Analysis of the Anisotropic Cyclic Material Behavior of EN AW-1050A H24 Derived from Strain-Controlled Testing Using a Clip-On Extensometer and an Optical System

Abstract

Due to its good conductive properties, unalloyed (pure) aluminum, such as EN AW-1050A H24, finds new fields of application in electromobility. To optimize components, the cyclic material behavior must be understood and described precisely as a foundation of a proper fatigue life estimation. Various cyclic tests were performed to not only derive the cyclic parameters to describe the material but also to find the most suitable procedure to deal with the challenges faced during the experiments. The main point of interest is the comparison between a surface-mounted clip-on extensometer and an optical system both used for strain control in cyclic tests. For the analysis of the anisotropic behavior of EN AW-1050A H24, un-notched flat specimens were extracted from sheet metal lengthways and crossways in respect to the rolling direction. While the cyclic material behavior for specimens of both directions of extraction is characterized by cyclic softening in general, the specimens extracted crossways show a strain-amplitude-dependent cyclic softening with strong strain localization especially at the contact points of the knives of the clip-on extensometer leading to an increased quantity of invalid experiments as well as sudden fractures. In the study, it was possible to show the benefits of a contactless optical strain control system when dealing with very soft metallic materials such as EN AW-1050A H24.

1. Introduction

To meet climate goals, electromobility is growing in popularity. Due to the shift of a whole industry, various metals now find new fields of applications. As an example, the aluminum alloy EN AW-1050A H24 is of great interest due to its good conductive properties. Consequently, the material behavior of that material needs to be accurately understood to assess fatigue life. During extensive testing of the cyclic material behavior of specimens extracted from sheet metal, anisotropy was detected in the stress–strain behavior. While the specimens extracted both lengthways and crossways in respect to the rolling direction of the sheet metal showed softening behavior, the softening from the crossways-extracted sheet metal was more distinctive, depending on the strain amplitude and most importantly, not according to the behavior described by Ramberg–Osgood [1]. In this study, the anisotropic material behavior of EN AW-1050A H24 is laid out, the resulting challenges in fatigue life testing are discussed and the consequences for fatigue life estimation are shown.

To face the challenges in fatigue life testing, a comparison between an optical strain measurement system and a clip-on extensometer is carried out. In various papers like [2,3,4], the advantages of optical strain measurement is shown and the performance of the systems are proven to be sufficient to process pictures in the sampling rate needed. In [5], the problem with tactile extensometers and incompatible material surfaces of very soft and weak solids is discussed; as a result, an optical system and clip-on extensometers are compared and found to be in good agreement. This is confirmed in [6], in which the two types of strain measurement were also compared, and the data were found to be similar. In [7], a difference between the two testing methods was shown and the reason could be traced back to the indentation of the clip-on extensometer pins and the fact that the polymeric material tested tends to creep.

In past studies, the anisotropy of EN AW-1050 or comparable, commercially pure aluminum alloys, was already investigated in various contexts. In [8], the anisotropic ductile fracture was investigated and anisotropic behavior in tensile tests was proven. In contrast, the tests on compression specimens in [9] showed no anisotropy, which was attributed to the soft-annealed condition of the material tested. In [10], the flow stresses within commercially pure aluminum were measured and a significant anisotropy was detected both at low and high strains. Ref. [11] analyzes the anisotropy of an aluminum alloy and specifies the influence of the loading direction on strain localization. The result here is that the more orthogonal the extraction direction, the more stochastically the localization is and therefore, the more slip bands are created. In [12], the anisotropic behavior of AA1050 produced by accumulative roll bonding (ARB) was investigated. The main finding is that with increasing ARB cycles and therefore increasing plastic deformation, the normal anisotropy increases, whereas the planar anisotropy decreases.

