Artificial neural networks for prediction of best image reconstruction parameters for concrete elements

• Grant number: 23/04604-0
• Support Opportunities: Scholarships abroad - Research Internship - Master's degree
• Start date: August 01, 2023
• End date: October 31, 2023
• Field of knowledge: Engineering - Civil Engineering - Structural Engineering
• Principal Investigator: Vladimir Guilherme Haach
• Grantee: Lara Guizi Anoni
• Supervisor: Lev Khazanovich
• Host Institution: Escola de Engenharia de São Carlos (EESC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Institution abroad: University of Pittsburgh (Pitt), United States
• Associated to the scholarship: 22/15224-0 - Ultrasound tomography of concrete elements: Backprojection x Algebraic Methods, BP.MS

Abstract

Ultrasound tests have been widely applied for the evaluation of concrete elements since they cause no damage to the inspected structure. Fast detection of non-homogeneities of the structure in a simple analysis and the generation of tomographic images of internal sections in a more complex analysis are among of its achievements. Concrete tomography research involves various image reconstruction techniques, including algebraic and projection methods, with specific definitions that contribute to image accuracy. Tomography can be generated based on ultrasound wave time of flight or variations in the wave amplitude. Data can be manipulated with interpolation rules, and trajectory update algorithms can be used. The quality of generated images depends on these definitions. This study proposes using Artificial Neural Networks (ANN) to predict the best image reconstruction technique and related specific parameters. Several simulated images generated with ultrasonic test analysis software will aid in training, testing, and optimizing the ANN. The goal is to improve the image generation capacity from ultrasonic test results for voids and damage detection in concrete specimens. (AU)