Effects of Holding Time on Thermomechanical Fatigue Properties of Compacted Graphite Iron Through Tests with Notched Specimens

The material investigated in this work is CGI with a pearlitic matrix. The nominal chemical composition of the material is given in Table I.

Figure 1 illustrates the complex morphology of compacted graphite iron. The microstructure is composed of graphite particles embedded in a pearlite matrix, with a variety of lamellar structures and a small fraction of ferrite.

TMF experiments were conducted using a 25 kN MTS servohydraulic fatigue machine (MTS Systems Corporation, Eden Prairie, MN) capable of imposing independent temperature and strain profiles on a test specimen. Solid, smooth, dogbone specimens were taken out of cylinder heads and machined to a final smooth cylindrical gauge length of 22.0 ± 0.25 mm and a diameter of Ø6.00 ± 0.025 mm. The reduced section of the specimens was machined by a turning operation, with the final two steps removing only 0.05 mm of the thickness to prevent internal stresses near the surface. In a next step, specimens were provided with a circumferential notch in the middle of the gauge length, with a depth of 0.2 mm and a 0.04 mm tip radius (cf. Figure 2). Cracks were always found to initiate from graphite particles at or near the notch tip, and in this respect, a possible influence of surface roughness on TMF lifetime (as would be the case when using smooth specimens) can be excluded when using notched specimens.

A high-frequency induction generator was used for heating. Cooling was accomplished by blowing compressed air from three sides onto the specimen and by thermal conduction into the water-cooled specimen grips. The temperature was measured and controlled by two K-type thermocouples that were pressed onto the surface of the specimen. One thermocouple was placed at a location approximately 0.5 to 1.0 mm above the notch. The other thermocouple was used to control the temperature located at 0.5 to 1.0 mm below the notch.

The contact pressure of the thermocouples is regulated by means of a spring configuration. This is found to lead to a good contact with the specimen surface and an accurate temperature measurement. The temperature gradient observed is within the requirements of the European code of practice for strain-controlled thermomechanical fatigue testing.[11]

The axial strain was measured via an air-cooled, high-temperature, ceramic rod extensometer with 12 mm gauge length. TMF tests were performed under full constraint conditions, meaning that the total strain measured by the extensometer was kept constant. By cycling the temperature between 323 K and 693 K (50 °C and 420 °C), OP TMF loading was accomplished. This TMF test procedure is intended to replicate the thermal and mechanical conditions in the valve bridge area of cylinder heads. The heating and cooling rates were 9 K s^–1 and 6 K s^–1, respectively.

The thermal cycles are shown schematically in Figure 3. Holding times were introduced to represent the in-service conditions that valve bridges are subjected to. However, holding times were selected to have reasonable testing times. Three different holding times at the maximum temperature 693 K (420 °C) of 30 seconds, 480 seconds, and 1800 seconds were chosen in combination with 30 seconds for the minimum temperature 323 K (50 °C). Using different holding times at 693 K (420 °C) allows stress relaxation to occur to different degrees. The choice of holding times is based on previously performed compressive stress relaxation experiments at 693 K (420 °C).[7]

The microstructural evolution during a typical TMF heating stage was evaluated by performing EBSD measurements in order to find the mechanisms associated with the stress relaxation phenomena. To this purpose, a CGI sample with a rectangular cross section of 9 × 7 mm², and a height of 17 mm was fabricated, which is designated as a parallelepipoid-shaped specimen. The small size of the specimen accommodates the positioning of the specimen within the confined space of the scanning electron microscope (SEM) sample chamber.

The EBSD measurements were performed for three conditions of the CGI material, being (1) as-cast (point A in Figure 3), (2) heated in total constraint but without holding time (point B in Figure 3), and (3) heated in total constraint with a 1800 seconds holding time (point C in Figure 3).

