3.1. Fatigue Behaviour in Air
To evaluate the fatigue damage behaviour, the material reactions of brazed specimens under cyclic loadings in air were recorded using various measurement techniques. Hence, the progressions of the total mean strain εm,t, the total strain amplitude εa,t, the loss energy density w, the change in temperature ΔT and the change in AC voltage ΔU are presented as functions of the load cycles N in the following. The loss energy density w describes the areas of the hysteresis loops and is calculated by w = ∮ σ dε.
The characteristic curves for a fatigue test with a constantly controlled maximum stress σ
max = 530 MPa are presented in
Figure 4 a linear grit to improve the correlation of the different techniques for the fatigue state close to failure. During the initial fatigue stage up to about 10
4 cycles, a considerable total mean strain ε
m,t up to 23% was reached. The increase of the total mean strain was a result of the directional accumulation of plastic strains due to the stress ratio of R = 0.1. In addition, the loss energy density w decreased regressively due to cyclic hardening effects. Therefore, the hysteresis loops were continuously shifted to higher strain values, while the size of the hysteresis loop areas and the absorbed energy per cycle decreased. As an increase of the magnetic portion of Δζ = 6 vol.% was detected with the feritscope after 3000 cycles, a deformation-induced phase transformation to martensite was the reason for the cyclic hardening in the initial stage. The amount of the formed martensite is known to increase with increasing plastic deformations. Thus, a decreasing ratcheting strain rate within the first fatigue stage was expected to contribute to the degressive trend of the martensite formation and to a regressive change of the loss energy density. Within the following saturation stage, the cyclic deformation curves indicated a steady strain rate with slowly progressing ratcheting fatigue and cyclic hardening effects. Within the third stage, starting at about 5000 cycles before failure at N
f = 7.4 × 10
4, progressively increasing ε
m,t, ε
a,t and w values indicated a crack propagation until failure.
For the overall fatigue life, the ΔU curve corresponded well with the total mean strain. The direct relation of the curves may be used for a quantitative analysis of the cyclic deformation behaviour of brazed joints in future studies. In addition, there was a direct relation of the cyclic hardening behaviour and the specimen’s temperature, because 95% of the loss energy density w is known to dissipate into heat. At the life fraction N/N
f = 0.5, a change in temperature of ΔT = 8.7 K, with an absolute temperature within the gauge length T
1 = 42.8 °C, was determined. As the maximum change in temperature ΔT was below 20 K, a significant influence of the self-heating on the martensite formation could be excluded [
11]. Consequently, temperature and electrical measurements are well applicable for brazed AISI 304L/BAu-4 joints to characterise ratcheting fatigue and cyclic hardening effects.
A comparable evolution of ratcheting strains is described and explained for austenitic stainless steels in [
10,
12,
14]. During the first fatigue stage, the increasing density of the mobile dislocations and the formation of dislocation cells lead to a degressive increase of the total mean strain. The saturation is expected to be a result of a stable configuration of the newly generated dislocations in the base material [
14]. During the third stage, the progressive increase indicates rapid dislocation activities and crack propagations until failure [
12].
The cyclic deformation curves, determined in four fatigue tests with constantly controlled maximum stresses σ
max within the range of 410–560 MPa, are presented in
Figure 5 and
Figure 6. A greater accumulation of ratcheting strains at the beginning of the test and pronounced ratcheting strain rates in all three characteristic fatigue stages could be detected for higher stress levels of the brazed AISI 304L/BAu-4 joints (
Figure 5a). For all specimens, the ratcheting strain rate decreases considerably during the saturation stage compared to the rate at the test beginning. The highest stress level of σ
max = 560 MPa led to total mean strains of approximately 28% within the saturation stage. The accompanying increase of the total strain amplitudes with increasing fatigue stresses is visualised in
Figure 5b. The decreasing σ
m and σ
a, combined with decreasing ε
m,t and ε
a,t lowered the total damage and hence extended the fatigue life.
For austenitic stainless steels, the degree of deformation during cyclic loading, depending on the applied stresses, has already been investigated and the results published [
9,
10,
12,
13,
14]. During cyclic loading, the highest density of mobile dislocations in the base material is expected to be obtained at the time of the maximum tension loading [
12]. As there are no compressive stresses and no reverse plastic deformations during unloading, formed dislocations cannot be annihilated. Consequently, a net storage of mobile dislocations in the base material after each loading cycle is expected, which is more pronounced for higher maximum stresses [
12,
14]. Furthermore, inelastic ratcheting strains during fatigue loading lead to a reduction of the cross-sectional area of the brazed specimens and to an increase in true stresses [
12]. This effect is also expected to be more pronounced for higher fatigue stresses, leading to higher ratcheting rates, especially in the third fatigue stage [
12]. The described decrease of the total damage and the extension of the fatigue life for the brazed joints with decreasing σ
m and σ
a, combined with decreasing ε
m,t and ε
a,t, was also reported for the base material [
13,
14].
