidate fatigue crack initiation and propagation behavior. The findings provide both theoretical foundations for fatigue-resistant design and practical guidelines for engineering applications of tubular flange connections.

2. Materials and Methods

Fatigue testing is one of the most common means to study fatigue behavior. Seven fatigue specimens of TFCs were designed and fabricated. A T-joint configuration was adopted for load transfer during testing. The specimens were subjected to constant-amplitude cyclic loading using an MTS fatigue testing machine.

2.1. Test Specimen

The T-joint configuration (dimensions detailed in Figure 1) was fabricated using Q355B steel, with tubular members measuring ϕ75.5 × 3.75  and flange plates of 160 mm diameter and 16 mm thickness. Single-bevel groove welds were executed via CO² gas-shielded arc welding to meet Class II weld standards [14]. The high-strength bolts employed are of the 10.9S grade, crafted from 20MnTiB steel, and they are in full compliance with the provisions of GB/T 1228-2006 [15].

The experimental investigation employed an MTS servo-hydraulic fatigue testing system to complete the fatigue tests of flange welds (Figure 2). The testing machine features a measurement accuracy of 0.00028 mm for maximum displacement loading, is capable of applying a maximum load of 500 kN, and can reach a maximum loading frequency of 100 Hz. In the setup, one end of the T-joint assembly was rigidly clamped to the MTS actuator, while cyclic loading was applied directly to the flange weld region. To replicate the operational stress state encountered in service conditions, two T-joint specimens were interconnected via four M12 Grade 10.9S high-strength bolts, ensuring proper load transfer through the bolted interface.

2.2. Loading Scheme

The components that bear the load on the test specimen in this article include steel pipes and bolts, so the maximum loading force of the test depends on the smaller value of the steel pipe tensile strength and the bolt tensile strength. The loading scheme for this experiment was selected based on the experimental conditions, as shown in

Table 1. Loading scheme table.

Number	Load		Stress Amplitude	Stress Ratio
	Fmax	Fmin	Δσn	ρ
TF-1	185.46	18.546	184.48	0.1
TF-2	168.60	16.860	167.70	0.1
TF-3	168.60	16.860	167.70	0.1
TF-4	151.74	15.174	150.93	0.1
TF-5	151.74	15.174	150.93	0.1
TF-6	134.88	13.488	134.17	0.1
TF-7	101.16	10.116	100.62	0.1

, where $\mathsf{\Delta }{\sigma }_{\mathrm{n}}=\left(F_{\max }-F_{\min }\right)/A={\sigma }_{\mathrm{n}}^{\max }-{\sigma }_{\mathrm{n}}^{\min }$. The loading rate is set to 5 Hz.

2.3. Fatigue Test Program

The process of the fatigue testing adheres to the stipulations of GB/T 13682-1992 [16]. During the fatigue test, the load is applied in a sinusoidal curve pattern, as depicted in Figure 3. The load decreases from F_max to F_min and then increases back to F_max within one cycle. When the fatigue crack propagates to 1/4 of the perimeter and the deformation becomes too large to continue loading, this moment is regarded as fatigue failure.

3. Results

3.1. Fractured Specimens

A total of seven welds at the connections of steel tube–flange plates underwent fatigue failure, as shown in Figure 4. Cracks occurred at the weld toes of the steel pipes in the steel pipe–flange connection welds of all seven specimens. The cracks gradually propagated along the weld toes. With the increasing displacement, fatigue fracture eventually occurred. After the machine is shut down, a universal testing machine should be used to break the specimen, exposing a complete fracture surface, which is convenient for subsequent fracture surface analysis.

3.2. Fatigue Failure Process

Five typical specimens were selected to illustrate the fatigue failure process of the steel tube–flange connection welds. The location where fatigue-induced fracture occurred in the weld connecting the flange and steel pipe of TF-1 is present in Figure 5. The stress amplitude is 197.46 MPa, and the stress ratio is 0.1. The fatigue fissure commences at the weld root, which is the juncture where the flange interfaces with the steel conduit, in proximity to the high-strength bolt. There is a slight undercut at this weld toe. Subsequently, with the load cycles, the fatigue crack progresses in the direction of the weld toe contour, and part of the crack extends into the interior of the weld. When the crack propagated to a quarter of the circumference, the fatigue test was stopped, and the specimen eventually failed.