To enable the EBSD measurements, one side of the parallelepipoid-shaped specimen (9 × 17 mm²) was prepared with a final polishing step using OPS (0.25 μm), manually polishing the surface for 40 minutes. In the center of the polished surface, three microindentions were made to mark a specific location. At this location, ESBD measurements were performed, characterizing the situation of the as-cast state of the material (point A in Figure 3).

To obtain the condition represented by point B of the CGI material (Figure 3), two cylindrical pushing rods (Ø = 12 mm) were clamped in the grips of the TMF fatigue machine. The parallelepipoid-shaped specimen was placed between the ends of the cylindrical pushing rods and heated to 693 K (420 °C) (at the rate of 9 K s^–1) under total constraint. When reaching the temperature to 693 K (420 °C), a peak compressive stress was reached and the test was stopped, after which the specimen was removed. One side of the specimen was then polished and analyzed by ESBD, giving microstructural information in the condition represented by point B in Figure 3. The ESBD measurements were done at the same location as was done for the as-cast condition.

Finally, to characterize the condition represented by point C (Figure 3), the specimen was again placed in the TMF machine and heated to a temperature of 693 K (420 °C) (at the rate of 9 K s^–1), after which the peak compression level of the condition represented by point B was applied, while maintaining the temperature of 693 K (420 °C). At the point of reaching the peak compressive stress, the strain was kept constant for a period of 1800 seconds, allowing stress relaxation to occur (point C in Figure 3). After this holding time, the test was ended and the specimen surface was repolished for observation by EBSD at exactly the same location as studied before. The EBSD measurements were done on a scanning area of 430 × 300 μm² with a step size of 1.4 μm.

The results of tests with different holding times are listed in Table II. In this table, the TMF lifetime N ₁₀ is defined as the number of cycles at which the maximum (tensile) stress (σ _max) has dropped by 10 pct relative to its maximum value.

The TMF lifetime results for different holding times show a clear reduction in lifetime, at prolonged holding times (Table II). Relative to the lifetime at 30 seconds holding (i.e., 213 cycles), the lifetimes for the 480 seconds and 1800 seconds holding times show a reduction of 28 pct and 57 pct, respectively. As reported in our previous work,[7] for a holding time of 480 seconds, a reduction in lifetime of the same order was found for the smooth specimens (i.e., 35 pct reduction in N ₁₀).

The effect of holding time can also be evaluated in relation to the developed stress levels during the progress of the TMF experiment. In Figure 4, a typical development of the minimum, mean, and maximum stress is shown as a function of elapsed TMF cycles for the holding times tested. The development of stress levels can be divided into three distinct regions.[3]

The first region is characterized by increases in minimum and maximum stress level for subsequent TMF cycles. In the second region, the maximum stress level remains more or less stable. Finally, in the third region, a drop in stress level is observed. For the total constraint tests applied here, the drop in maximum tensile stress is induced by a decrease in specimen stiffness, associated with the opening of a macroscopic crack. Therefore, the drop in stress level can be used to detect a macroscopic crack and can be applied as a criterion to define TMF lifetime. In Reference 3, the number of cycles at a 25 pct drop in stress level (i.e., N ₂₅), is suggested as the TMF lifetime. However, in our previous research,[7] it was found that a 10 pct drop in stress is a more consistent criterion (i.e., N ₁₀), leading to less scatter. In Section III–B, the underlying reasons for using N ₁₀ as the lifetime criterion are discussed in more detail. For the TMF lifetime results presented in Table II, the N ₁₀ lifetime criterion was adopted.

TMF testing under total constraint is equivalent to cyclically inducing a fixed amount of mechanical strain. This amount is equal (and opposite in sign) to the value of the thermal strain, which was 0.6 pct for the TMF tests under consideration, corresponding to the temperature range of 323 K to 693 K (50 °C to 420 °C).

Therefore, during the entire duration of the TMF tests, the CGI material is exposed to strain changes of 0.6 pct in the tensile direction during cooling and the compressive direction during heating. The resulting stress levels largely depend on the initial stress levels from which individual strain excursions start. Therefore, the development of tensile and compressive stresses during a TMF test, under conditions of total constraint, are mutually dependent.