With increasing fatigue stresses, the cyclic hardening in the first and second stage was more pronounced, although the respective decrease of w always showed a higher gradient at the test beginning in comparison to the second steady state (
Figure 6). The third stage constantly signified cyclic softening due to fatigue damage, crack propagation and failure. Presenting the results in a linear grit of the life fraction N/N
f enabled evaluating the relative proportions of the three deformation stages during the fatigue lives. For higher stresses, cyclic hardening at the beginning of the test and cyclic softening effects close to failure were present for a higher proportion of life. For the highest stress level σ
max = 560 MPa, a cyclic hardening within the first cycles could be determined, followed by a cyclic softening and a subsequent secondary cyclic hardening. For lower stresses, a cyclic softening of the brazed joints could not be shown with the areas of hysteresis loop.
An earlier phase transformation to martensite and a higher martensite formation rate for higher fatigue stresses and ratcheting strains are also described for austenitic stainless steels [
9,
14,
27]. It is well documented in the literature that the deformation-induced formation of martensite is significantly influenced by the chemical composition of the metastable austenitic stainless steels and the loading condition, which influences the degree of deformation and the deformation temperature [
10,
15,
28]. A self-heating of the specimens due to cyclic loading with constant amplitudes leads to changes in temperature ΔT < 20 K. Thus, a significant effect on the formation of martensite can be neglected [
11].
The evolution of the loss energy density w of the brazed joint, tested at the highest stress level σ
max = 560 MPa, could be correlated to test results of the austenitic stainless steel [
10]. Within the first cycles, the formation and interaction of new dislocations were expected to lead to a cyclic hardening, followed by a cyclic softening due to the increasing mobility of the dislocations and the formation of a dislocation structure. The subsequent secondary cyclic hardening can be explained by a deformation-induced martensitic transformation [
9,
10]. Consequently, the cyclic deformation behaviour of the brazed stainless steel joints using the gold-based filler metal BAu-4 is characteristic for the metastable austenitic base material [
9,
14]. A detailed analysis of the fatigue damage behaviour with respect to microstructural changes in the brazing seam requires additional experiments and metallographic investigations.
The results of eight fatigue tests were analysed by relating the controlled maximum stress levels σ
max to the number of cycles to failure N
f in a diagram with a double logarithmic scale (
Figure 7). In addition, results of the AISI 304L base material from earlier studies [
5] are presented. The S-N curve for the brazed joints is described with the fit line σ
max = S
D ∙ (N
D/N
f)
1/k according to Basquin, with the coefficient of determination R
2 = 0.9018. Thereby, the fatigue strength of 397 MPa at 2 × 10
6 cycles and the gradient k = 12.4 were calculated [
8]. The results of the base material [
5] were positioned within the scatter range of the experimental data of the brazed joints. It can be concluded that the fatigue strength up to 2 × 10
6 cycles of the stainless steel AISI 304L is not significantly reduced due to the brazing process used.
3.2. Corrosion Fatigue Behaviour in a Synthetic Exhaust Gas Condensate
The results of fatigue tests with superimposed corrosion fatigue loadings, performed in situ with σ
max = 410 MPa in a synthetic exhaust gas condensate K2.2, are shown in
Figure 8. The measured values of the total strain amplitude ε
a,t, total mean strain ε
m,t, loss energy density w and the electrochemical open-circuit potential E
OCP are visualised in dependence on the load cycles N. Comparable to the fatigue tests in air, a continuous increase of the total mean strain ε
m,t with three characteristic stages was detected due to ratcheting effects. In contrast, ε
a,t remains at a constant level up to 5000 cycles prior to the failure at N
f = 8.8 × 10
4. Within the first stage, a decrease of w indicates cyclic hardening effects due to the deformation-induced phase transformation to martensite. The strain values, which were measured integrally with an extensometer with knife edges positioned at the shafts of the specimen, only allow a qualitative assessment of the cyclic deformation behaviour. For local strain measurements at the brazing seam in the condensate, the use of drop-down extensions at the knife edges of the extensometer will be optimised in future studies.
At the very beginning of the cyclic loading, a sharp decrease of EOCP down to −84 mV was observed. Up to 3 × 104 cycles, EOCP was shifted to higher values, whereby the rate of change decreased. Subsequently, a continuous cathodic shift of EOCP was determined during the steady strain rate stage until up to approximately 8.3 × 104 cycles. Finally, a significant decrease of EOCP indicated the failure of the specimen.
