1. Introduction
The operational behavior of components produced by forming technology depends largely on their properties [
1]. While machining mainly affects the surface and the boundary layer of the components by up to a few micrometers, forming affects the entire material’s volume. Depending on the magnitude and direction of the stress state during forming, microstructural changes are induced. These in turn affect properties like the yield strength, ultimate tensile strength, hardness, residual stresses and forming-induced ductile damage. Ductile damage refers to the nucleation, growth and coalescence of pores during plastic deformation [
2]. This results in a deterioration in the effective mechanical properties. Modifications to the forming process, such as introducing back pressure or changing the shoulder opening angle in cold extrusion, can lead to a reduction in damage and open up the possibility of improving the mechanical properties [
3]. Full forward rod extrusion is a forming process used to produce mechanical components such as shafts and bolts. Meya et al. [
4] observed that radially superimposed stresses during bending reduce the number of blowholes. Depending on the stress level, these local material discontinuities can occur both during the various forming stages and in service up to significant length scales relevant to material or component failure. Understanding the relationships between the material properties, the microstructure and the operational behavior is therefore of great importance. This importance arises from the ability to select process parameters, considering the criteria of design and functionality, service life, and service load adaptation, which have a positive effect on component performance in terms of energy absorption in notched bar impact tests and life in high-cycle fatigue (HCF) tests. A similar observation has been made during back-pressure extrusion [
5]. Higher back pressures reduce the hydrostatic stress state and inhibit shrinkage growth [
6], which improves the fatigue life in HCF tests. However, previous research has focused on fatigue behavior after hot forming or axial and torsional fatigue. The low-cycle fatigue (LCF) behavior under axial–torsional loading after full forward extrusion has not been investigated. This combination of forming-induced ductile damage and fatigue is of great economic and environmental importance. A significant number of component failures can be attributed to LCF damage, where the cyclically induced stress exceeds the yield strength, leading to irreversible plastic deformation of steel [
7] and non-ferrous alloys [
8].
A number of parameters such as the strain rate or strain amplitude have a significant influence on LCF behavior [
9,
10]. In addition, defects and the near-surface microstructure in particular have a significant influence on LCF fatigue properties. Hardness changes due to work hardening, induced residual stresses or different phase fractions, as well as non-metallic inclusions and the general microstructure, have been investigated. Forming-induced ductile damage occurs in the form of pores in the matrix material or at non-metallic inclusions, leading to local stress increases [
11,
12]. The resulting material or component failure under service conditions can be attributed to inhomogeneities [
13], which have been addressed by various authors in their studies on the correlation between fatigue performance and defects, as the defects are determined by the threshold condition for small crack nuclei at the defect tip [
14,
15]. In addition, a correlation has been shown between the torsional fatigue limit and the threshold for non-propagating branched cracks. This leads to a multiplication of dislocations and a reduction in the average grain size, which affects the grain boundaries associated with the intervening particles. As a result, dislocation blockage occurs, leading to significant microstructural changes that can have a significant impact on fatigue properties [
16]. A fundamental analysis of the effects of room-temperature pre-straining and subsequent LCF loading on steel was performed by Sonsino [
17]. Other studies have examined the influence of cyclic loading on material properties after severe plastic deformation processes such as equal channel angle pressing [
18]. These included multi-stage fatigue tests of up to 15,000 cycles on hot-rolled steel sheets and fatigue studies of damage-induced effects after hot-rolling with up to 25,000 cycles [
19,
20]. Teschke et al. performed fatigue tests in the range of 10 to 10
6 cycles on hot-rolled sheets [
21]. Moehring and Walther [
22] investigated the influence of forming-induced ductile damage in the LCF range under torsional loading. The influence on forming-induced ductile damage in the LCF range under axial loading was investigated by Langenfeld et al. [
12], who showed a significant impact of forming-induced damage on the fatigue performance. In other studies, this influence was also confirmed for axial–torsional loads [
23].
