Development and Comparison of Quantitative Phase Analysis for Duplex Stainless Steel Weld

Phase Quantification in Duplex Steels 2018 62 3 Abstract In duplex stainless steels the ideally 1:1 ratio of austenite-to-ferrite phases ensures the outstanding mechanical and corrosion properties compared to other, conventional stainless steel grades. However, this phase balance can be easily shifted to a mostly austenitic or mostly ferritic microstructures, depending on the welding process and heat input. In order to determine the phase ratio, several methods are available to use, such as Feritscope measurements, ASTM E562 manual point count method (on metallographic images) or quantitative image analysis. From these methods, Feritscope measurements cannot be applied to determine phase quantification in the narrow heat affected zone of duplex stainless steel welds – because of the very limited heat input. The manual point count method is very dependent of the assessor and cannot be automated. In this paper a histogram-based image analyzing process was developed, using Beraha’s etchant solution. The results were compared to Feritscope measurements and a very good correlation (R2 = 0.9995) was found. This method will give the ability to easily and automatically measure phase ratio in weld metal, heat affected zone or in subsurface regions of multi-pass welds.

(EBSD) [21,22] or most often quantitative optical microscopic method can be used for phase ratio evaluation [23,24]. Quantitative optical microscopy can be used to determine the volume fraction by a manual point count method (according to ASTM E562 standard [25]) or quantitative metallography using an image analyzer. In case of the ASTM E562 method, an array of points formed by a grid line is superimposed upon a magnified image, and the number of points falling within the microstructural constituent of interest is counted and averaged for a selected number of fields. If the amount of volume fraction of interest is higher than 20 % (which is true for almost all cases of DSS welds), 100 point should be evaluated of 20 fields for a 10 % relative accuracy. This method is slow, highly subjective, not automated, and not repeatable for all users. On the contrary, image analyzing is more efficient, because it can be automated and gives faster results than the manual point count method.
For the proper usability of image analyzing an adequate metallographic preparation and contrast etching is required.
For DSS etching different etchant solutions are available (see Table 1); however, Beraha's reagent is widely used for phase selective color etching [26][27][28]. Beraha's reagent paints the ferrite phase dark and the austenite phase remains light. In case of Beraha's etchant it was found the etching time has a great influence on the usability of metallographic images for quantitative phase analysis.
In our research we investigated the effect of Beraha's reagent's etching time of 2205 duplex stainless steel (EN 1.4462) welds on the image quality. The optimum etching process was determined, which also gives the best correlation to Feritscope results in the weld metal and base metal.