Figure 6 shows the fatigue failure diagrams of specimens numbered TF-2 and TF-3. The stress amplitude applied to these two types of specimens is 179.51 MPa, and the stress ratio is 0.1. For the TF-2 connection, the fatigue crack begins at the weld toe of the steel pipe. With the increase in stress cycles, a small number of cracks deviate from the weld toe seam during propagation and penetrate into the interior of the weld. In the case of the TF-3 connection, the fatigue crack, from its initiation to propagation, adheres to the weld toe seam of the steel pipe. Notably, the cracks in both specimens originated near the high-strength bolts. This phenomenon suggests that there is a considerable degree of stress concentration at this location, rendering it more prone to fatigue failure.

Figure 7 shows the fatigue failure diagrams of specimens numbered TF-4 and TF-5. The stress amplitude applied to these two specimens is 161.56 MPa, and the stress ratio is 0.1. For specimen TF-4, its fatigue crack starts at the weld toe of the steel pipe and spreads along the weld toe gap. Unlike prior specimens, the fatigue crack of TF-4 is relatively slender and fails to penetrate the steel pipe wall thickness. As for specimen TF-5, its fatigue crack also initiates at the steel pipe weld toe. As stress cycles accumulate, a portion of the fatigue crack propagates into the steel pipe’s base metal and has already gone through the wall thickness.

It is found that the initiation positions of cracks in the welds of the steel pipe–flange connections all occur at the welds of TFCs, and the specific positions are at the weld toes close to the high-strength bolts. In the whole structural system, high-strength bolts play an important role in connection and fastening. However, due to the characteristics of force transmission and distribution in the proximity of the bolts, stress is highly concentrated in this area. From the perspective of mechanical principles, when external loads are applied to the structure, the forces will be transmitted and distributed through the steel pipes, flanges, and high-strength bolts. Near the bolts, the force lines will be significantly distorted and converge, thus causing the phenomenon of stress concentration. Compared with other parts of the structure, the stress level in this area is significantly higher. When the structure experiences alternating loadings over an extended period, this high-stress state will continuously cause damage accumulation to the materials. During the repeated stress cycles, tiny defects and cracks begin to initiate and propagate, and, eventually, fatigue failure occurs.

3.3. S-N Curve and Fatigue Design Method

Frequently employed approaches for fatigue design encompass the nominal stress approach, the hot spot stress approach, and the structural stress approach [17,18,19]. The nominal stress method is the most common method. Nominal stress refers to the stress value calculated according to the basic mechanical principles based on the macroscopic geometric shape of the component and the external forces that it bears, without considering complex factors such as the local stress concentration in the structure.

Therefore, the nominal stress approach is adopted to establish the limit state design approach for TFCs. The seven sets of fatigue data of the TFC welds obtained from the fatigue tests in this section are fitted with a power function to acquire the S-N curve (Figure 8).

The expression for the stress–life curve of the steel tube–flange weld is shown in Equation (1):

$$\mathsf{\Delta }\sigma =1378.43{\mathrm{N}}^{-0.174} r^2=0.869$$

(1)

To reduce the impact of a small sample size on the reliability of the statistical results, the formula specified in the IIW code [20], which is applicable to small samples, was used to calculate the S-N curve. The IIW code stipulates that, when the fatigue test data are limited (n ≤ 10), the S-N curve shall be calculated according to Equation (2):

$$\begin{gathered} x_i={\log }_{10}(\mathsf{\Delta }{\sigma }_i) \\ {y}_i={\log }_{10}(N_i) \\ {\log }_{10}(C_i)={\log }_{10}(N_i)+m\cdot {\log }_{10}(\mathsf{\Delta }{\sigma }_i) \\ {\log }_{10}(C)={\displaystyle \frac{\sum \left({\log }_{10}(C_i)\right)}{n}} \\ \mathrm{Stdv}\left({\log }_{10}(C)\right)=\sqrt{{\displaystyle \frac{\sum {(d{\log }_{10}(N_i))}^2}{n-1}}} \\ {(\mathrm{d}{\log }_{10}(N_i))}^2={\left(\begin{array}{c}{\log }_{10}(C)-m\cdot {\log }_{10}(\mathsf{\Delta }{\sigma }_i)-{\log }_{10}(N_i)\end{array}\right)}^2 \\ {\log }_{10}(C_k)={\log }_{10}(C)-k\cdot \mathrm{Stdv}\left({\log }_{10}(C)\right) \end{gathered}$$

(2)

where $\mathsf{\Delta }{\sigma }_i$ represents the i-th nominal stress amplitude and $N_i$ represents the i-th fatigue life. m is obtained according to the test data and is 4.375. $\mathrm{Stdv}\left({\log }_{10}(C)\right)$ is the standard deviation, and k is the multiple of the standard deviation, taking the value of 2.32.