As can be seen in Figure 4, for all holding times, the mean stress level shifts towards a tensile value during the progress of the TMF test. This is typical for the out-of-phase nature of the TMF loading condition under total constraint. One reason for the shift is that the material exhibits a lower yield strength in the compressive region (at high temperature) compared to the value in the tensile region (at low temperature). In addition, the compressive stress may decrease during the holding time at high temperatures because of stress relaxation. As a consequence, the strain cycle associated with cooling starts at a relatively low compressive stress and will, thus, produce a relatively high tensile stress. Moreover, the subsequent heating cycle starts from this relatively high tensile stress, resulting in a relatively low compressive stress. This process is repeated until the ratio of tensile and compressive stresses settles to a certain stable value.

The stress–mechanical strain hysteresis can also give useful information about the TMF behavior. The mechanical strains calculated as a function of stress for one of the specimens (N ₁₀ = 96, cf. Table II) with a 1800 seconds holding time are shown in Figure 5. The black circle indicates the parts of the TMF cycles where the compressive stress relaxes. It can also be observed that the lower the compressive stress, the higher the tensile stress in the next cycle, confirming the interdependency of tensile and compressive stress levels, as described before.

With progressing TMF cycles, the amount of stress relaxation at 693 K (420 °C) gradually decreases and the shape of the hysteresis loop becomes narrower, indicating a reduction of cyclic plasticity.

In Figure 6, the stress is plotted as a function of time for two cycles (55 and 56) for the specimen discussed above (Figure 5). The stress relaxation at 693 K (420 °C) for these two cycles is approximately 22 MPa. On the other hand, the increase of σ _max by changing the holding time from 30 seconds to 1800 seconds is approximately 24 MPa (cf. Table II). Therefore, it can be inferred that the higher stress relaxation occurring during 1800 seconds holding time is responsible for the increase of σ _max.

The evolution of the amount of stress relaxation during the holding time under compression with the number of TMF cycles is plotted for all three holding times in Figure 7. This stress relaxation is calculated as the stress change from the moment the temperature has stabilized at 693 K (420 °C) (after 10 seconds exposure) until the last moment of exposure to 693 K (420 °C). As can be seen, the stress relaxation for the 480 seconds and 1800 seconds holding times show a similar trend, starting at a high level of about 50 MPa and reducing to a level of about 22 MPa after 50 cycles, and then remaining at that level. The stress relaxation for the specimen with 30 seconds holding time has a much lower value of about 10 MPa for the entire duration of the TMF test. Relaxation of stresses is thought to be related to viscous phenomena and recovery processes.[3] This is investigated by EBSD techniques and will be discussed in Section III–D.

In Figure 5, the decrease in cyclic plastic strain can already be observed, but in Figure 8, the reduction of cyclic plasticity is presented more clearly. The calculated plastic strain range vs the number of elapsed cycles is plotted for all three holding times. The plastic strain range is calculated by subtracting the elastic strain range from the mechanical strain range. The maximum and minimum elastic strains needed to calculate the elastic strain range are calculated as the maximum and minimum stresses divided by the Young’s moduli at the minimum and maximum temperature, respectively.

All holding times show a similar trend of a rapidly decreasing plastic strain range in the first few cycles, continuing to decrease at a somewhat slower pace in the first tens of cycles, saturating at a nearly constant value from 50 cycles onwards, and increasing somewhat at the end of the test. The reduction in cyclic plasticity is in accordance with an increase in yield strength found at increased numbers of cycles. This increase in strength indicates the occurrence of cyclic strain hardening. It should be noted that at the start of the test, the total constraint condition was applied at the minimum temperature, and consequently, the first TMF cycle produces a large compressive plasticity in the CGI. In the 2nd cycle (the first complete cycle), the plastic strain range Δε _p is already significantly lower.