Therefore, it is important to extend metallographic characterization methods and combine them with mechanical testing methods as well as structure-sensitive measurement techniques to investigate the influence of forming-induced ductile damage on performance. In addition, non-destructive investigation methods such as computed tomography (CT) are not available for condition characterization due to the microscopic size of the initial forming-induced ductile damage and the density of the steel being examined [
24]. For this purpose, focused ion beam scanning electron microscopy (FIB-SEM) investigations with atomic force microscopy (AFM) measurements and a self-developed 3D volume model approach are correlated with the mechanical testing methods to better understand the damage mechanisms. A similar approach has been used by Otto et al. [
25] to visualize brittle phases in solder joints to better understand their distribution and the influence on the fatigue and corrosion properties. Other work has focused on carbon-based materials [
26] or the influence of voids and inclusions in steels [
27], where 3D microstructure imaging has provided new insights into material behavior. Regarding the suitability of the 3D volume reconstruction, the work of Osuch et al. [
28] shows that 3D volume reconstruction is suitable for the visualization of carbides in steel materials. Furthermore, the influence of the selected detectors of the SEM on the images and the subsequent segmentation is relevant.
Resistometry-based methods will also be used to investigate fatigue damage assessments in terms of the change in electrical resistance due to pore formation or growth [
29]. The long-term goal is to quantify the various influencing factors and to separate the damage mechanisms, especially with regard to forming-induced ductile and fatigue-induced damage. The coupling of microstructural analyses and performance characterizations with respect to fatigue properties, as well as the goal of separating damage mechanisms, is shown in
Figure 1.
2. Materials and Methods
2.1. Material
The fatigue specimens were fabricated from cylindrical billets of case-hardening steel 16MnCrS5 (DIN 1.7139, AISI/SAE 5115) via full forward rod extrusion. The base material, supplied by Georgsmarienhuette (Georgsmarienhuette, Germany), consisted of rolled and drawn cylindrical rods with a ferrite–pearlite microstructure. The chemical composition is shown in
Table 1.
No additional heat treatment was applied to the material. In the extrusion process, cylindrical billets with an initial diameter (d0) of 30 mm and length (l0) of 71 mm were forced through a die to reduce their diameter, resulting in an extrusion strain (εextr) of 0.5. This specific extrusion strain induced hydrostatic stress in the forming zone, leading to the development of forming-induced ductile damage. The diameter of the billet after extrusion (d1) was observed to be 23.4 mm.
To control the amount of damage developed, different shoulder opening angles were considered. Specifically, shoulder opening angles of 2α = 90° and 2α = 30° were used during the extrusion process. The choice of the shoulder opening angle influences the hydrostatic stress states at the center axes, thereby affecting the level of the forming-induced ductile damage. The full forward rod extrusion process was performed on a triple-acting hydraulic drawing press (HZPUI, SMG, Mannheim, Germany) with a maximum punch force of 2600 kN and a punching speed of 10 mm/s. The billets were coated with Beruforge 179 (Carl Bechem, Hagen, Germany), a MoS2-containing coating lubricant, and the extrusion process was conducted at room temperature.
The specimen geometry was designed to maintain comparable levels of strain hardening and residual stress even when subjected to different forming process parameters [
3]. The specimens were produced by machining the formed semi-finished products. Hering [
3] demonstrated that machining does not significantly affect the damage state or the distribution of forming-induced ductile damage within the specimen. The geometry of the specimens used in this study, along with the extraction position of the specimens from the workpiece after full forward rod extrusion, is detailed in
Figure 2.
2.2. Fatigue Testing Setup and Methods
In order to separate the ductile-forming-induced damage from the fatigue-induced damage and to analyze their interaction, constant amplitude fatigue tests were performed on an Instron 8801 servo-hydraulic fatigue-testing system (Instron, Norwood, MA, USA) at a stress amplitude of σ
a = 380 MPa with a stress ratio of R = −1 and a test frequency of f
ax = 10 Hz (
Figure 3a). Axial–torsional fatigue tests were performed using a Walter + Bai, LFV-T250 T2500 HH servo-hydraulic axial–torsional testing system (Walter + Bai, Loehningen, Switzerland). These tests were performed at a test temperature of T = 20 °C. The total strain amplitude was set to ε
a,t = 1.0%, with a strain ratio of R = −1 and a test frequency of f
axto = 0.01 Hz, along with superimposed torsion controlled by an angular amplitude of θ = 10° with a test frequency of f = 0.01 Hz (
Figure 3b). The specimens were clamped using hydraulic chucks, with care taken to ensure that the specimens were clamped with a specimen clamping area of at least 25 mm and in the same position on the chuck to prevent twisting and to ensure comparable test conditions.
