Mechanics of Materials
Statistical Analysis of Influential Factors on the Stability of Retained Austenite in Low Alloy TRIP Steel
2022 Volume 63 Issue 5 Pages 693-702
Details
2022 Volume 63 Issue 5 Pages 693-702
The factors representing the stability of retained austenite (γR) in low-alloy TRIP steel at a plastic strain of up to 0.04 were statistically analyzed by evaluating the size, strain level, crystal orientation, morphology, and precipitation site of a large number of γR grains (>1000). The ratio (Rarea) of the area of γR after deformation to that before deformation was adopted as a representative parameter for the stability of γR. The frequency of low Rarea in fine γR was significantly high, indicating that they were unstable. Moreover, it was indicated that γR in the high-strain regions tended to be unstable, whereas γR in the bainite region was stable. The crystal orientation and aspect ratio did not correlate with Rarea in TRIP steel. The strain level and precipitation sites in coarse γR were uncorrelated with Rarea, indicating that the size of the γR has a greater effect on its stability. The stability of γR was not dominated by the carbon concentration but was affected by other factors. The size of γR may represent the stress conditions at the γR/ferrite or bainite boundaries, which is why this factor represents the stability of γR in the TRIP steel at a plastic strain of up to 0.04.
Fig. 13 Schematic illustration of the relationship among the stability of γR, stress concentration and each representative factor.
The transformation-induced plasticity (TRIP) effect improves both the strength and ductility of metal materials, and TRIP steels are widely used in automobile applications owing to their high energy absorption.1–3) The deformation-induced martensitic transformation of metastable retained austenite (γR) in TRIP steels is the key to the TRIP effect. Sugimoto et al.4–6) investigated the mechanical properties of TRIP steels and the effect of the stability of γR on these properties. They demonstrated that the highly stable γR transforms into martensite at a late stage of deformation, providing a good balance between strength and ductility. Therefore, the stability of γR is the most important factor controlling the mechanical properties of the TRIP effect.
The chemical composition of the alloy affects the stability of austenite, owing to the change in the chemical free energy between austenite and martensite. Several relationships between chemical composition and martensite start temperature have been proposed,7,8) and the chemical composition of metastable austenite has been regulated through the control of alloy elements and thermomechanical treatment.9–11) However, in TRIP steel, the morphology and precipitation site also affect the stability of γR.12–16) Matsuda et al.15) used transmission electron microscopy to demonstrate that the globular γR that precipitated at the triple-junction points of bainite grain boundaries was unstable compared to the filmy γR that precipitated at the bainite lath boundary. Tirumalasetty et al.16) applied the ex-situ electron backscattered diffraction (EBSD) method during tensile deformation and reported that the γR inside ferrite grains was more stable than that precipitated at ferrite grain boundaries. Using an identical ex-situ EBSD method, Yamashita et al.17) revealed that the γR near the zone normal to ⟨111⟩ along the tensile direction was mechanically stable. A similar result was obtained using an in-situ neutron diffraction method.18) Koga et al.19) also demonstrated that γR within the high-strain region tended to preferentially transform into martensite. Several factors influencing the stability of γR have been proposed. Using a three-dimensional high-energy X-ray diffraction microscope, Melero et al. investigated the effects of carbon concentration and grain size on the stability of γR at the same time.20) However, the contribution of factors mentioned above to the stability of γR has not been comprehensively studied because it is difficult to evaluate these factors simultaneously even with a three-dimensional high-energy X-ray diffraction microscope.
In this study, the factors representing the stability of γR in low-alloy TRIP steel were statistically analyzed by automatically acquiring data on the size, strain level, crystal orientation, morphology, and precipitation sites of a large numbers of γR grains (>1000) using the ex-situ EBSD method.