The levels of cyclic plastic strain differ for the different holding times, with the highest cyclic plasticity for the longest holding time. It seems likely that a high cyclic plasticity contributes to the shorter lifetime at prolonged holding times.

It was found that the load drop related to the development of a macroscopic crack is not always consistent. Sometimes, a considerable degree of load drop occurred before final fracture, while in some other cases, sudden fracture took place without any load drop. It should be mentioned that the (final) main crack is typically asymmetrical (cf. Figure 9) and initiates randomly relative to the position of the extensometer. Therefore, the extensometer could be positioned at the crack mouth location or opposite to it. To analyze this matter, we marked the extensometer position on the tested specimens. From this, we observed a relation between the aforementioned load drop behavior and the position where the main fatigue crack developed relative to the extensometer. Typically, when the crack mouth of the final crack was on the same side as the extensometer, a substantial load drop occurred before failure, and the final crack lengths were relatively large.

The effect of the extensometer position can be related to the asymmetry of the dominant fatigue crack, which introduces a bending component in the specimen. If this crack develops at the extensometer position, the displacement experienced by the extensometer during tensile loading is relatively large, due to opening of the crack. In this case, the control system applies only a relatively low tensile load on the specimen to maintain the total constraint condition at low temperature. Since the crack is now loaded to a lesser extent, it can grow to a larger size before it becomes critical and leads to failure. More cycles are needed for failure to occur, during which the maximum tensile load will drop substantially as a result of the decreasing specimen stiffness. Conversely, when the dominant crack develops diametrically opposed to the extensometer, an opposite observation can be made. The extensometer now experiences relatively low displacements, and correspondingly high external tensile loads are required to maintain the total constraint condition. Failure will occur at a shorter length of the dominant crack, involving only a small load drop.

In order to reduce the influence of the extensometer position on TMF lifetime, the failure criterion for TMF lifetime is defined as the number of cycles at a 10 pct load drop. At such a small load drop, the crack is still relatively short and the bending effect will be less pronounced; consequently, the effect of the extensometer position will be effectively reduced.

In our previous work,[7] the failure times were calculated successfully by applying the Paris' fatigue crack growth lawwhere a is the crack size (m), N is the number of cycles, ∆K _I is the stress intensity range (MPa√m), and C and m are material-dependent parameters. The stress intensity range is corrected for crack closure at compressive loads by simply assuming ∆K = K _max. In the calculation, it is assumed that a crack initiates immediately from the notch, i.e., the depth of the circumferential notch is taken as the initial crack length. Assuming specific values for C and m, the number of cycles is calculated for a first crack extension of 0.01 mm around the entire circumference of the specimen. This process is repeated for subsequent steps of 0.01 mm, adjusting ∆K in each step in accordance with the increased crack length. By applying this numerical procedure, the number of cycles N is calculated for reaching 1.5 mm of crack extension, which corresponds to 50 pct of the sample diameter. This limit is considered here as the critical crack length for sample failure. Thus, lifetime is calculated on the basis of crack growth kinetics, i.e., the Paris's law, rather than by the crack-initiation rate. In the case of a notched sample in CGI, this condition is likely to be fulfilled. The values for ∆K ranged from 10 MPa√m to 40 MPa√m, representing the levels of ∆K for the initial crack length (0.2 mm) and final crack length (1.5 mm), respectively.

The model accounts for the maximum stress level developed, so the theoretical effect of different maximum stress levels found at different holding times can be evaluated. In Table II, the calculated values for the parameter C are shown for which the calculated and measured lifetimes are the same at each holding time. The m parameter value is assumed to be 5, as used in our previous work.[7] It must be noted that the C value for the 30 seconds holding time 7.3 × 10^–12 is not the same as found earlier, i.e., 9.5 × 10^–12.[7] The difference is probably related to the fact that the tested specimens were sampled from different cylinder heads with different batch properties.