In order to record damage development during fatigue testing, resistivity-based, structure-sensitive methods in the form of direct current potential drop (DCPD) measurements were used. These were conducted with a newly developed experimental setup which allows for sensitive reproducible measurements due to a constant contact pressure and a constant measuring point distance. The electrical current was applied using a Keithley 2602B specific source meter (Keithley, Cleveland, OH, USA) and the electric voltage was measured by using a Keithley 2182A nanovoltmeter (
Figure 4a).
The innovative combination of a source meter and voltage measurements allows for the implementation of the ‘delta mode’. In this mode, a highly constant current is applied for 0.09 s, followed by a polarity reversal of the signal, which triggers a nanovoltmeter reading at each polarity (
Figure 4b). Therefore, any thermoelectric offsets caused by contact can be avoided. The sampling rate of the resistance-based measuring system is 5.74 Hz. The self-developed software used for data acquisition and system control was specifically programmed for this experimental setup and measurement application. The height of the current is important for electrical resistance measurements. A high current strength leads to an improved measurement accuracy, but simultaneously promotes specimen heating. Through the one-factor-at-a-time method (OFAT), all important adjustable measuring parameters were determined. In addition to the current intensity, the duration of the delay (
Figure 4b), which describes the delay of the voltage measurement after the polarity reversal, and the number of power line cycles (NPLC), which describes the sampling rate of the grid cycles of the general power grid, were considered (
Table 2). An optimal current of I = 0.8 A, an NPLC of 1 and a delay of 0.02 s were determined, which represents a reliable compromise for the measurements [
29].
2.3. Microstructural Investigations
The following sections describe the methods used for microstructural analyses. This includes image segmentation, quantitative microstructure analysis, and 3D model generation. In addition, electron channeling contrast imaging (ECCI), scanning transmission electron microscopy (STEM) and AFM methods are briefly explained. For these investigations, a Zeiss Crossbeam XB 550 L FIB-SEM (Zeiss, Oberkochen, Germany) was used.
2.3.1. Image Acquisition
For 3D model generation, thin sections were automatically ablated using an FIB, with an image taken every 20 nm using both Inlens and SE detectors. The milling process was carried out using an FIB acceleration voltage of 30 kV and a sample current of 200 nA. For subsequent fine milling, an acceleration voltage of 5 kV and a sample current of 30 pA were used. By selecting the above parameters, significant heating of the sample can be excluded. A layer width of approx. 9 µm was chosen to obtain a meaningful volume in terms of the pore distribution and sulfide structure distribution while minimizing preparation artifacts due to FIB preparation. Since it is not possible to prepare an image free of artifacts, these were removed with the help of deep learning AI image segmentation.
2.3.2. AI Image Segmentation
Zeiss Zen Analyzer software version 3.5.96.06000 (Zeiss, Oberkochen, Germany) and the Zeiss Zen Intellesis module version 3.5.96.06000 (Zeiss, Oberkochen, Germany) were used for image segmentation. Using a deep learning algorithm, models have been trained to automatically and reliably segment the pores from the SEM SE images. This facilitated the automatic identification and segmentation of the pores (
Figure 5a). Using Zen Analyzer’s built-in analysis capabilities, a range of valuable information such as the number, area, coordinates, and geometric properties (e.g., diameter or elliptical semi-axis) could be directly extracted without the need for a specially programmed algorithm.
For sulfide segmentation, a similar approach was used with the Zeiss Zen Analyzer software and the Zen Intellesis module version 3.5.96.06000 with the advantages and features mentioned above. However, SEM Inlens images were used for the segmentation of the sulfide structures instead of the SEM SE images (
Figure 5b). To validate the segmentation results, the segmented images were independently compared with the original images by several individuals. Additionally, the image stacks were processed independently by several individuals to compare the quantitative results. The order of the image stacks was also changed to eliminate errors in the algorithm.