After taking the double logarithm of the above formula, the S-N curve under a confidence level of 95% is shown in Figure 9.

The S-N curve expressed in double logarithm of the steel pipe flange weld under an assurance rate of 97.72% is as follows:

$$\begin{gathered} lg\left(N\right)=15.005-4.375lg\left(\mathsf{\Delta }\sigma \right)\pm 0.322 \\ {r}^2=0.882 \end{gathered}$$

(3)

The correlation coefficients in both Equations (2) and (3) are relatively large, indicating that the fatigue data exhibit good regularity. Under this circumstance, the allowable stress amplitude for 2 million cycles, ${[∆\sigma ]}_{{2\times 10}^6}$, is 82.41 MPa. Therefore, in the Eurocode 3 and the ‘Code for Design of Steel Structures’, the member and connection categories of the steel tube–flange welds should be classified as 80 and Z7, respectively.

The derived S-N curve is applicable to the fatigue design of welded connections in TFCs while specifically considering its application range. This S-N curve is validated for single-bevel groove-welded TFC joints fabricated via gas-shielded arc welding under the tested geometric configuration (steel tube dimensions: Φ75.5 × 3.75 mm). Extrapolation to other material grades or geometric parameters (e.g., larger diameter-to-thickness ratios) requires further experimental or numerical validation to ensure reliability in fatigue life prediction.

Due to the defects existing in the welding process, such as cracks, undercuts, porosity, etc., the discreteness of fatigue data may increase. These defects may serve as crack initiation sources, thus increasing the likelihood of fatigue failure. Therefore, welding defects should be minimized as much as possible during the actual construction process to improve the fatigue resistance of TFCs.

4. Discussion

4.1. Comparison with the Current Codes and Standards

As shown in Figure 10 and

Table 2. Fatigue design details of steel tube–flange connection welds in specifications.

Number	Reference Category	Illustration		Fatigue Strength Corresponding to 2 Million Cycles (MPa)
1	IIW: 913 category			50
2	IIW: 921 category		K-butt weld, toe ground	90
			Fillet weld, toe ground	90
			Fillet welds, as welded	71
3	Euro 3: Class 71Steel structure design standards: Z8 category			71
4	Euro 3: Class 40Steel structure design standards: Z11 category			40
5	This paper			82.41

, the existing codes and standards provide clear regulations on the fatigue strength for different types of flange weld forms [20,21,22]. The fatigue design standards recommended by the IIW have clear classifications and recommended values for the fatigue strength of different welding types and treatment methods. For example, the recommended fatigue strength value for a K-type butt weld with the weld toe ground is 90 MPa, while the recommended fatigue strength value for a fillet weld without grinding is 71 MPa. The European standards (such as Class 71 and Class 40) also classify the fatigue strength of different welding types and treatment methods. For instance, the fatigue strength of a socket joint with a full penetration butt weld of 80% is 71 MPa, while the fatigue strength of a socket joint with a fillet weld is 40 MPa.

Compared with the existing codes and standards, this paper conducts a fatigue test study on the tube–socket joint with groove welding and obtains a fatigue strength value of 82.41 MPa corresponding to 2 million cycles. There are obvious differences when compared with the fatigue strength of similar types of welds in the existing codes and standards. For example, in the European standard, the fatigue limit of the socket joint with fillet weld (Category Z11) is only 40 MPa. The value of the groove welded joint studied in this paper is much higher than that, indicating that different welding methods have a significant impact on the fatigue performance of the flange welds.