While the challenges of testing with extensometers as well as the anisotropy of aluminum is discussed in the current literature, it is of importance for use in the automotive sector to characterize the exact material used in more detail, which has not yet been performed. Additionally, to the best of the authors’ knowledge, the challenges of testing soft aluminum with tactile extensometers has not yet been discussed, although soft, pure aluminum alloys will become more important in industrial application in the future.

2. Materials and Methods

To assess the anisotropic cyclic behavior of EN AW-1050A H24 fatigue life tests were carried out. The acquired experimental data are analyzed using the cyclic stress–strain curves according to Ramberg-Osgood [1] and the strain–life curves according to Basquin [13], Coffin [14], Manson [15] and Morrow [16].

2.1. Material

The commercially pure aluminum alloy EN AW-1050A H24, the material investigated in this study, is part of the 1xxx series of the European norm EN 573-1 [17], containing more than 99% of aluminum. The whole chemical composition is shown in

Table 1. Chemical composition of EN AW-1050A in % of mass, remaining parts to 100% aluminum [18].

Si	Fe	Cu	Mn	Mg	Cr	Zn	Ti	Others
0.25	0.4	0.05	0.05	0.05	-	0.07	0.05	0.03

, and the mechanical properties are shown in

Table 2. Mechanical properties of EN AW-1050A H24 according to [19].

Tensile Strength [MPa]	Yield Point [MPa]	Hardness [HBW]
105–145	75	33

.

The material condition is indicated as H24. According to DIN 515, the condition H means that the material was work-hardened. More precisely, with the first digit, H2x, it is denoted that the material was work-hardened beyond the desired final strength and re-annealed for better formability. The second digit in the material condition indicates the degree of hardening. Possible digits are 2, 4, 6 and 8, where 8 is defined as fully hardened. Therefore, H24 is half-hardened, right between the annealed and the xx8 material condition [20].

2.2. Geometry of the Specimens

Figure 1 shows the geometry of the specimens that were used. The specimens were extracted from a sheet metal lengthways as well as crossways in respect to the rolling direction. These flat specimens have a mandatory 10 mm gauge length for the strain measurement with a surface-mounted clip-on extensometer and are elliptically rounded towards the surface used for clamping to ensure a stress concentration factor of K_t ≈ 1. The specimens are 1.19 mm in thickness and have a width of 6 mm, resulting in a cross section at the measurement area of 7.14 mm².

To investigate the cyclic material behavior while minimizing the influence of the surface, the edges of the tested specimens were polished and rounded. This allows for an R_a-value below 0.2 μm and an R_z-value below 0.7 μm as necessary according to [21]. Compliance with the values was randomly checked with a roughness meter. The rolling surface of the sheet metal stays untreated since crack initiation is expected to initiate from the edges of the specimens.

2.3. Cyclic Tests

The extracted specimens are used for cyclic testing carried out on a so-called E-cylinder test system as shown in Figure 2 and described in more detail in [22]. The E-cylinder test system has been designed by the Fraunhofer LBF to perform cyclic testing in the LCF regime.

In the test bench, the specimens are fixed with a clamping system that holds the specimens between the upper rigid part and the lower part which can be moved uniaxially to apply loads. The actuator is a stepper motor that can move very precisely and rotates a spindle which translates the movement from rotation to axial displacement. To measure force, a load cell by Interfaceforce, model 1710DLL—2.5 kN, was used. These load cells measure force with resistance strain gauges, giving an output voltage that is correlated to a certain force based on calibration factors. For measuring the strain at the surface of the specimens, either a clip-on extensometer or an optical strain control system can be used.

For strain-controlled tests, the strain value is the control parameter, while the force-signal is used to detect a decrease in stiffness, ratcheting and an initiating crack, which is the criterion of the test end.

The procedure used in this study uses a sinusoidal waveform of the strain-signal. Since the tests were performed under alternating fatigue loading without mean strain, the maximum and minimum strain equals the strain amplitude. Since a velocity is defined in the test setup, the test frequency is dependent on the distance and in the case of the specimens of this study never exceeded 1 Hz. Because of the very slow testing, a heating of the sample can be neglected. The tests were carried out under ambient temperature (t ≈ 21 °C).

