New methods for automatic quantification of microstructural features using digital image processing

Abstract

Thermal and mechanical processes alter the microstructure of materials, which determines their mechanical properties. This makes reliable microstructural analysis important to the design and manufacture of components. However, the analysis of complex microstructures, such as Ti6Al4V, is difficult and typically requires expert materials scientists to manually identify and measure microstructural features. This process is often slow, labour intensive and suffers from poor repeatability. This paper overcomes these challenges by proposing a new set of automated techniques for 2D microstructural analysis. Digital image processing algorithms are developed to isolate individual microstructural features, such as grains and alpha lath colonies. A segmentation of the image is produced, where regions represent grains and colonies, from which morphological features such as; grain size, volume fraction of globular alpha grains and alpha colony size can be measured. The proposed measurement techniques are shown to obtain similar results to existing manual methods while drastically improving speed and repeatability. The benefits of the proposed approach when measuring complex microstructures are demonstrated by comparing it with existing analysis software. Using a few parameter changes, the proposed techniques are effective on a variety of microstructure types and both SEM and optical microscopy image

Texture-Based Metallurgical Phase Identification in Structural Steels: A Supervised Machine Learning Approach

“…It was important to note that the textural features of the metallurgical phases extracted in this study were continuous variables. Therefore, to evaluate the conditional probabilities of these textural features, conditional probability density functions were used as proxy measures in place of actual conditional probabilities in Equation (9). The probability density functions considered in this study were a normal distribution and a Weibull distribution.…”

“…The images obtained from the light-optical microscope are then analyzed manually following the standard protocols provided by the American Society for Testing and Materials (ASTM) standards E114 [7] and E562 [8]. However, this process is labor-intensive and a subjective process prone to poor repeatability and interpretation of results [9]. Therefore, automated digital image processing-based techniques have been developed in recent years to overcome these issues and accurately quantify the microstructural features for better design of engineering components.…”

“…Zhang and Liu [21] implemented the canny edge algorithm for identification of phases in Ti-6Al-4V titanium alloy, which involved a series of preprocessing steps such as noise removal and uneven illumination. Campbell et al [9] developed a watershed algorithm-based technique with pre-and post-processing steps to segment the touching grains in a titanium (Ti-6Al-4V) microstructure. An automated microstructural characterization software MiPAR ® was developed by Sosa et al [19] that included a segmentation Automated image segmentation procedures have been proposed and developed in recent years by numerous researchers to identify and quantify microstructural features [20].…”

Microstructure Instance Segmentation from Aluminum Alloy Metallographic Image Using Different Loss Functions

“…Image processing-based methods segment the microstructure using typical image processing methods. For example, the watershed method, the typical digital image processing method, is used to segment microstructures [12]. The improved mean shift algorithm is used to segment the grains based on the characteristics of each region [13].…”

Computer Methods for Constructing Parametric Statistically Equivalent Models of High-Strength Cast Iron Microstructure to Analyze It’s Elastic Characteristics

“…Другу групу досліджень [7][8][9][10][11][12][13] об'єднує застосування технологій комп'ютерного бачення. Розпізнавання образів зображення використовують для: класифікації структури за відповідними ознаками [7; 10]; оцінки кількості дефектів (вироджених графітових вузликів [8]); сегментації складних мікроструктур [11], знаходження розмірів частинок і розподілу їх на площині, а також застосовуються як вхідні дані для тренування нейронних мереж для прогнозування властивостей матеріалу відповідно до зображення його мікроструктури [12][13].…”