A 0.31C–1.74Si–1.49Mn–0.006P–0.0010S–0.008Al–0.0014N (mass%) cold-rolled steel plate with a thickness of 2.5 mm was used in this study. The steel was cold-rolled and annealed in a salt bath at 1063 K for 300 s in the ferrite and austenite region, and then austempered in another salt bath at 673 K for 600 s. The initial volume fraction of γR was 17.2% and its carbon content was calculated to be 1.32 mass% using the X-ray diffraction method.17)
The microstructure of the surface of the normal direction (ND) specimen was observed with a secondary electron image using field-emission scanning electron microscopy (FE-SEM). The specimen was electropolished in a perchloric acid and ethanol solution (1:9 by volume) at 253 K using an electrolytic polishing machine (Lectropol 5) with an applied voltage of 31 V. The EBSD technique with FE-SEM was employed to analyze the volume fraction of γR before deformation and at a plastic strain of 0.04. Data were recorded using a beam scanning step of 50 nm. Data points with a confidence index of greater than 0.1 were provided for phase identification using the software program OIMTM analysis 7.0.1. The EBSD method tends to underestimate the volume fraction of γR because the fine γR grains can not be detected by this method;21) however, the difference between the volume fraction of γR measured by the EBSD17) and neutron diffraction methods22) was only 2% for the plastic strain of up to 0.04 in the TRIP steel used in this study. Therefore, the relationship between the stability of γR and each factor was investigated at a plastic strain of 0.04 to ensuring the accuracy of the analysis.
The microstructures of the unloaded specimens after reaching a plastic strain of 0.04 at 293 K were analyzed using the digital image correlation (DIC) method. The analysis was conducted using the software program VIC-2D for SEM images using a subset size of 81 pixels and step size of 3 pixels.
2.2 Measurement of the stability, size, and aspect ratio of retained austeniteThe stability of γR was evaluated using the ratio of the area of γR after deformation to that before deformation (Rarea) measured from the phase map. Figure 1 shows examples of the austenite phase maps (a) before and (b) after deformation (0.04 nominal strain). After deformation, a part of the γR disappeared, indicating that a deformation-induced martensitic transformation occurred. The areas of γR before and after deformation were measured automatically using OpenCV and Rarea was calculated. The high and low Rarea values denote high and low stability of γR, respectively. The area of γR before deformation was used as the size of γR. The aspect ratio, defined as the ratio of the short side to the long side of a γR grain, was measured from an identical phase map before deformation using OpenCV. The analyzed γR grains were larger than 0.2 µm2, and a total of 1204 were used in this analysis. The analyzed full-phase map was presented in our previous study.19)
Examples of the austenite phase map (a) before and (b) after deformation.
The crystal orientation of γR along the tensile direction was evaluated based on EBSD data. Figure 2 shows the orientation map before deformation colored based on the color of the standard stereographic triangle in Fig. 2. The γR grains with a crystal orientation within 15° of ⟨001⟩, ⟨101⟩, ⟨111⟩, and ⟨113⟩ along the tensile deformation were analyzed in this study, whereas the other crystal orientations, that is, the white regions in the inverse pole figure, were excluded.
Orientation map colored based on the color of the standard stereographic triangle before deformation.
The γR existed inside the ferrite or bainite regions in TRIP steel.23) Figure 3 shows part of the orientation map of the ferrite phase in the tensile direction overlaid on the image quality map before deformation. White lines denote ferrite/bainite boundaries. Because the image quality is decreased by the existence of strain, the martensite regions are of a low image quality, that is, dark contrast, as indicated by the white arrows in Fig. 3. The ferrite was recrystallized and had low strain, whereas the bainite had a certain amount of strain owing to its transformation mechanism. Thus, the image quality in the bainite region are generally lower than that in the ferrite region.24) Furthermore, the bainite had a finer grain size than the ferrite, and the γR within the bainite region had an identical crystallographic orientation.23) The precipitation site of γR was classified as ferrite or bainite based on the aforementioned points, as shown by the white lines in Fig. 3.
Part of orientation map of ferrite phase on tensile direction overlaid on image quality map before deformation. White lines denotes ferrite and bainite boundaries.