2.3.1. Cyclic Tests with Clip-On Extensometer

A common method of strain measurement uses surface-mounted clip-on extensometers which are in contact with the specimen by two knives that move towards or away from one another. In this study, an extensometer by Sander Messtechnik, model EX A 10–0.25 U with a maximum strain to be measured of 2.5% was used, as shown in Figure 3.

The strain is calculated by the displacement between the two knives divided by the base length of the extensometer of in this case 10 mm. While it is an advantage of this kind of extensometer that the technology is long proven and relatively simple, it comes with some downsides, such as hard handling when mounting the extensometer on the surface of the specimen, a measurement of an integral size over the base length and possible damage to the surface due to the knives.

2.3.2. Cyclic Tests with Optical Strain Control System

An enhanced method to measure strain is using an optical system, such as the Fraunhofer RODiS system used in this study. A further description of the Fraunhofer RODiS system is provided in [23]. Those optical strain measurement and control systems use cameras to observe the surface of the specimen. Just before the start of the test a reference picture is taken, and two reference areas are defined. For comparability with the data from the clip-on extensometer, the reference areas were located right beneath the knives of the clip-on extensometer, as shown schematically in Figure 4.

These reference areas have a unique pattern on the surface, either coming from production or by application of a statistical pattern on the surface. That makes the identification of a precise point on the surface of the specimen possible during testing. The movement of these two areas are tracked in two dimensions and the strain is calculated by dividing the movement between the two references by the base length, which was selected to be 10 mm just like for the clip-on extensometer. The result is a plane strain.

The advantage of optical strain measurement and control systems is a simpler setup procedure as well as the possibility to vary the distance between the two tracked areas. That allows the measurement of strain in a small area of interest. Although the measured strain is still integral, it can now be calculated for a far smaller base length. The third advantage is the contactless strain measurement allowing to eliminate the risk of pre-damaging the surface, especially for soft materials.

3. Results

With the described test setup and equipment, strain-controlled cyclic tests in the range 1 × 10² < N_f < 1 × 10⁵ were performed on a total of 35 specimens and evaluated. A total of 25 of the specimens were tested with the clip-on extensometer, 15 of which were extracted lengthways and 10 crossways in respect to the rolling direction of the metal sheet. An overview is given in

Table 3. Overview over the specimens tested.

	Clip-On Extensometer	Optical System
lengthways	15	0
crossways	10 (+13 invalid)	10
∑	25	10

.

The remaining ten specimens were extracted crossways and tested with the optical strain measurement system. In the following section, the test results in respect to anisotropy are presented and the consequences of the anisotropy for fatigue life evaluation are discussed. Additionally, the challenges in fatigue life testing due to the investigated anisotropy are shown.

3.1. Analysis of the Experimental Data Dependent on Extraction Direction

The following paragraphs show the cyclic material behavior dependent on extraction direction, shown with the strain–life curves and the cyclic stress–strain curves. A strain amplitude dependent cyclic softening behavior is discovered for the crossways-extracted specimens.

3.1.1. Strain–Life Curve

One of the most fundamental pieces of information for the description of materials is the number of cycles until crack initiation dependent on strain amplitude. In Figure 5 and Figure 6, the strain–life curves according to Basquin [13], Coffin [14], Manson [15] and Morrow [16] are given both for the lengthways and crossways-extracted specimens. While the evaluation of the strain–life curves was performed on the basis of the data of the clip-on extensometer, Figure 6 shows the location of the optically controlled data points for comparison. Even though the description of Basquin–Coffin–Manson–Morrow results in a satisfactory representation of the test results, improvements deriving a more accurate strain–life curve might be possible, e.g., applying the tri-linear strain–life curve [24] or even the so-called Fatigue Life Curve [25], especially if test results at lower loading are included.