Compared with the steel tube–flange welds with bolt connections specified in the codes and standards, the groove welding form studied in this paper has a higher fatigue strength and better fatigue performance. The weld form has a significant influence on the fatigue strength and also indicates that the groove welds have a relatively high fatigue strength, which is worthy of being popularized in practical engineering projects.

Moreover, owing to the comparatively steep slope of the S-N curve in this research, Fatigue fracture is more prone to happen in high-stress areas for steel pipe–flange connections with high-strength bolts compared to those without high-strength bolts. This might be because the presence of high-strength bolts provides additional prying forces [23,24], which are more pronounced under high stress levels. Therefore, the influence mechanism of the prying forces of bolts on the fatigue behavior of TFCs deserves to be analyzed with emphasis in future research.

4.2. Numerical Simulation

4.2.1. Modeling Process

The steel tube–flange connection consists of a steel tube, a flange plate, and high-strength bolts. The model was established using ABAQUS, which includes these components, as shown in Figure 11. The steel tube along with the flange plate are made of Q355B steel, having an elastic modulus of 206,700 MPa and a Poisson’s ratio of 0.3. The C3D8R element is adopted for the model. Due to the presence of high-strength bolts in the model, the contact problem between the friction surfaces needs to be considered. In ABAQUS, the mutual contact relationship between the two components is simulated by defining the “contact pair”. The tangential anti-slip coefficient is taken as the friction coefficient used for the contact surfaces treated by sandblasting in engineering, that is, 0.4. The normal properties of all contact pairs are defined as hard contact, and separation between the elements after contact is allowed.

Since the fatigue tests are all carried out within the elastic range, the plastic stage of the steel is not considered. As the main focus is on the stress state of the welds, the threads of the bolts are not taken into account. During the test process, the steel tube–flange connection transfers the force to the bolts, and the bolts then transmit the load to the next flange plate. In terms of boundary conditions, the vertical end plate of the lower connecting component is fixed, and a force of 185 kN is applied to the upper end plate.

4.2.2. Calculation Results

To evaluate the stress concentration degree of the welds in the steel pipe–flange connections, this study uses the hot spot stress (σ_max) concentration coefficient (K_t) to analyze the stress concentration [25], and its expression is defined as follows:

$$K_t={\displaystyle \frac{{\sigma }_{max}}{{\sigma }_n}}$$

(4)

Figure 12 displays the stress contour diagram of the finite element calculation results. Figure 13 depicts the axial stress distribution over the steel pipe’s surface. As shown in Figure 12 and Figure 13, the stress level rises gradually as one approaches the weld toe. Notable stress concentration occurs in the steel tube–flange connection welds, especially at the steel pipe’s weld toe, a key zone for fatigue failure. Figure 14 shows the circumferential stress distribution at the steel pipe’s weld toe, including both the weld toe and the steel pipe’s middle section. Clearly, the circumferential stress distribution in the flange weld is uneven. The stress in the welds close to the high-strength bolts is relatively high, while the stress reduces at positions more distant from the bolts. At the point where the weld is closest to the bolt, the stress level at the weld toe peaks. By means of finite element analysis, the hot spot stress concentration factor is found to be in the range of 1.52 to 1.92, with the highest value appearing near the bolts.

In the entire structural system, the loads borne by the flange welds near the high-strength bolts exhibit a distinct uneven pattern. When an external axial load is applied, during the process of force transmission through the high-strength bolts, flanges, and related connecting components, due to the abrupt changes in the structural geometry and the interaction among various components, the stress in the area surrounding the high-strength bolts becomes relatively complex. This unevenness of the load leads to an extremely significant stress concentration at this position.

Owing to the fastening effect of the bolts, there are obvious differences in the constraint conditions and acting forces experienced by the welds in the circumferential direction. Near the bolt fastening points, the welds are more strongly constrained, while, in the parts far away from the bolts, the constraints are relatively weaker. This situation of uneven force in the circumferential direction further intensifies the degree of stress concentration.

Under the condition of fatigue load, this situation of stress concentration and uneven force distribution poses a challenge to the fatigue resistance of the welds. With the continuous cyclic application of the fatigue load, tiny cracks gradually initiate. These cracks gradually grow and propagate along the weaker paths within the material, eventually causing the entire structure to fail. The analysis results align well with the experimental observations in this study, which strongly corroborates the accuracy of the numerical simulation results and the reliability of the revealed mechanical mechanisms.