The strain levels of γR were determined by the strain distribution measured using DIC analysis in accordance with our previous study19) and were classified into four groups: 0–0.02 (low), 0.02–0.04 (low medium), 0.04–0.06 (high medium), and 0.06–0.08 (high). The average strain was 0.04. The strain level of γR denotes the strain not only the γR but including the ferrite or bainite, owing to the low resolution of the strain distribution resulting from the large subset. Furthermore, the strain generated by deformation induced martensitic transformation can be estimated to be below 0.4%, and can therefore be ignored in the analysis; the details were discussed in our previous study.19)
2.4 Measurement of carbon concentration distributionThe carbon concentration in the γR was measured using two wavelength-dispersive X-ray spectroscopy (WDS) detectors equipped with FE-SEM. These two detectors enabled a short measurement time for carbon concentration. The acceleration voltage was 2.5 kV. To avoid contamination of the specimen surface, which can cause an overestimation of the carbon concentration, a point analysis was conducted because the short measurement time is effective for avoiding contamination. The carbon concentration of γR was calculated based on a calibration curve prepared from standard carbon steels.
2.5 Machine learning for the stability of retained austeniteLightGBM (an ensemble learning algorithm) in scikit-learn (a machine-learning library in Python) was adopted for the prediction of Rarea. The relationship between Rarea and factors such as the size, strain level, and precipitation site of γR was learned, and the feature importance was calculated. Frequency and cumulative frequencies were evaluated for each factor, and the relative importance was demonstrated by the baseline for cumulative frequencies. The importance of a feature was computed as the normalized total reduction of the criterion brought by that feature. This is also known as Gini importance. The hyperparameters of LightGBM, such as the rate drop, max drop, and number of leaves, were tuned using Optuna.
Figure 4 shows histograms of the γR size before deformation for each (a) strain level, (b) precipitation site, (c) crystal orientation, and (d) aspect ratio. Here, although there was no strain distribution before deformation, the γR size before deformation for each strain level was measured from the region in each strain level after deformation. The frequency of fine γR of less than 0.5 µm2 in the high-strain level (0.06–0.08) was significantly higher than for the other strain levels (Fig. 4(a)), indicating that the ratio of the number of fine γRs to the number of γRs in the high-strain level was high compared to that for other strain levels. The γR was a hard phase in the low alloy TRIP steel,19,22) and the ratio of the area of γR to the subset size for the fine γR was small, which could lead to the high frequency of fine γR at the high-strain level. The histograms for the other strain levels were approximately equal, indicating that the strain was distributed independent of the γR size for these strain levels. The histogram in the ferrite region (Fig. 4(b)) indicates that the amount of fine γR in the ferrite region was larger than that in the bainite region. The γR in ferrite regions are austenites that have been nucleated from deformation bands inside the ferrite grain during heat-treatment and retained until reaching room temperature, while the γR in bainite regions are untransformed austenite divided by bainites during heat-treatment and retained at room temperature owing to carbon concentration in the austenite. Therefore, it is reasonable that the size of the precipitated γR in the ferrite region would be finer than the size of the divided γR in the bainite region. The histograms for each crystal orientation (Fig. 4(c)) were similar; however, the frequency of fine γR in the ⟨111⟩γ was larger than that in the other orientations. Our previous study23) revealed that the γR inside the ferrite grain obeys the Kurdjumov-Sachs (K-S) orientation relationship, (111)γ//(110)α and [1-10]γ//[1-11]α.25) Because the size of the γR inside the ferrite grains tended to be fine, it is possible that the K-S orientation relationship may affect the amount of fine γR in the ⟨111⟩ through inheriting the texture of the ferrite grain. The coarse γRs tended to have low aspect ratios, as shown in Fig. 4(d), meaning that these γRs had acicular shapes. This coincides with our previous study,23) which found that the coarse acicular γRs were often observed in both ferrite and bainite regions.
Histograms of the γR size before deformation for each (a) strain level, (b) precipitation site, (c) crystal orientation, and (d) aspect ratio.
