A study on CT detection image generation based on decompound synthesize method

Journal of X-Ray Science and Technology, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 16, 2024

Background Nuclear graphite and carbon components are vital structural elements in the cores of high-temperature gas-cooled reactors(HTGR), serving crucial roles neutron reflection, moderation, insulation. The integrity stable operation these reactors heavily depend on quality components. Helical Computed Tomography (CT) technology provides a method for detecting intelligently identifying defects within structures. However, scarcity defect datasets limits performance deep learning-based detection algorithms due to small sample sizes class imbalance. Objective Given limited number actual CT reconstruction images sparse distribution defects, this study aims address challenges imbalance model training by generating approximate augment dataset. Methods We propose novel image generation algorithm called Decompound Synthesize Method (DSM), which decomposes process into three steps: conversion, background generation, synthesis. First, STL files various industrial converted voxel data, undergo forward projection obtain corresponding images. Next, Contour-CycleGAN is employed generate synthetic that closely resemble Finally, randomly sampled from an existing library added using Copy-Adjust-Paste (CAP) method. These steps significantly expand dataset with mimic reconstructions. Results Experimental results validate effectiveness proposed tasks. Datasets generated DSM exhibit greater similarity images, when combined original data training, enhance accuracy compared only Conclusion shows promise addressing Future research can focus further optimizing refining structure models.

Language: Английский

Prediction of mechanical properties of Q345qE steel welded joints based on machine learning

Published: Dec. 22, 2024

Language: Английский