Digital advances spark greater productivity in metallurgical laboratory environments.

The Olympus BXiS digital metallurgical microscope may be used to analyze grain size, rate non-metallic inclusions or evaluate cast iron, in compliance with prevailing national and international standards. Source: Olympus America

A reticle typical can be placed in a microscope’s optical path to help operators estimate grain size. Source: Olympus America

Metallurgical quality control has evolved from a manual, subjective and laborious process to a methodology that is accurate, repeatable and automated-thanks to advancements in digital microscope systems that include the latest image analysis software. With systems that integrate the latest digital technology, tasks that used to be time consuming and prone to error-such as estimating the average grain size in aluminum, rating non-metallic inclusions in steel or evaluating nodularity in cast iron-are now easier, faster and performed to higher standards.

Until recently, many quality control professionals regularly examined microstructures one by one under a microscope. The operator made estimates about each sample’s quality based on the image viewed through the microscope’s eyepieces in comparison with a reference reticle that had been placed in the microscope’s optical path and appeared superimposed over the image (Figure 1). At other facilities, microscope operators looked back and forth between the eyepieces and a poster on the wall, comparing what they saw in the microscope to an image or micrograph posted for reference.

Using these methods, quality control staff often became adept at judging, but they were unable to provide assured accuracy, as no one knew for certain that his or her visual perception or estimation was correct. These processes were cumbersome and often inaccurate. Further, there was no guarantee that results from one operator repetitively analyzing successive samples were repeatable. Results might change as an operator became increasingly fatigued over the course of the workday. In addition, results from each operator were subjective enough so that reproducibility was affected, as different operators came up with varying results using the same measuring criteria. But as tough as these issues were, they were only the beginning. After the inspection itself, additional errors could creep in during the documentation process, where operators entered their findings manually into either a paper or electronic workbook.

At the dawn of digital microscopy about two decades ago, software developers proposed the idea of performing these common metallurgical quality control research tasks automatically via image analysis. Early applications focused on samples that had abundant inherent contrast, involving dark features on bright, highly reflective backgrounds, with a minimal number of overlapping structures. With early image analysis programs, users could digitize an analog image projected from the microscope, and, via a series of complex algorithms, rate the microstructure based on gray-scale values and morphological parameters.

But early image analysis technology for metrology faced significant challenges. Analog closed-circuit television (CCTV) cameras were commonplace during this time, and they provided only about 640 x 480 pixels (0.3 Megapixel) of resolution. Because resolution was limited, accurate digitization of the original image was sometimes impossible. Early adopters of this methodology also struggled with throughput. The systems of the 1990s had relatively little processing power, making image analysis a very time-consuming process. Once the image had been digitized, it could take up to several minutes to analyze. Data storage issues presented still another barrier to widespread early implementation. With cost at a premium, most users could not afford the required storage media. In the early years of Internet technology, sharing large files was difficult. So although a great idea in theory, automated image analysis was, in the 1990s, ahead of its time.

Fast-forward to today’s digital metallurgical laboratory. Pre-configured, integrated solutions consisting of a light microscope, digital camera, computer and material-science-specific application software have proven to be indispensible tools for driving fast, accurate and repeatable metallurgical sample data (Figure 2). Today, these robust systems are widely used for such vital metallurgical tasks as grain sizing, non-metallic inclusion rating, cast iron analysis, porosity and phase measurement, dendrite arm spacing, decarburization measurement, layer thickness measurement and particle analysis.

Modern computers accomplish data crunching tasks in seconds, and with data storage no longer at a premium (thanks to highly developed compression algorithms and more affordable memory), storage capacity for many thousands of images can be accommodated in most laboratory environments. Newer microscope-specific digital cameras meet Nyquist criteria, so they provide enough pixels (digital resolution) to properly digitize fine detail.

New digital imaging and analysis solutions comply with prevailing American Society for Testing and Materials (ASTM), International Organization for Standardization (ISO) and other international standards. In recent years, organizations such as ASTM have re-written their standards to accommodate the rapidly growing transition towards automated image analysis.

Enhanced Accuracy, Repeatability, Reproducibility

Among the most significant advantages quality engineers have reported after implementing an automated digital image-analysis solution are increased accuracy, repeatability and reproducibility. Accuracy is improved because the microstructure is evaluated by its gray-scale values to determine if, for instance, a non-metallic inclusion is a sulfide or silicate. Pre-programmed thresholds objectively determine the nature of the inclusion. Similarly, a series of complex algorithms is used to evaluate the specimen’s morphological parameters; shape may be evaluated to determine whether a non-metallic inclusion is a globular oxide or alumina.

Digital image analysis also enhances repeatability. Because algorithms are written to comply with specific standards, there is no human source of manipulation. Data is always computed consistently and accurately. Reproducibility is also improved; the results attained by one operator will be identical to that of another, as long as all experiment parameters remain consistent. Importantly, results also are fully traceable, allowing for smooth and stress-free audits. Users can display reference images, show a calibration report, and even perform a second analysis on an archived image, should the need arise (Figure 3).

In the quest for accurate inspection and analysis data, many metallurgical quality control laboratories and their customers now demand digital data organizing, archiving and reporting. Unlike earlier protocols where operators manually entered findings into an electronic spreadsheet or paper workbook, today’s image analysis software yields automated spreadsheets with full results and related statistical data. Throughput is much faster and there is far less chance that an operator might introduce errors by reversing numbers or making other data entry slips. Detailed reports encompassing product images, measurement results and statistical data in a tabular or graphic (i.e. histogram) format can be generated easily and automatically in Microsoft® Word using a pre-defined template, complete with the corporate address and logo.

Cast iron may be analyzed digitally to determine percentage, size, form and nodularity of graphite in compliance with ASTM A247. Source: Olympus America

Archiving and Reporting Data

Most people find it frustrating to rummage through computer folders searching for an important document. Image analysis packages offer integrated database solutions that can handle many thousands of images, so laboratories can quickly and easily store, archive and search for images, projects and related data. An administrator defines database fields and data tags, and the operator populates these fields when the image or data is first archived. Any qualified user can later perform a query and perform a quick and accurate search (Figure 4). Typical database fields may include “Customer Name,” “Operator Name/Number,” “Project Due Date,” “Grain Size” and “Measurement Results.” The laboratory can organize massive amounts of data in a structured environment, and many data fields populate automatically when creating a report. Thus, software tools provide a totally seamless workflow from image acquisition through analysis, reporting and archiving. A customer request, even for original images and measurement results dating back 18 months, can be handled routinely and quickly.

Although the reflected-light optical microscope has been used widely for decades to analyze metallurgical microstructures, digital microscopes and image analysis software have rapidly eclipsed manual analysis methodologies. This change in paradigm can be attributed to the significant gains in accuracy, repeatability and reproducibility achieved with digital image analysis. What’s more, digital microscope systems can automatically generate measurement, statistical and analysis spreadsheets and reports, all with minimal operator interaction. Images and related data can be archived in a highly structured, organized digital medium, allowing for fast and effortless retrieval.V&S


For instance, ASTM E45-11 Standard, ASTM International, 100 Barr Harbor Drive, PO Box C700, West Conshohocken, PA, 19428-2959 USA


Prior to the 1990s, analog closed-circuit television (CCTV) cameras were commonplace.

These CCTVs provided only about 640 x 480 pixels (0.3 Megapixel) of resolution.

Modern computers accomplish data crunching tasks in seconds, and with data storage no longer at a premium

Storage capacity for many thousands of images can be accommodated in most laboratory environments. Newer microscope-specific digital cameras provide enough pixels to properly digitize fine detail.