Although, the comparison of both strain–life curves shows that the expected life is not strongly influenced by the direction of extraction, it is worth mentioning that for the strain–life curve in Figure 6, 13 experiments were invalid according to SEP1240 [21] and VDA 239–300 [26] due to a sudden fracture of the specimen right beneath the knives of the clip-on extensometer. The reason for that can be found in a significantly more distinctive strain localization for the specimens extracted crossways from the sheet metal. While for those 13 specimens, the notches created by the specimens were so significant that they represent the weakest point, for the other valid specimens, another point between the knives of the extensometer were the weakest point or the notches created by the knives were minor. As a result, the first dislocation in the material started there instead of under the knives, initiating the strain localization there. The material mechanics of the strain localization, which has a self-reinforcing effect once an initial damage is created along a certain dislocation of the crystal structure, also explain why for the valid specimens the potential damage underneath the knives of the extensometer can be neglected.

3.1.2. Cyclic Stress–Strain Curve

Strain localization is also the reason for the discovery to be found in the cyclic stress–strain curves according to Ramberg–Osgood [1]. In Figure 7 and Figure 8, the cyclic stress–strain curves can be seen both lengthways and crossways with respect to the rolling direction of the sheet metal. Analogous to the strain–life curves, the evaluation only considers the data of the clip-on extensometer, but Figure 8 provides the position of the optically controlled datapoints for comparison.

The theory provided by Ramberg–Osgood [1] gives a satisfactory description of the stress–strain behavior in the lengthways direction when deriving the stress–strain curves from Basquin–Coffin–Manson–Morrow strain–life curves. Nonetheless, it can be expected to achieve a further improvement when using the compatibility with the tri-linear strain–life curve [24] or the Fatigue Life Curve [25], if the results of stress-controlled fatigue tests in the third regime defined in [24,25] (HCF to VHCF) are included.

The cyclic stress–strain curve in Figure 8 fails to describe the cyclic material behavior in the crossway direction. While it would be expected to see an increase in stress with an increase in strain, a maximum in stress can be found at a total strain amplitude of ε_a,t = 0.25% followed by a drop until ε_a,t = 0.4% and a plateauing behavior afterwards (cf. Figure 8).

The explanation for this behavior can be found when investigating the cyclic softening behavior of EN AW-1050A H24 in respect to the total strain amplitude, as plotted in Figure 9. For total strain amplitudes of 0.2%, 0.4% and 0.6%, the upper stress of each cycle normalized by the maximum upper stress is plotted over the cycle number normalized by half the cycles to crack initiation to show the accelerated cyclic softening and the normalized “stabilized states” at N/(N_i/2) = 1.0.

When applying higher total strain amplitudes to the specimen more energy per cycle is available to move the dislocations in the crystal structure of the metal. Consequently, a strain localization and cyclic softening happens significantly more quickly, as shown by the slopes in Figure 9. According to SEP1240 [21], the cyclic material behavior must be described at 50% of the cycles to crack initiation, i.e., damage dependent. As a result, for higher total strain amplitudes, the softening already made further progress and therefore leads to a decreased stress amplitude at a higher strain amplitude as to be seen in the cyclic stress–strain curve provided in Figure 8.

3.1.3. Consequences of the Anisotropy for Fatigue Life Evaluation

In order to obtain the parameters K’ and n’ used to describe the stress–strain curve according to Ramberg–Osgood, there are essentially two ways. The first is performing a linear regression through the plastic data points, the second is using the compatibility equations [21]. While ideally those parameters should be known for both directions of extraction for the life prediction, by looking at Figure 8, it can be understood that a linear regression through the plastic data points will not work properly. The compatibility equations can be applied on the parameters of the strain–life curve provided in Figure 6 which results into the curve drawn into Figure 8 but fails to describe the actual material mechanics. Since there are no methods at hand to describe the explained behavior of the crossway-extracted specimens, it is recommended to either use K’ and n’ obtained from the compatibility equations or to use the parameters of the lengthways-extracted specimens for both directions of extraction as an approximation solution. In Figure 10, the stress–strain curve with the parameters of the lengthways-extracted specimens is plotted over the data points of the crossways-extracted specimens.