The evolution of E
OCP (
Figure 8) wass expected to be mainly influenced by dissolution and repassivation mechanisms, which depend on the cyclic deformation and damage behaviour of the brazed joints. The abrupt decrease of E
OCP at the beginning of the test indicated a quick rupture of the passive layer of the stainless steel due to the first loading. As a result of the high ratcheting strain rate within the first stage, additional active material surface was exposed to the electrolyte. Here, the increasing E
OCP was attributed to the repassivation of the active stainless steel surface. Despite the decreasing ratcheting strain rate in the steady state, leading to a smaller formation rate of the active surface, a cathodic shift was determined. The sharp cathodic shift within the third stage was correlated with the cyclic softening due to fatigue damage, crack propagation and failure.
In the literature, a continuous cathodic shift of E
OCP of stainless steels, mechanically loaded in a medium, is also described as a result of a steady rupture of the passive film [
29,
30,
31]. The rupture leads to an anodic dissolution of the active material surface below the passive oxide layer, which is exposed to the electrolyte. In general, a formation of a passive film can lead to a gradual increase of the E
OCP [
30]. The destruction and formation of an oxide layer during cyclic loading in a medium depend on various parameters such as the temperature, the test frequency, loading waveform and stress ratio [
32]. By increasing the load and the loading speed, the degree of the cathodic shift of the E
OCP of the austenitic stainless steel AISI 304 is known to be more pronounced because of a reduced time for the repassivation [
4,
30].
In the following, possible reasons for the cathodic shift during the steady state are discussed. To evaluate the E
OCP of brazed joints, electrochemical potential differences, which occur due to different compositions of the gold-base filler metal and the austenitic stainless steel, have to be taken into consideration. Furthermore, local sensitised areas and micro-compositional differences can occur at the interfaces and diffusion zones, as already reported for welded and brazed joints [
4,
21,
22]. The resulting localised galvanic cells in the area of the brazing seam generate active local regions, which are prone to local corrosive attacks [
20,
21]. In addition, the thermal expansion mismatch between the austenitic stainless steel and the gold-base filler metal during the brazing process may have generated residual stresses in the area of the brazing seam [
2]. Residual stresses are also known to increase the susceptibility of environmentally assisted cracking with respect to crack initiation and propagation, as shown for weldments [
21]. Thus, the cathodic shift during corrosion fatigue loading of brazed joints is expected to be influenced by the fatigue crack initiation and propagation as well as by the passivation kinetics, passive film thickness [
33] and local galvanic corrosion effects [
21,
22]. The fracture position and fracture behaviour as well as the deformation-induced formation of martensite may also affect the corrosion fatigue behaviour. In this study, the possible reasons for the cathodic shift during the steady state were not further investigated.
The controlled maximum stress levels in relation to the numbers of cycles to failure N
f, determined in eight fatigue and eight corrosion fatigue tests of brazed specimens, are given in
Figure 9a. The S-N curve of the specimens, tested in the synthetic condensate, is described with a fit line in the double logarithmic scale. A fatigue strength of 202 MPa at 2 × 10
6 cycles and a steeper gradient of k = 4.4 was determined with R
2 = 0.9334. Thus, the fatigue strength at 2 × 10
6 cycles of specimens tested in air was significantly reduced down to 51% due to the superimposed corrosive loading. As corrosion mechanisms are time-dependent [
32], longer test durations are known to increasingly affect the fatigue properties of the brazed joints. For the design of the brazed components, the effect of the test frequency on the corrosion fatigue behaviour has to be taken into consideration.
The authors investigated stainless steel joints brazed with the nickel-base filler metal BNi-2 in earlier studies [
34]. The S-N curves of the brazed AISI304L/BNi-2 joints determined for fatigue and corrosion fatigue loadings are presented in a semi-logarithmic scale in
Figure 9b. In contrast to the high-strength brazed AISI304L/BAu-4 joints as well as the applied austenitic base material, a significant reduction of the fatigue properties down to 50% is observed. At 2 × 10
6 cycles, a fatigue strength of 210 MPa is obtained in air. The number of cycles to failure N
f of brazed AISI304L/BNi-2 joints cyclically tested with maximum stresses above 210 MPa is reduced with a factor of approximately 5 in the synthetic condensate K2.2. Smaller stresses increasingly influence the fatigue behaviour due to longer test durations. A fatigue strength of 90 MPa was determined in the synthetic condensate for 2 × 10
6 cycles. Thus, a reduction of the fatigue strength down to 43% due to the superimposed corrosive loading has to be taken into consideration for the brazed AISI304L/BNi-2 joints [
34].