Figure 5 shows the frequencies of the strain levels corresponding to each (a) crystal orientation and (b) precipitation site. Here, the frequency refers to the ratio of the number of γRs for each strain level and each orientation or precipitation site to the total number of γRs (1204). The frequency for ⟨001⟩ was low compared to the other crystal orientations, indicating that the number of γR was low. Although the absolute values of the frequencies differed among the crystal orientations and precipitation sites, each orientation and precipitation site exhibited a similar frequency of strain levels, that is, the strain had no correlation with the orientation or precipitation sites.
Frequencies of strain levels belonging to each (a) crystal orientation and (b) precipitation site.
Figure 6(a) shows the cumulative frequency of Rarea for all the data, that is, 1204 γRs. The cumulative frequency increased linearly with increasing Rarea, indicating that the stability of γR was completely randomly distributed in the TRIP steel. Figures 6(b)–(f) show the cumulative frequencies of Rarea for each (b) size, (c) strain level, (d) crystal orientation, (e) aspect ratio, and (f) precipitation site. The dotted line denotes the baseline, which is the change in the cumulative frequency for all the data. The fine γR (0.3–0.5 µm2) clearly exhibited a higher cumulative frequency than the baseline, which continuously decreased with increasing size of the γR (Fig. 6(b)). Here, there is a possibility that Rarea becomes smaller in fine γR compared to that in coarse γR when the same amount of martensite forms. Therefore, Rarea in fine (<1 µm2) and coarse (>1.9 µm2) γRs that had the same total area before deformation were calculated to confirm the effect of size on the stability of γR. Table 1 summarizes the number of γRs, sum of the area before and after deformation, and Rarea calculated from the sum of the area in fine and coarse γRs. The number of fine γRs was approximately five times larger than that of the coarse γR. After deformation, the sum of the area for the fine γR was significantly smaller than that of the coarse γR, thus, Rarea was smaller for the fine γR. Therefore, it can be concluded that fine γRs are unstable, and the size of γR is significantly correlated with the stability of the γR.
Cumulative frequencies of Rarea for (a) all data, (b) size, (c) strain level, (d) crystal orientation, (e) aspect ratio, and (f) precipitation site.
The γR within the high-strain region (0.06–0.08) exhibited a tendency similar to that of the fine γR (Fig. 6(c)); the cumulative frequency was higher than that of the baseline. However, the cumulative frequencies for other strain levels (0–0.02, 0.02–0.04, and 0.04–0.06) were approximately identical to the baseline. Thus, although the γR within the high-strain region was unstable, the influence of the strain level on the stability of γR was smaller than that of the size.
The crystal orientation was hardly correlated with the stability of γR (Fig. 6(d)); all the cumulative frequencies exhibited a linear increase as functions of Rarea, which is identical to the trend for all data. The cumulative frequency for ⟨001⟩ within the range of Rarea from 0 to 0.5 was slightly higher than that for the other crystal orientations; however, this is likely to be an error because the number of γRs in the ⟨001⟩ were smaller than those of the other crystal orientations, as shown in Fig. 5(a). Application of the in-situ neutron diffraction method on TRIP steel revealed that γRs whose ⟨111⟩ was parallel to the tensile direction, increased as the tensile test proceeded, and it was concluded that these γRs were stable, although the amount of such γR also increased with the crystal rotation.26) However, in this study, the effect of crystal orientation on the stability of γR was low. Considering that the volume fraction of γR near the zone normal to ⟨111⟩ along the tensile direction increased from 0.1 plastic strain in low alloy TRIP steel,17) it can be reasonably concluded that the crystal orientation hardly affects the stability of γR at low tensile strain levels.
