The Simulation-Based Approach for Random Speckle Pattern Representation in Synthetically Generated Video Sequences of...

Area of Science:

• Engineering
• Computer Science
• Materials Science

Background:

• Vision-based structural health monitoring (SHM) is crucial for infrastructure safety.
• Generating synthetic vision data is essential for training machine learning models and reducing experimental costs.
• Digital image correlation (DIC) methods analyze random speckle patterns (RSP) to measure structural deformation.

Purpose of the Study:

• To develop a methodology for generating synthetic vision data of deformable structures with random speckle patterns (RSP).
• To integrate finite element modeling (FEM) and Blender graphics for realistic synthetic data creation.
• To validate the accuracy of the synthetic data against real-world measurements.

Main Methods:

• Developed a novel RSP modeling methodology for synthetic image generation.
• Combined finite element modeling (FEM) with Blender's 3D graphics environment.
• Generated synthetic video sequences of mechanical structures with deformable RSPs.
• Processed synthetic images using digital image correlation (DIC).

Main Results:

• The proposed approach successfully generated synthetic video sequences of structures with deformable RSPs.
• Comparative analysis demonstrated high compliance between DIC-processed synthetic images and numerical data.
• The synthetic data accurately represents structural displacements.

Conclusions:

• The developed methodology provides a viable solution for generating synthetic vision data for SHM.
• This approach enhances the training of machine learning models for structural analysis.
• It enables more accurate predictions and reduces the need for extensive physical experiments.