Materials and methods
The used base material for automated, autogenous tungsten inert gas (TIG) welding was a 6 mm thick 2205 (EN 1.4426 or X2CrNiMoN22-5-3) duplex stainless steel sheet. The chemical composition of: 22 wt.% Cr, 5.5 wt.% Ni, 2.8 wt.% Mo and 0.166 wt.% N. The heat input during welding was 1 kJ·mm -1 , calculated with 0.6 thermal efficiency factor for TIG welding.
The Feritscope measurements were done in 10 points on the face side along the 100 mm long weld seam with equal distances. The used instrument was a Fischer FMP 30 Feritscope, which measures according to the magnetic induction method.
For quantitative image analyzing measurements standard metallographic specimens were prepared from the transverse direction of the weld and polished to 1 µm diamond polisher. For the microstructural evaluation of duplex stainless steels different etchants are recommended (Table 1). As mentioned, for phase ratio quantification Beraha's etchant is the most suitable [29], because it colors ferrite phase dark and leaves the austenite phase bright. However, the sufficiency is strongly dependent on the etching time and the number of etching cycles. In order to investigate the effects of etching time and cycle of Beraha's etchant, different total etching times and different cycles were used (see: Table 2) to quantify the phase ratio in the weld. In Table 2 the total etching time is to be interpreted as the sum of the etching times of each etching cycles, (e.g. 3 cycles and 12 seconds total etching time means 3×4 s etching time). The etchant should be freshly made before every process [29]. Before every etching cycle the metallographic samples were grinded back to 2000 grit paper and polished again. Before the etching the samples were cleaned with ethanol and completely dried. During etching the sample was constantly stirred in the etchant solution. After the etching process the sample was washed off in running water and cleaned with ethanol and dried again. If multiple etching cycles were used, the cleaning process was done before every etching cycle.
According to Table 2, an image was taken using Olympus PMG3 optical microscope from the same 700×700 µm area after each etching process. The phase quantification of the metallographic images were done using an image threshold method [32].
The steps of the developed threshold process (also illustrated in Fig. 2) are: (I) the original image is loaded to an image analyzer software and (II) the grayscale histogram is taken up on the 0 to 255 range (8 bit). The histogram has two peaks; one at a darker gray level (DGP, closer to 0 value) and one at a lighter gray level (LGP, closer to 255 value). The difference between the two gray level peaks (ΔG) is equivalent to the level of contrast of the image. In order to count the darker and lighter number of pixels (which is in correlation to the austenite -ferrite ratio) the original image should be converted into a black and white image. The most adequate way to do this is to take the average value of the gray level difference (0.5 · ΔG) and add it to the DGP value (III). The resulted value can be taken as a boundary (DGP + 0.5 · ΔG) (IV), and from this boundary the lower values will be painted to black and the higher values to white (V). Finally, the ratio of the white and black pixels can be measured and correlated to the austeniteferrite ratio (VI). This process can be automated and does not depend on the type of image analyzer software used. It is also found, in order to have a corresponding result to the Feritscope measurements, the microstructure image should be taken from at least a 500×500 µm area.
Although the Feritscope measurement method is the easiest way to determine the ferrite content in a DSS, it's application has boundaries. Since it is a magnetic inductive method, the result of the measurement strongly depends on the right contact between the probe and the measured material. This is the reason only flat surface can be measured correctly. For example, the measurement of weld root can be often misleading because of the strong curvature and the lack of magnetic contact. The other limit of the Feritscope method is the measurement of smaller volumes. The HAZ in DSS welds is usually narrow at the typical heat input range (e.g. Fig. 1). Thus, image analyzing and ASTM E562 manual point count are the only proper methods for determining the HAZ microstructural phase ratio.

Results and discussion 3.1 Feritscope measurements
Fischer type Feritscope measurements were done in the weld metal in order to compare the results of metallographic phase ratio quantification of different etching procedures. The average measured austenite content, along the whole length of the welded seam was 28.9±2.7 %. All the results of histogram based image analyzing will be compared to these results.

Phase ratio quantification with image analyzing
In Fig. 3 the ΔG values are plotted as a function of total etching time vs. etching cycles. In Fig. 3 it is visible the level of ΔG depends on the used etching cycle and etching time. The optimal etching process can be found, where ΔG (the level of contrast) is the maximum. The highest ΔG values were found at the multiple etching cycles. Accordingly, in Fig. 3 two optimums can be predicted. For visualization, cropped areas taken from the etched specimens (according to Table 2) are shown in Fig. 4. In Fig. 4 it is also visible that multiple etching cycles will give better results in contrast. For a single cycle of etching the 6 s etching time gave the best result. The 24 s one cycle etching gave the least contrast value, which means the grain structure is over-etched and no optical difference is visible between the two phases (it would make the ASTM method virtually impossible). The first etching cycle is suitable to reveal grain boundaries, but in order to increase the contrast (darken the ferrite phase) another cycle of etching is needed. Three etching cycle is generally not recommended, because in case of shorter individual cycles insufficient contrast was observed (e.g. 3×2 s and 3×4 s etching cycles). One optimal etching process was found at 3×16 s (ΔG = 149) and another at 2×12 s (ΔG = 152) etching cycle, respectively. According to Fig. 3, the optimal etching process seems to have a gradient toward the multiple cycles, with ~10-15 s etching time in each cycle, but the possible gain is not sufficient as more effort and attention is needed for sample preparation. Therefore, for further evaluation the shorter and simpler 2×12 s total etching cycle was chosen in order to minimize the possible etching defects coming from multiple etching cycles (ethanol cleaning, drying).
For all of the etching cycles the austenite ratio (AR image ) is measured (Table 3) with the image analyzing method described in Chapter 2 (Materials and methods). Taking the Feritscope's result as etalon (AR Feritscope = 28.9±2.7 %) the difference of the image analyzing results compared to the Feritscope's result are also calculated in percentages as, RD = (AR Feritscope -AR image ) × AR Feritscope -1 .
From RD a unit less number, the degree of usability (DoU) is calculated as DoU = 100 × RD -1 . Representing DoU as a function of the DGP and LGP values of the different etching cycles on a contour color fill plot (Fig. 5), it is visible that, the lower DGP and the higher LGP value will result in a higher DoU.
The evaluation of the image with the chosen optimal etching process (2×12 s) gives 30.3 area % austenite content (AR image ) and 20.6 DoU value. In accordance with Fig. 5, the DoU value can be plotted as a function of the ∆G contrast level (Fig. 6). The relationship is a quasi-exponential correlation; the increasing contrast level means exponentially increasing usability.
According to Fig. 6 the two highest usability values were found at the optimal etching times, where the shorter and simpler etching process (2×12 s etching cycle) gave the highest value. Therefore, this cycle is recommended for histogrambased image analyzing.

