Author ORCID Identifier
ORCID: 0000-0001-6375-211X
Document Type
Report
Publication Date
Winter 2024
Abstract
The primary objective of this project is to investigate the framework of internal damage identification in inhomogeneous media (IHM) using deep learning techniques and wave scattering. The study focuses on two main steps: 1) analyzing the wave response pattern in homogeneous media (HM) and 2) characterizing the wave response variation (WRV) in IHM with different aggregate sizes and random distributions. The WRV patterns are then classified and used to predict internal damage using a convolutional neural network (CNN). The accuracy of the finite element (FE) model is verified using laboratory tests with artificial internal crack models of varying depths. The machine learning (ML) procedure is explained using the accumulated local effect (ALE) and cross-validated with the WRV pattern. This project is divided into five tasks: Task 1. Conduct a literature review; Task 2. Analyze WRV in HM using analytical solutions and FE models; Task 3. Investigate WRV in IHM and verify with laboratory tests; Task 4. Develop a CNN damage identification model; and Task 5. Verify the CNN prediction results with field test data.
Recommended Citation
Ham, S., Romanoschi, S., Wu, Y., & David, D. (2024). Framework of Internal Damage Identification in Inhomogeneous Medium Interweaving Wave Scattering Model and Deep Learning. Retrieved from https://repository.lsu.edu/transet_pubs/166
Comments
Tran-SET Project 22PUTA33