2.3.3. Three-Dimensional Model Generation
Figure 6 shows the first steps in processing the FIB-SEM SE images. A region with a low density of artefacts was selected from the 510 images to construct a 3D model measuring 8.54 × 8.77 × 10.18 µm
3. Automated image cropping was performed using the open-source software FIJI ImageJ version 1.54f. The pores were then segmented from the image stacks using Zen Intellesis version 3.5.96.06000. The pixels were then converted to voxels using FIJI ImageJ and several small mesh models were generated and saved in .obj format. A mesh enabler was used to create a solid model from these mesh structures, and all the parts were assembled into a final model using Autodesk Inventor computer-aided design software version 2024 (Autodesk, San Francisco, CA, USA). The model was then rendered to improve its visual representation. The same workflow was used to create a 3D model of the sulfide distribution. This method is also described in detail in [
25] for precipitation reconstruction in brazed joints.
2.3.4. Electron Channeling Contrast Imaging (ECCI) and Scanning Transmission Electron Microscopy (STEM)
For the ECCI investigations, the specimens were prepared according to the specific requirements. An acceleration voltage of U = 25 kV and a specimen current of I = 1 nA were used. Dislocations can be visualized by a specific orientation of the electron channels. Other defects such as pores can also be clearly seen from these images and can be correlated with the other methods. For STEM studies, the specimens were prepared according to the specific requirements for STEM analysis (
Figure 7a) by preparing the specimens to a thickness of less than 100 nm for transillumination (
Figure 7b) using the FIB.
The milling process was performed with an FIB acceleration voltage of 30 kV and a sample current of 200 nA. For the subsequent fine milling of the FIB, an acceleration voltage of 5 kV and a sample current of 30 pA were used. For the microscopic examination of the FIB lamella, an acceleration voltage of U = 30 kV and a sample current of I = 1 nA were used. These parameters were found to be optimal for imaging pores and dislocations with the highest possible resolution and contrast. In order to correlate the results of the STEM studies with the 3D SE studies, the images were then segmented and analyzed in ZEN-Intellesis with respect to the pores.
2.3.5. Atomic Force Microscopy (AFM) Measurements
For an in-depth assessment of the shape, topography and separation of pores and cracks inside sulfides in the nanometer range, AFM investigations were performed for selected specimens, using a piezo-based AFM (LiteScope, Nenovision Brno, Czech Republic). AFM investigations were performed in tapping mode with Akiyama probes as single-pass measurements. The topography was studied with setpoints of around 5 Hz at a resonance frequency of the probe of around 45 kHz. AFM investigations were performed under vacuum inside a Crossbeam XB 550 L FIB-SEM (Zeiss, Oberkochen, Germany). This ensures stable oscillation at resonance and also helps to find the region of interest. After or even during acquisition of AFM data, SEM images can be acquired using this setup. Post-processing of AFM data was achieved using the free open-source software Gwyddion Version 2.63.
4. Conclusions and Outlook
The present study has shown that forming-induced ductile damage has a significant influence on the fatigue properties of full forward rod extruded case-hardened steel 16MnCrS5 (DIN 1.7139, AISI/SAE 5115). This has been demonstrated for both stress-controlled uni-axial and axial–torsional strain-controlled fatigue tests. The damage mechanisms were validated by a microstructural analysis using AFM, ECCI and STEM. In addition, advanced metallographic techniques were used to visualize pores and sulfide distributions in 3D for the first time to infer the shape and distribution of defects in the form of pores and MnS inclusions.
With respect to axial–torsional stress, cyclic softening and tensile–compressive asymmetry were observed, which can be attributed to the Bauschinger effect. This was confirmed by ECCI and STEM images of the dislocation structures in the fully forward extruded material, with an increased dislocation density occurring in both ferrite and pearlitic phases. Under cyclic loading, the dislocation density decreases due to mechanisms such as dislocation annihilation, which is reflected in the cyclic softening of the material. It has also been shown that the cyclic axial–torsional fatigue behavior can be improved by adjusting the forming parameters during full forward rod extrusion and the resulting forming-induced ductile damage. The specimens with a shoulder opening angle of 2α = 30° required an approximately 24% higher number of load cycles to failure, Nf, compared to a shoulder opening angle of 2α = 90°.