The cumulative frequency of the high aspect ratio (>0.7), that is, circular γR grain, was slightly higher than that of the other aspect ratios within the range of Rarea from 0.3 to 0.9 (Fig. 6(e)), which agrees with the previous finding that globular γR was unstable compared to filmy γR.14) However, the difference in the cumulative frequency among the aspect ratios was smaller than that of the size and strain level; thus, the effect of the aspect ratio on the stability of γR can be considered negligible. The filmy γRs at the lath boundary in a martensite or bainite structure are stable because of the high carbon concentration.14) Because these filmy γRs could not be detected in the TRIP steel using the EBSD method owing to their fine size, the volume fraction of filmy γR could be estimated to approximately 2% from the difference between the volume fraction of γR measured by the EBSD17) and neutron diffraction methods.22) Therefore, most of the γR was formed not at the lath boundary but along the deformation band in a ferrite grain, and was untransformed austenite in the bainite region. This may be the reason why the effect of aspect ratio on the stability of γR was small in the TRIP steel.
The cumulative frequency in the bainite region was lower than that of the baseline (Fig. 6(f)). Although the frequency of fine γR was low within the bainite region (Fig. 4(b)), the value of the cumulative frequency was significantly low and it can be inferred that γR within the bainite region tended to be stable. The reason for this will be discussed later.
Figure 7 shows the cumulative frequencies of Rarea by strain levels and precipitation sites for both the fine γR smaller than 0.5 µm2 and the coarse γR larger than 1 µm2. The cumulative frequencies of the fine γRs were higher than baseline frequencies (Figs. 7(a), (c)), whereas those of the coarse γRs were lower than the baseline frequencies (Figs. 7(b), (d)). This indicates that the size of the γRs significantly affects its stability, as shown in Fig. 6(b). In the fine γR, the cumulative frequencies in the high-strain region and ferrite region were higher than those for the other strain levels and bainite region, respectively (Figs. 7(a), (c)). In contrast, in coarse γR, all cumulative frequencies were approximately identical (Figs. 7(b), (d)). Therefore, it can be concluded that the strain levels and precipitation sites of γR affect the stability of only fine γRs, and that, the size of γR was the most dominant factor controlling its stability. Figure 8 shows cumulative frequencies of Rarea by strain levels for the γR smaller than 0.5 µm2 in (a) ferrite and (b) bainite regions. The cumulative frequencies in the ferrite region were higher than those of the baseline, regardless of the strain level (Fig. 8(a)), indicating that the γR in the ferrite region was unstable. Notably, the cumulative frequency for the low-strain level in the bainite region was remarkably lower than that of the baseline and approximately equal to that of the coarse γR (see Fig. 8(b) and Figs. 7(b), (d)). Thus, the fine γRs were only stable at a low-strain level in the bainite region.
Cumulative frequencies of Rarea by (a), (b) strain levels and (c), (d) precipitation sites for the γR lower than 0.5 µm2 (a), (c) and greater than 1 µm2 (b), (d).
Cumulative frequencies of Rarea by strain levels for the γR lower than 0.5 µm2 in (a) ferrite and (b) bainite.
The relationship between Rarea and the size, strain level, and precipitation site of γR was learned using the LightGBM. However, the coefficient of determination between the predicted and actual Rarea values was approximately 0.15, indicating that Rarea was difficult to predict based on these parameters. There are two possible reasons for the low coefficient of determination: all factors affect Rarea in a complex manner or there is an unknown factor affecting Rarea. Although the accurate prediction of Rarea was difficult in this study, the feature importance indicated that size was the most influential factor on Rarea, that is, on the stability of γR, as shown in Fig. 9, which agrees with the results of Figs. 6 and 7.
Feature importance calculated from LightGBM learned relationship between Rarea and size, strain level, and precipitation site of γR.
The carbon concentration was not considered in Section 3.2 owing to the difficulty of measuring the carbon concentration due to the contamination of hydrocarbons that occurred during the EBSD measurement. It has been reported that carbon concentration affects the stability of γR20) because carbon is a strong austenite-stabilizing element. The carbon concentration of 30 γRs was measured using a WDS. Figure 10 shows a secondary electron (SE) image indicating the WDS analysis points numbered from 1 to 30, and (b) carbon concentration at the points shown in (a). The WDS measurement was carried out before EBSD to avoid contamination by hydrocarbons; thus, we identified γR from the contrast in the SE image. The average carbon concentration was 1.60 mass%, which was slightly high but approximately equal to the carbon concentration (1.32 mass%) measured using the X-ray diffraction method.17) Furthermore, the measured carbon concentration within the ferrite region was less than 0.1 mass%; thus, the WDS measurement had a relatively high accuracy.
