January AMP_Digital

A D V A N C E D M A T E R I A L S & P R O C E S S E S | J A N U A R Y 2 0 1 9 1 7 Fig. 1 — (a) Original BSE image; (b) outline of automatically identified grains (red) on original image. (a) (b) EBSD is often used purely for grain size analysis [3] , primarily when prior at- tempts to identify grains using SE and BSE imaging have failed. While EBSD can accurately report grain size, short- comings of the technique include: • Time intensive: EBSD collection time of a single field-of-view commonly ranges from two to eight hours. High-speed cameras reduce collec- tion time in some applications, but cameras are expensive and even more sensitive to sample prepara- tion and material composition. • Limited resolution: The nature of the electron beam/sample interaction in traditional SEM imaging versus EBSD typically means that SEM im- aging can resolve smaller features than EBSD. • Tedious sample preparation: To produce useful data using EBSD, samples must be relatively free of internal strain and their surfaces meticulously polished to a mirror finish with extreme care taken at each step. CASE STUDY: REPLACING EBSD FOR GRAIN SIZE ANALYSIS A manufacturer wanted to replace EBSD grain size measurement with au- tomated BSE SEM image analysis. Fig- ure 1(a) shows an example image of the microstructure of interest. Even with optimal sample preparation, sev- eral software solutions failed to detect the microstructure’s BSE-imaged grain boundaries with acceptable accuracy. BSE imaging provides an incomplete representation of the microstructure’s grain boundaries (Fig. 1a). Experienced metallographers can mentally connect the dots to delineate each grain, but the task can be very difficult to automate. In this case, the need to connect the dots, together with the faint contrast exhibited by visible boundaries were the primary challeng- es that previous automated solutions failed to overcome. Recognizing that manual grain analysis from SEM images is no longer feasible nor acceptable, the company was forced to continue with EBSD for automated grain sizing. The company worked with Mipar Software to pursue automated BSE grain size analysis. A promising auto- mated solution was quickly developed including an adaptive feature detection capability that captures subtle bound- ary contrast, and a “separate features” function, which mimics human in- terpretation to complete the partial- ly revealed grain structure. Figure 1(b) shows an outline of the automatically identified grains on the original image. VALIDATION To complement the grain detec- tion in Fig. 1(b), Mipar wanted to quan- tify the accuracy with which grain size could be measured from BSE imaging. Figure 2 compares the raw images, grain detections, and grain size distributions extracted from BSE and EBSD images. Table 1 compares grain size statistics fromeachmethod, where statistics and distributions were produced by collect- ing measurements from four random fields of view. Edge grains were exclud- ed frommeasurement in each method. M icrograph analysis and character- ization is a key function of many materials laboratories support- ing manufacturing, quality control, and R&D. While classic methods are subjec- tive and resource intensive, advance- ments in capture technology along with novel approaches to computer algorithm development enable automated tech- niques not possible previously. This arti- cle discusses the benefits of automated micrograph characterization in scanning electron microscopy (SEM), the necessity of image-based analysis in particle char- acterization, and challenges inmoderniz- ing industry standards. ALTERNATIVES TO EBSD FOR GRAIN SIZE MEASUREMENT Grain size is a critical microstruc- tural parameter that directly influenc- es the mechanical properties of nearly all structural materials. Accurate quan- tification of a material’s average grain size and distribution is therefore of paramount importance, as inaccurate measurements can lead to poor quali- ty control, inaccurate property predic- tions, and inefficient R&D cycles. Etching procedures do not suffi- ciently reveal grain boundaries for opti- cal microscopy in some microstructures and grains are too small to be optically imaged in others. In both cases, SEM is often enlisted to image grains. However, the two common SEM imaging modes, i.e., secondary electron (SE) and back- scattered electron (BSE) imaging, can produce less grain contrast than optical microscopy, even with proper sample preparation. In these cases, electron backscatter diffraction (EBSD) [1] has historically been the go-to technique for identifying discrete grains for auto- mated analysis. With EBSD, the crystal structure and orientation of each pix- el are determined, and these data are subsequently processed to reveal the material’s grain structure. EBSD LIMITATIONS In some applications, grain orien- tation data are used to report crystal- lographic texture in addition to grain size [2] . For this purpose, EBSD is typical- ly an irreplaceable technique. However,