When comparing the percentage values of 51% (AISI304L/BAu-4) and 43% (AISI304L/BNi-2), the corrosion fatigue behaviour of the brazed joints using the gold-base filler metal seems to be improved. Considering the absolute values, the fatigue strength at 2 × 106 cycles was reduced by 195 MPa for the AISI304L/BAu-4 joints and by 120 MPa for the AISI304L/BNi-2 joints. Therefore, a greater degradation of the fatigue properties due to the loading in synthetic condensate was determined for the high-strength and corrosion-resistant brazed joints with the gold-base filler metal. A comparison of the microstructure-based corrosion fatigue damage mechanisms of brazed AISI304L/BAu-4 and AISI304L/BNi-2 joints will be presented in future studies.
3.3. Fractographic Analysis
The fatigue and corrosion fatigue damage mechanisms were analysed based on fracture surfaces and polished sections using SEM. Thereby, the dark-grey stainless steel and the light-grey filler metal BAu-4 could be distinguish with a BSE detector. The damage mechanisms of the brazed specimens, cyclically tested in air before and after pre-corrosions according to the VDA test procedure 230-214 [
23], are presented in earlier studies by the authors [
8,
24,
25]. In air, the fatigue crack was initiated on the left side of the fracture surface, as presented in
Figure 10a. The origin of the fatigue crack was an imperfection close to the surface within the brazing seam. Such an imperfection with a free solidified surface of the filler metal appeared light-grey on the fracture surface and was described by the authors in detail in [
8]. After initiation, the crack propagated alternatingly within the brazing seam, the interfaces and the base material close to the diffusion zone. In this context, dark-grey areas can be seen next to the light-grey areas in the crescent-shaped fatigue fracture area. The final fracture with pronounced deformation characteristics was detected in the centre of the light-grey brazing seam [
8,
25].
After the long-time pre-corrosion, a circumferentially corrosive attack in combination with pits in the area of the brazing seam was detected on the fracture surface (
Figure 10b). The local corrosive attack in the area of the diffusion zones is a result of micro-compositional differences at the brazing seam, which lead to the formation of local galvanic cells. The corroded diffusion zones serve as a crack initiation zone and lead to numerous dark-grey fatigue fracture areas. Here, the cracks mainly propagated within the base material. The final fracture occurred in the light-grey brazing seam [
8,
25].
The fracture surface of a brazed specimen after corrosion fatigue loading in the synthetic condensate K2.2 with σ
max = 410 MPa for a test duration of 2.5 h, is shown in
Figure 10c. A single fatigue crack initiation can be identified on the left side of the top-view. The polished section of the crack initiation zone shows a fracture in the centre of the brazing seam that commonly occurred at the transition between the gold-rich and nickel-rich phases (Label I,
Figure 11a). A secondary crack between these phases can be seen in more detail in
Figure 11b.
In accordance with the as-received and pre-corroded specimens tested in air, the fatigue fracture (Label II) is mainly located in the base material close to the brazing seam and the final fracture (Labels III and IV) occurred within the brazing seam (
Figure 12a,b). A polished cross-section of the area, where the crack position changes from the base material to the brazing seam, is shown in
Figure 12a. In contrast to the earlier investigations on brazed AISI 304L/BNi-2 joints [
4], corrosive products were not detected on the fracture surface of specimens that were cyclically tested in the medium. Therefore, a repassivation of the active stainless steel fracture surface was expected. Such corrosion fatigue damage mechanisms have also been described for non-brazed austenitic stainless steels [
35,
36].
The fracture characteristics for the fatigue fracture and final fracture areas are quite comparable for fatigue and corrosion fatigue loaded brazed specimens. The fatigue fracture areas of specimens, cyclically tested in the synthetic condensate (
Figure 10c), were more pronounced than the fatigue fracture areas of specimens tested in air (
Figure 10a). It was assumed that the brazing seam was susceptible to environmentally assisted cracking and, therefore, the fatigue crack was initiated in the medium at lower fatigue stresses. Lower stresses in combination with the medium promoted the fatigue crack propagation in the base material, which led to a larger fatigue fracture area when compared to specimens tested in air.
The fractographic investigations enabled a comparison of the damage mechanisms of brazed AISI 304L/BAu-4 joints caused by corrosion fatigue loading and fatigue loading after a long-term pre-corrosion. After pre-corrosions, the local corrosive attack in both diffusion zones reduced the remaining cross-section area and facilitated several fatigue crack initiations (
Figure 10b). In contrast, isolated fatigue crack initiations occurred during testing in the synthetic condensate K2.2 and the fatigue crack initiation was environmentally assisted. Finally, the application-relevant corrosion fatigue behaviour of brazed joints could not be sufficiently characterised by fatigue tests of pre-corroded joints. Corrosion fatigue tests in the synthetic condensate need to be additionally taken into consideration.