(a) Secondary electron image indicating the WDS analysis points numbered from 1 to 30, and (b) carbon concentration at the points shown in (a).
Figure 11 shows an orientation map of γR on tensile direction in the same region as Fig. 10 on the specimen, (a) before deformation, and subjected to plastic strains of (b) 0.04, (c) 0.08, and (d) 0.12, respectively. The numbers in the figures indicate the analysis points in Fig. 10(b). The portion of the γR that was already transformed to martensite before deformation was thermally induced martensite. Beyond a plastic strain of 0.04 (Fig. 11(b)), several γRs disappeared and transformed to deformation-induced martensite. The number of γRs continuously decreased with increasing plastic strain (Fig. 11(c)), and finally, only four measured γRs remained. Figure 12 shows the frequency of the carbon concentrations in the (a) thermally induced martensites, (b) deformation-induced martensites transformed until a plastic strain of 0.12, and (c) γR remaining at a plastic strain of 0.12. The carbon concentrations of the thermally induced martensites were not low, but they were randomly distributed. Using specific electron probe microanalysis suppressing hydrocarbon contamination for high accuracy determination of carbon concentration,27) a similar result had been demonstrated by Tanaka et al.28) It has been considered that the thermal stability of austenite is dominated by its chemical composition; interestingly, these results indicate the existence of other factors influencing the thermal stability of austenite, and additional investigation is needed to elucidate this point. The mechanical stability of γR was also hardly correlated with carbon concentration (Figs. 12(b), (c)). There were various carbon concentrations in the γR transformed to martensite below a plastic strain of 0.12; two of the remaining γRs had high carbon concentrations (>2.0%), but the other two were low (approximately 1.3%). Although there was no clear correlation between the stability of γR and influencing factors such as the size, strain level, and precipitation site owing to the small number of measured γRs in this analysis, it can be concluded that the mechanical stability of γR was not determined by only the carbon concentration. Therefore, the low coefficient of determination of the machine learning was not due to a lack of carbon concentration.
Frequency of the carbon concentrations in the (a) thermally induced martensites, (b) deformation-induced martensites transformed until a plastic strain of 0.12, and (c) γR remaining at a plastic strain of 0.12.
The main representative factors of the stability of γR in the TRIP steel were the size, strain level, and precipitation site. Of course, carbon concentration should also be one of the representative factors of the stability of γR; however, it was not considered because it could not be investigated quantitatively in this analysis. Martensitic transformations can be induced by either stress or strain. Stress provides the driving force for martensitic transformation, whereas strain decreases the driving force required for martensitic transformation.29,30) Kawata et al.31) concluded that γR in low alloy TRIP steel transforms to martensite induced by stress, that is, stress-induced martensitic transformation is dominant. Furthermore, in our previous study,19) we demonstrated that the strain introduced to γR during tensile deformation was lower than that in the ferrite or bainite regions. Thus, we also concluded that the stress concentration is due to the difference in strain between γR and ferrite or bainite induced the martensitic transformation.