A fatigue life assessment is performed applying 0.1%, 0.2%, 0.4% and 0.8% of total strain amplitude and 500 N, 600 N and 700 N of force amplitude not considering transient effects and using the damage parameter according to Smith, Watson and Topper [27], Equation (1).

$$P_{\mathrm{S}\mathrm{W}\mathrm{T}}=\sqrt{({\sigma }_{\mathrm{a}}+{\sigma }_{\mathrm{m}})·{\epsilon }_{\mathrm{a},\mathrm{t}}·E}$$

(1)

In

Table 4. Fatigue life evaluation dependent to load-to-build direction.

Load Amplitude	Nf,exp,long (Lengthways)	Nf,exp,orth (Crossways)	Deviation (Nf,exp,long/Nf,exp,orth) − 1
εa,t = 0.1%	530,143	386,851	37.0%
εa,t = 0.2%	4257	3364	26.5%
εa,t = 0.4%	509	481	5.8%
εa,t = 0.8%	104	107	−2.8%
Fa = 500 N	175,863	303,927	−42.1%
Fa = 600 N	3793	21,768	−82.6%
Fa = 700 N	115	64	79.7%

, the experimentally assessed fatigue life using the parameters for the lengthways (longitudinal)- and crossways (orthogonal)-extracted specimens calculated with the compatibility equations are given as well as the deviation between both in percentage.

To gain a conservative fatigue life assessment, it is recommended to use the parameters of the lengthways-extracted specimens for both directions when calculating with a given force in this particular case. If a (total) strain amplitude is given, it is suggested to use the parameters from the compatibility equations of the crossways-extracted specimens for the crossways direction. Only for high loads and strains and therefore very low lifetimes, the conservative trend reverses, which needs to be considered. For fatigue life estimations beyond an approximation, it becomes evident that these fatigue life estimations with only the cyclic stress–strain curves and the strain–life curves as available data are not sufficient. For that reason, in a further study a more sophisticated approach with a damage-dependent material model is evaluated.

3.2. Resulting Challenges for Strain Measurements

3.2.1. Resulting Challenges for Strain Measurements with Surface-Mounted Extensometer

Due to its low hardness, EN AW-1050A H24 proves to be challenging when performing experimental fatigue life evaluations using the surface-mounted clip-on extensometer. The extensometer is clipped onto the surface of the specimens using one clamp for each knife. These clamps press the knives onto the surface by applying a spring force that results in a pressure on the very small contact area of the knives and the specimen surface (line load). For soft materials, this pressure is enough to create a small but sharp notch over cyclic loading, as shown in Figure 11.

In most cases, the crack initiation will take place beneath the knives due to the created notch. Thus, the measured material behavior is affected by the notch and might not correctly represent the actual behavior of the material since the behavior of a notched specimen is labeled as the material behavior of an unnotched specimen. This is especially true for the crossways-extracted specimens due to the anisotropic microstructure and resulting behavior. Since the strain localization for those specimens is more noticeable, the created notches are usually the starting point of the localization. In Figure 12, the upper lengthways-extracted specimen shows how the crack ideally initiates between the knives of the extensometer. The red arrows indicate the position of the knives of the extensometer. The crossways-extracted specimen below shows a crack initiation underneath the knife of the extensometer.

3.2.2. Resulting Challenges for Strain Measurements with the Optical Strain Measurement System

The problems associated with the knives of the surface-mounted clip-on extensometer being in physical contact with the specimens’ surfaces can be rectified by using the optical strain measurement system described above. While the optical system provides reliable results for the lengthways-extracted specimens, the strain localization in the crossways-extracted specimens leads to further challenges as presented in Figure 13.