For uni-axial stresses in the HCF range, a resistometry-based method was used to separate forming-induced ductile and fatigue damage. The local deformation behavior was analyzed using DIC. By using the newly developed experimental setup, the measurement scatter of the resistometry-based measurement was minimized by more than a factor of 102 compared previous measurements using conventional DCPD systems. The electrical resistance was correlated with the deformation behavior, which was measured using a proprietary test setup. The cyclic softening that occurs under axial–torsional stress could also be observed from the plastic strain amplitude in the axial-tested specimen. Thus, the resistometry-based method provides a good opportunity to quantify forming-induced ductile damage and to record damage accumulation in situ during the fatigue test to detect damage mechanisms such as pore coalescence or crack propagation. The information provided by the non-destructive measurement technique was used to validate the quantitative determination of the void content, and this technique offers a fast and cost-effective method for quantifying the degree of forming-induced ductile damage in the future.
The influence of pores and MnS inclusions on crack initiation and propagation was emphasized by fractographic studies, as these play a significant role, especially in areas close to the surface. High-resolution AFM measurements of MnS inclusions confirmed the hypothesis that full forward rod extrusion leads to fracturing of the MnS inclusions and thus to pore formation. In addition, AFM was used for the first time to visualize the formation of nanopores in fully forward rod extruded 16MnCrS5 (DIN 1.7139; AISI/SAE 5115). These were found preferentially in the pearlite phases between the cementite clusters, and also in the ferrite. This was confirmed via ECCI and STEM.
By combining FIB-SEM, deep learning AI image segmentation and CAD, high-resolution 3D models of the MnS inclusions and pore distribution were generated for the first time in the full forward rod extruded case-hardening steel 16MnCrS5 (DIN 1.7139; AISI/SAE 5115). It was found that the pores are distributed throughout the representative volume and not, as previously thought, only in the sulfide structures. This is mainly due to previously undetected nanopores in both the ferrite and pearlite phases, as confirmed via ECCI, STEM and AFM measurements. By visualizing the distribution of the MnS inclusions, this influence could also be considered separately from the pores with respect to crack initiation and propagation, as a significant influence of these was determined in the fractographic investigations. The platelet-shaped sulfide structures also show a microstructural notch effect, as the brittle phases cannot compensate for the deformation of the matrix. By combining the two 3D models, it was possible to determine that the largest pores were located in and around the MnS inclusions. Considering the pore distribution in the present phases, it can be seen that about 1.9% of the pores occur in MnS inclusions, and thus the probability of pore formation within a MnS inclusion is about ten times higher than in the remaining representative volume or in the ferrite and pearlite phases, since the sulfide structures account for only 0.27% of the representative volume. Mechanisms such as decohesion between MnS inclusions and the surrounding ferrite–pearlite matrix could not be considered in this work because only the intersections between pores and sulfides were compared.
The segmentation of MnS inclusions and pores thus provides new insights into their distribution in the volume as well as the separation of damage mechanisms. This will allow future conclusions on crack propagation and crack initiation, especially for complex axial–torsional load cases. With this information, future forming processes must be designed to minimize forming-induced ductile damage in order to improve the tensile as well as fatigue properties, as it has a significant impact on them. This concerns both the suppression of pore formation and pore coalescence. The combination of in situ damage measurements during fatigue tests with the developed resistometry-based measuring system and the determination of pores, MnS inclusions and crack development in 3D can be used in the future to carry out damage-controlled forming processes.
Further investigations of short-term crack growth and defect development or damage accumulation due to different fatigue states are the next logical steps to develop a deeper understanding of the influence of initial forming-induced ductile damage using the 3D models. In the future, in situ tests inside the SEM will be performed to evaluate the microstructural damage development. In addition, the studies will be extended to the very high cycle fatigue range using a Rumul Gigaforte 50 to cover all application-relevant cycle ranges.