Assuming that martensite is mainly induced by stress, the relationship between the stability of γR, stress concentration, and each representative factor is schematically illustrated in Fig. 13. Here, the shape of martensite is not correctly described in Fig. 13 for simplifying the illustration. The stability of fine γRs smaller than 0.5 µm2, which is half of the average size of γR (Savg.), in high-strain levels (over 1.5 times of average strain (εavg.)) in the ferrite region was the lowest in TRIP steel (Fig. 8(a) and Fig. 13(a-2)). The strain measured in this study included the strains in both ferrite or bainite and γR owing to the low resolution of the strain distribution, as discussed in a previous study.19) Furthermore, γR is harder than ferrite or bainite in this TRIP steel; thus, in the high-strain region, the difference in strain between ferrite or bainite and γR was large. The lowest stability of fine γRs in the high-strain level in the ferrite region can be attributed to the large stress concentration region owing to the high ratio of γR/ferrite interfacial boundaries to the volume of fine γRs and the large difference in strain between ferrite and γR (Fig. 13(a-1)). The stability of fine γR in the ferrite region improved with lowering strain level (Fig. 8(a) and Fig. 13(b-2)), which can be reasonably understood as the stress concentration region narrowing because of the decrease in the difference in strain between ferrite and γR (Fig. 13(b-1)). The fine γRs at high-strain level in the bainite region were more stable than those at low-strain level in the ferrite region (Fig. 8 and Fig. 13(c-2)). There are three possible reasons of the high stability of γR in the bainite region: (1) bainite is harder than ferrite, providing a narrow stress concentration region in γR (Fig. 13(c-1)) and a high stability of γR, although the internal stress in γR may be high in the bainite region owing to its relatively high strain; (2) the formation mechanism of γR is different between the ferrite and bainite regions, as mentioned in Section 3.2, and the carbon concentration of γR may be high in the bainite region, leading to the high stability of γR; (3) bainitic transformation accompanies volume expansion owing to phase transformation,32) and the strain is introduced into the neighboring untransformed austenite,33) thus, the γRs in bainite may have hydrostatic stress which provides a high stability for γR. The stability of coarse γRs larger than 1 µm2, which is equal to Savg, was significantly higher than that of fine γRs regardless of the strain levels and precipitation sites. The low ratio of γR/ferrite or bainite interfacial boundaries to the volume of γR should provide high stability to the coarse γRs (Figs. 13(d-1), (e-1)). Although the strain level slightly affected the stability of coarse γR (Figs. 13(d-2), (e-2)), it can be concluded that the size is the most representative factor for the stability of γR in TRIP steel subjected to a plastic strain of 0.04. Exceptionally, the fine γRs at the low-strain level in the bainite region had high stability equal to that of the coarse γRs (Fig. 8(b) and Fig. 13(e-2)), indicating that the combinational effect of the strain level and precipitation site on the stability of γR is equal to the effect of size. This complicates the effect of each factor on the stability of γR, and makes machine learning difficult.
Schematic illustration of the relationship among the stability of γR, stress concentration and each representative factor.
The importance of stress for deformation induced martensitic transformation has been highlighted by relationship between the macroscopic applied stress and the volume fraction of γR.31) Our results suggest that stress concentration occurs at the microscopic scale owing to microstructure, that is, γR, ferrite and bainite, and it is important for the deformation induced martensitic transformation in each γR. The size of γRs should be closely related to the stress concentration, which represents the stability of the γR. If the stress distribution can be visualized by correlating the microstructure, it will be strongly related to the size of γR. Stress distribution analysis methods such as synchrotron dark-field X-ray microscopy34) may directly reveal the effect of stress distribution on the stability of γR in TRIP steel; thus, further investigation is needed. Rarea could not be accurately predicted from machine learning, which may have been caused by the lack of a parameter for the stress condition. Although various parameters may affect the stress conditions of γR and provide various influencing factors on the stability of γR, our results strongly suggest that the size of γR can effectively control the stress concentration and stability of γR compared to other factors in the TRIP steel at plastic strain values of up to 0.04.
The factors representing the stability of retained austenite (γR) in low-alloy TRIP steel were statistically analyzed by automatically acquiring data on the size, strain level, crystal orientation, morphology, and precipitation sites of a large number of γR grains (>1000) using the ex-situ EBSD method. The main results are summarized as follows:
The authors are grateful to Nihon Parkerizing Co., Ltd., for performing wavelength-dispersive X-ray spectroscopy. The authors acknowledge the financial support of the Grant-in-Aid for Scientific Research (KAKENHI) Grant No. 20K1460 and the Iketani Science and Technology Foundation (0331027-A).
