GRAIN SIZE DISTRIBUTION ANALYSIS ON POROUS GRANULAR MEDIA MODEL USING PRINCIPAL COMPONENT ANALYSIS

• Main Author: Aliy Andra Putra, Hilmy
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/46634
• Institution: Institut Teknologi Bandung

In this research, a modeling of porous granular media (MGB) has been carried out, followed by processing, and image analysis to obtain information about the model’s grain size distribution, including the sorting type of the model. This distribution refers to the scale of rock sedimentation sizes with various types and sizes of rocks. In the modeling scheme, binary images are produced by deposition of pure spherical grains at randomly determined positions and sizes of grains. The deposited grains can overlap with the degree of overlap (?) which satisfies three kinds of conditions, namely penetrable (? = 100%), semi-penetrable (0<? <100%), and impenetrable (? = 0%). This degree of overlap affects the grain structure to be further analyzed. To produce a good analysis, image processing is carried out using watershed segmentation to separate overlapping grains by means of the H-minimum transformation, in order to set local minimums and to prevent over-segmentation. This segmentation aimed to obtain the identity of each objects, which can later be easily analyzed. From the results of the image processing, further analysis of the characteristics of the radius of the grain size is done using Principal Component Analysis (PCA). The results of the application of the PCA method are orthogonal eigenvectors. From the vectors, the direction of the main axis of the grains that are perpendicular to each other can then be obtained. Thus for 2D, four principal axes were obtained, while for the case of 3D, six principal axes were produced. The output from the modeling, processing, and analysis of the image are the average radius (obtained from the principal axes), area/volume, surface area, and centroid which can then be made in the form of distribution for further analysis. Subsequently, analysis were done by a comparison between the distribution of the radius and grain area/volume of the generated models, against the modeling parameters used. The purpose of the comparison is to obtain an image processing and analysis procedure that can yield an accurate grain size distribution analysis for both 2D and 3D media.