A roughening of the surface had been detected for the crossways-extracted specimens, which can be explained by the formation of slip bands. In individual crystals cyclic flow occurs at low amplitudes of shear stress, usually in 45 degrees in respect to the loading direction. Within the material, the flow is more obstructed than on the surface leading to an inhomogeneous stress distribution due to elastic anisotropy. Therefore, one crystal offers the best conditions for slipping, initiating the formation of a slip band by creating a displacement. After the first half-cycle of loading a neighboring crystal is activated due to the transient material behavior and creates a displacement in the other direction of the slip band. That way, the first intrusion is formed [28].

While the formation of intrusions is more likely to occur due to the potential strain energy, extrusions can be formed as well. This happens if after load reversal the slip occurs at the opposite side of the slip band. When repeating the mechanisms of intrusion- and extrusion-creation, visible slip bands, as shown in Figure 13, will be created [29].

As explained in Section 2, the strain measurement with the optical systems works by tracking the movement of two areas with respect to each other. The areas are predefined by a unique pattern on the surface that is automatically searched in each frame. If a strain localization occurs in the reference areas, due to the surface roughening the appearance of the surface changes too much compared to the reference frame. As a result, the strain measurement does not work anymore or give imprecise values, which will in either way lead to an invalid test result.

3.2.3. Comparison of Data Acquisition for the Two Strain Measurement Devices

As presented in Section 3.2.1 and Section 3.2.2, the state of the art provides strain measurement methodologies that both reach their limits when performing experimental fatigue life evaluations on EN AW-1050A H24. Using the surface-mounted clip-on extensometer led to 68% of the specimens to have their crack initiation beneath the knives, making those experiments invalid according to SEP1240 [21] and VDA 239–300 [26]. At the same time, only one experiment failed using the optical strain measurement system due to the roughening of the surface. While these numbers are not statistically validated, they do show the tendency of the optical system being superior for that field of application minimizing interferences. For that reason, a contactless strain measurement system is recommended when performing experiments on EN AW-1050A H24 or materials with similar properties.

4. Conclusions

The study shows the challenges in both the experimental part of material characterization as well as in the theoretical part of fatigue life evaluation for EN AW-1050A H24.

It becomes evident that for soft (aluminum) materials, as investigated here, it is advisable to use contactless strain measurement systems to minimize invalid experiments, as performed using the optical strain control system in this investigation. Regarding the strain–life curves and the cyclic stress–strain curve in the lengthways (longitudinal) direction for description of the cyclic material behavior, further progress can be expected from using the tri-linear strain–life curve or the Fatigue Life Curve and the parameters from its compatibility equations, especially if test results at low load amplitudes are included.

Regarding the cyclic stress–strain curve in the crossways (orthogonal) direction, it can be understood that due to the strain-amplitude-dependent softening, it is advisable to use a more advanced data basis for the cyclic material characterization and strain-based fatigue life evaluation. In further studies, a damage-dependent material model is evaluated to provide a procedure that considers the transient effects and therefore describes the cyclic material behavior closer to reality than the assumptions behind the cyclic stress–strain curves according to Ramberg–Osgood, which are derived for the so-called cyclic stabilized state. If an approximation of fatigue life for loads in the crossways direction without a comparatively work-extensive material model is wanted, it is recommended to use the parameters from the compatibility equations of the crossways-extracted specimens if a strain amplitude is given. If a force is given, using the parameters from the lengthways-evaluated stress–strain curve leads to a conservative life prediction and is therefore recommended.

A further aim of this study was to explain the anisotropic behavior, especially the strain-amplitude-dependent cyclic softening of the crossways-extracted specimens on a metallographic level. For that, a NaOH as well as hydrofluoric acid etching was performed on the specimens but proved to be not capable of showing the microstructure well enough. More experience and information in preparing and etching the samples is gathered in ongoing studies with the unchanged aim to analyze the structure of these soft aluminum materials.