Digital Image-based Identification Method for the Determination of the Particle Size Distribution of Dam Granular Material

• Published in: KSCE journal of civil engineering Vol. 22; no. 8; pp. 2820 - 2833
• Main Authors: Zhang, Shi-lin, Wu, Gao-jian, Yang, Xing-guo, Jiang, Wan-hong, Zhou, Jia-wen
• Format: Journal Article
• Language: English
• Published: Seoul Korean Society of Civil Engineers 01.08.2018; Springer Nature B.V; 대한토목학회
• Subjects: Algorithms; Case studies; Civil Engineering; Dam safety; Digital imaging; Earth; Engineering; Error correction; Evaluation; Genetic algorithms; Geotechnical Engineering; Geotechnical Engineering & Applied Earth Sciences; Graders; Grading; Granular materials; Hydroelectric power; Hydroelectric power stations; Identification; Identification methods; Image acquisition; Image processing; Industrial Pollution Prevention; Information processing; Labour; Mathematical models; Methods; Neural networks; Particle size; Particle size distribution; Size distribution; Statistical analysis; Structural engineering; Structural safety; Technology; 토목공학; digital image processing; large data; neural network algorithm; particle size distribution; parameterize; dam granular material
• ISSN: 1226-7988; 1976-3808
• DOI: 10.1007/s12205-017-0304-8

The Particle Size Distribution (PSD) properties of dam granular material plays an important role in the construction process of earth-rock dams, as it can affect the filling quality and structural safety. However, the conventional sieving method employed to check the PSD is labor-intensive, time-consuming and not highly accurate. In this study, a digital image-based identification method is presented for the determination of the PSD of dam granular material, which mainly incorporates image acquisition technology, a large database and a neural network. Digital Image Processing (DIP) technology is used to recognize the geometric size and grading curve of dam granular materials at a small scale, while statistical distribution models are used to determine the characteristic parameters of the grading curve and convert the graphical curve into mathematical variables. Furthermore, a large database and a BP neutral algorithm, which is improved using a genetic algorithm, are introduced as tools to reveal the implicit relationship between the DIP and sieving grading curves to correct the error of identification. A case study for the Changheba Hydropower Station is used to illustrate the implementation details of the presented method. The identification results demonstrate that the presented method can acquire and assess the gradation in spite of a degree of error, which can be decreased when more advanced DIP technologies are explored, the amount of data in the database is increased, and a more optimized network algorithm is adopted.