Results of manual point count method
The manual point count method according to ASTM E562 was done on two images, which represents the two extremes of the result of the etching process. The two selected images were the ΔG = maximum (2×12 s etching cycle) and the ΔG = minimum (1×24 s etching cycle). The original size of the images was a 700×700 µm area. For manual point count a grid was placed onto the images with 108 intersection points. Each intersection points were evaluated individually, deciding if the point is on austenite phase (= 1) on ferrite phase (= 0) or on phase boundary (= 0.5). After the evaluation, the sum of all the values are divided by the number of points (108 points) and multiplied by 100 to get the austenite phase percentage. On the image where ΔG = minimum, it was not possible to get results, because we were unable to decide whether the intersection points lied on the austenite or ferrite phase since there was a lack of color contrast. On the image where ΔG = maximum, the error of the subjective manual point count appeared; one result was 36.1 area % and the other was 33.3 area % on the same image with the same grid (Feritscope value was 28.9±2.7 %, while image analyzer read 30.3 area %). Also, if a 10 % relative accuracy is to be achieved 20 fields of a 108 points grids should be evaluated. Although the ASTM E562 manual point count method is widely used on metallographic images, its measurement error, slowness, and subjectivity can lead to misleading results. Because of the significant error and lack of reproducibility, the manual point count method is usually complemented with image analyzing as well [1,33,34]. On the other hand, the image analyzing process described in this paper will always give the same results, regardless of whom the assessor is.

Comparison of Feritscope and histogram-based image analyzing method
In order to validate the developed histogram-based image analyzing process (described in Materials and Methods section), previously welded samples were evaluated. The previous projects included different duplex and lean duplex stainless steel grades welded with different heat inputs, therefore  resulting in different austenite contents. The austenite content in the weld metal was determined with both (histogram-based image analyzer and Feritscope) methods. These results were compared in Fig. 7. To the measured points a 45° straight line is fitted with R 2 = 0.9995. The value of the optimally etched sample of the current work is highlighted with red color. The result of this comparison means the developed image analyzing process is suitable to use instead of Feritscope measurement for applications where Feritscope is hard to use (e.g. on heat affected zones, weld roots or on multi-pass welds to measure phase ratio of each pass).

Conclusions
• In this research, a histogram-based image analyzing method for phase quantification of duplex stainless steel welds was developed. The main conclusions are: • Beraha's type etchant is suitable for 2205 (EN 1.4462) duplex stainless steel etching for histogram-based image analyzing. • The optimal etching cycle for histogram-based image analyzing was determined as two etching cycles each for 12 seconds. • The histogram-based image analysis for phase analysis was compared and validated with Feritscope results in the weld metal as a very good correlation (R 2 = 0.9995) was found.
The histogram-based image analyzing method described in this paper is also applicable to measure phase ratio in the heat affected zones of the weld, where Feritscope measurement is not applicable.