4.3. Structural Stress Method Analysis

For the purpose of further assessing the fatigue comportment of the TFC welded joints, the structural stress method was adopted. This method takes into account the local stress distribution and the geometric effects of the welds, and aggregates thousands of sets of fatigue data into a unified master S-N curve.

4.3.1. Theoretical Basis

On the basis of structural stress (σ_s), equivalent structural stress (ΔS_s) can be further calculated. For tubular joints, the equivalent structural stress is defined as follows [26,27]:

$$\Delta S_S={\displaystyle \frac{\Delta {\sigma }_S}{t^{\ast (2-m)/2m}I{\left(r\right)}^{1/m}}}$$

(5)

where t is the plate thickness, m = 3.6 is the inverse slope of the master S-N curve, and I(r) is a dimensionless function accounting for the bending-to-membrane stress ratio r.

Subsequently, the number of failures could be determined by using the master S-N curve depicted in Figure 15.

4.3.2. Structural Stress Distribution

Figure 16 illustrates the distribution of structural stress along the circumferential direction of the weld toe. Similar to the pattern of hot spot stress, the structural stress reaches its maximum value of 662 MPa near the high-strength bolts and its minimum value of 505 MPa away from the bolts. This pattern further validates the conclusion drawn in Section 4.2.

4.3.3. Fatigue Life Estimation

Table 3. Fatigue life prediction.

$\Delta {\mathit{\sigma }}_{\mathbf{n}}$	$\Delta {\mathit{S}}_{\mathit{S}}$	Npre (×104)	Ntest	log(Ntest)/log(Npre)
197.46	662	8.20	5.05	0.96
179.51	598	10.89	16.18	1.03
179.51	598	11.26	18.91	1.04
161.56	534	15.51	21.77	1.03
161.56	534	16.18	25.63	1.04
143.61	470	24.27	47.00	1.05
107.71	343	70.00	101.63	1.03

presents the predicted number of failures and compares them with the experimental data. The experimental fatigue life (N_test) agrees well with the predicted fatigue life (N_pre), with the double logarithmic ratio between the two ranging from 0.96 to 1.05. This result validates the generality of the structural stress method, signifying that the structural stress method can be utilized to forecast the fatigue longevity of TFCs.

4.4. Fracture Morphology Analysis

4.4.1. Macro Image Analysis

Fatigue fracture can generally be delineated into three distinct zones: the crack nucleation zone, the crack propagation zone, and the instantaneous rupture zone [28]. Since the steel pipe undergoes large deformation when it is about to break and it is not suitable for continuing loading, the fatigue life of this specimen is regarded as the state when the crack propagates to one-quarter of the perimeter of the steel pipe. Subsequently, the specimen is subjected to tensile force until rupture occurs, which facilitates the analysis of both the macroscopic and microscopic aspects of the fracture surfaces.

Figure 17 displays the fracture images of TF-2 and TF-3, where obvious fatigue source regions and fatigue propagation regions can be observed. The fatigue crack initiation region of the weld is usually adjacent to the high-strength bolts. This is particularly evident when there are defects in the weld, such as undercutting and thick weld metal. This phenomenon arises owing to the relatively complex stress conditions in these regions, which are prone to stress concentration and the initiation of fatigue cracks. Should there be substantial defects within the material, fatigue cracks may also nucleate within the weld itself. Upon macroscopic examination of the fracture surface, the fatigue crack initiation zone typically exhibits a relatively uniform and subdued appearance. During the fracture propagation phase, variations in the resistance encountered at the fracture tip may cause the crack to diverge from its initial plane of propagation. Once such divergence happens, the crack continues to extend along the new plane, thereby giving rise to the formation of intersecting fracture surfaces, which are referred to as “fatigue striations.” Pronounced fatigue striations are discernible in Figure 17. These fatigue steps indicate the direction of crack propagation, and the steps usually point toward the development direction of the crack.

4.4.2. Microscopic Image Analysis

The origin of fatigue failure usually lies in sites of stress concentration on the surface of the material or inside it, like near notches, cracks, inclusions, and other non-uniform defects. These areas have a high tendency for stress concentration, making them the main places for crack initiation. The fatigue source region generally presents a relatively flat or slightly sunken surface, often with tiny cracks or slip lines marking where the crack first formed. Figure 18 shows the morphology of the fatigue source area and the propagation area, magnified 200 times. Four representative regions were chosen for scanning electron microscopy analysis, including the fatigue source region and the fatigue propagation region.