As can be seen in Table II and in Figure 4, increasing the holding time leads to higher maximum stress levels and shorter lifetimes. In order to see the isolated effect of holding time without the influence of increasing stress level, a corrected lifetime is calculated using the same maximum stress level for all holding times, while using the C values mentioned in Table II. This stress level was chosen to be the value that occurred for the shortest holding time, i.e., 324 MPa. The corrected calculated lifetimes (N _Paris) are shown in the last column of Table II.

In Figure 10, the experimental N ₁₀ values and the calculated N _Paris values are plotted for the three holding times under consideration. As can be seen, the reduction in lifetime is less pronounced for the calculated results. Therefore, according to the calculations of the corrected lifetimes, the higher tensile stress levels that developed during the 1800 seconds holding time tests do not completely account for the measured reduction in lifetime. This might point to an additional effect related to stress relaxation during holding at high temperature.

The evolution of the crystal misorientation angle in the range between 0 deg and 10 deg is an indication of local strains induced in the microstructure by the compressive stresses involved in these thermomechanical treatments. Such stresses produce plastic deformation, involving dislocation glide in the metal matrix. The dislocations remaining in the microstructure configure themselves in cells or subgrains, which will give rise to small orientation gradients that can be measured accurately by orientation contrast microscopy (EBSD with angular resolution of the order of less than 1 deg).

Table III contains the area fraction (pct) of low-angle grain boundaries (LAGB) measured with EBSD at a specific location of the parallelepipoid-shaped CGI specimen in three conditions. As can be seen Table III, the as-cast condition gives 64 pct of LAGB area fraction, whereas after heating to 693 K (420 °C) in total constraint (1st step), this has increased to 73 pct, clearly showing the effect of plasticity to which the sample was subjected. However, after a holding time of 1800 seconds (2nd step), the LAGB area fraction has regressed to the initial value. Alternatively, this trend can also be observed in Figure 11, exhibiting the image quality maps of the three conditions under consideration (the image quality factor represents a quantitative measure of the band contrast in the EBSD pattern, and therefore, it is indicative of the local density of crystal defects). In these maps, LAGBs (2 to15 deg) together with high-angle grain boundaries, 15 to 180 deg) are indicated by red (thin) and blue (thick) lines, respectively.

Table III lists the average KAM values for 1st nearest neighbors for the entire scan (excluding the graphite phase) observed in the three conditions considered. It should be noted that the average KAM value is sensitive to the plastic strain occurring after heating the sample from 323 K to 693 K (50 °C to 420 °C), as its value increases from 0.62 deg to 0.73 deg. After a holding time of 1800 seconds at this temperature, the average KAM value drops to 0.68 deg, which is precisely half way the value of the 1st step and the as-cast condition, cf. Table III.

Figure 12 also displays the KAM distributions between 0 and 5 deg for the three conditions under consideration. It can be observed that there is a slight difference in the tail end of the distribution, in the range of 1 to 3 deg, between the samples that were subjected to a plastic strain of 0.4 pct (imposed by the mechanical constraint) and the initial sample in as-cast condition. After holding for 1800 seconds at high temperature, the peak position or the distribution average is affected by the stress relaxation during isothermal holding time; i.e., one may readily assume that the sample was subjected to stress relieve recovery. By close inspection of the KAM distribution profiles, it can be observed that this gave rise to a slight shift of the distribution peak from 0.44 deg at the start to 0.31 deg at the end of the holding. This is also reflected in a drop of the average KAM value from 0.73 deg to 0.68 deg. This may be associated with the fact that the deformation substructure, established by plastic deformation, is predominantly reflected in the high tail of end the distribution, whereas statistical scatter of short-range orientation gradients more affects the bulk of the distribution. In this sense, one may draw the conclusion that the relaxation processes have left the plastic deformation structure unchanged but have reduced to some extent the local orientation scatter, corresponding to the well-known recovery of statistically stored dislocations in the metal matrix.[12] Whether these minute changes are truly representative for the underlying microstructural changes needs to be confirmed in further study.