In the area marked as “Fatigue step” in the figure, there is an obvious stepped morphology. This is because, during the propagation of the fatigue crack, the crack propagation rates on different planes are inconsistent, resulting in the formation of steps during crack propagation. The appearance of fatigue steps indicates that the crack has experienced different stress states and propagation paths during the propagation process.

The area marked as “River-like stripes” presents a striped morphology similar to that of a river. This characteristic is usually formed during the cleavage fracture process, indicating that the material has undergone cleavage fracture during the crack propagation process. The direction of the river-like stripes can indicate the crack propagation direction. Generally speaking, the converging direction of the stripes points to the crack source, while the diverging direction is the crack propagation direction.

The area marked as “Secondary cracks” shows the secondary cracks generated during the propagation of the main crack. The generation of secondary cracks is due to the local cracking of the surrounding materials caused by stress concentration during the propagation of the main crack. The existence of secondary cracks will further reduce the strength and toughness of the material, accelerating the crack propagation and the failure of the structure.

By examining the orientations of the chevron markings and fatigue striations, the direction of crack propagation can be ascertained. Typically, the crack originates from regions of stress concentration (such as the weld near high-strength bolt connections) and then extends along the direction of maximum stress. The convergent direction of the chevron markings indicates the crack source, and the arrangement of the fatigue striations is also correlated with the crack propagation direction.

The appearance of fatigue steps, fatigue striations, and secondary cracks all indicate the gradual buildup of fatigue damage. Fatigue striations arise from the step-by-step crack growth caused by each stress cycle. With the accumulation of cycles, the crack slowly elongates. The appearance of fatigue steps implies that the crack has met changes in material properties or stress conditions during its spread, causing it to deviate from its original path. The formation of secondary cracks denotes a further worsening of fatigue damage.

Figure 19 shows the shape of the area where fatigue propagates on the large-scale fracture surface at the weld toe of the steel tube, magnified 5000 times. It is clear that many fatigue striations are present in this area, which are typical signs of a fatigue fracture. The direction of these fatigue striations is perpendicular to the direction of crack propagation, and they typically form during the later phases of steady crack growth.

5. Conclusions

This research carries out a constant-amplitude fatigue test on the welded joints of steel tube–flange connections, generating seven groups of fatigue data. Using the allowable nominal stress amplitude approach, the constant-amplitude fatigue S-N curve for the flange welded joints is developed. This curve is compared with existing codes and standards. Stress concentration analysis is performed using finite element software, and the fatigue fracture surfaces are examined at both macroscopic and microscopic scales. The principal conclusions are as follows:

• The fatigue cracks of the high-strength bolts always nucleate at the weld toe where the flange plate meets the steel pipe, close to the high-strength bolts. Subsequently, these fatigue cracks spread along the weld toe direction under cyclic loading.
• Finite element analysis shows considerable stress concentration at the weld toe of the steel tube–flange weld near the high-strength bolts. The non-uniform stress distribution in the circumferential direction of the weld makes this location more prone to fatigue failure. The finite element analysis results align with the experimental observations in this study, thereby validating the accuracy of the finite element model’s analysis results and the reliability of the explained mechanical mechanisms.
• The allowable stress amplitude for 2 million cycles of the steel tube–flange weld is 82.41 MPa, which is greater than the fatigue strength of the relevant category in the current codes and standards. This indicates that the steel tube–flange weld has better constant-amplitude fatigue performance compared with the flange connections specified in the codes. This advantage is due to the fact that the specimens in this paper adopt the groove welding method for welding, which has better fatigue performance than fillet welds. Therefore, in practical engineering designs, for steel tube–flange welds that need to withstand cyclic loads, groove welding can be preferentially used for connection.
• The fracture analysis shows that there are obvious fatigue crack initiation regions and crack propagation regions in the fracture of the steel tube–flange welds. A large number of fatigue steps are generated in the fatigue source region, indicating the existence of multiple fatigue sources. Large numbers of river-like stripes and fatigue bands in the crack propagation region are typical characteristics of fatigue failure.