Machine learning removing method for in-situ electric heating electromagnetic interference of scanning electron microscope

CN120807358APending Publication Date: 2025-10-17ZHEJIANG UNIV
0 Cites 0 Cited by

Patent Information

Application Number
CN202510818242.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-06-18
Publication Date
2025-10-17

AI Technical Summary

Technical Problem

During in-situ high-temperature mechanical testing using a scanning electron microscope, electromagnetic interference introduced by direct current heating of metal foil samples causes electron beam deflection and 50Hz power frequency interference, affecting imaging accuracy and clarity. Traditional shielding methods are costly and hinder in-situ observations.

Method used

A machine learning method is used to construct a U-Net arc-structured generator and a PatchGAN discriminator. Combined with sliding window data expansion and feature area cropping, the interference-free in-situ electrically heated metal sheet sample image is reconstructed. The image quality is optimized by alternating training of the generator and discriminator.

Benefits of technology

It effectively removes electromagnetic interference noise, improves image clarity, provides high-quality in-situ dynamic observation images, reduces equipment costs, and facilitates material failure mechanism analysis and phase change process monitoring.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120807358A_ABST
    Figure CN120807358A_ABST
Patent Text Reader

Abstract

The invention discloses a machine learning removal method for scanning electron microscope in-situ electric heating electromagnetic interference. The method comprises the following steps: carrying out data acquisition on an image under a power-off condition and an SEM real-time image under different voltage modes; preprocessing the acquired image data and constructing a training set; constructing an electromagnetic interference picture reconstruction model comprising a generator and a discriminator, inputting an image comprising electromagnetic interference, and outputting an in-situ electric heating metal sheet sample image without interference after reconstruction; wherein the generator is of a U-Net arc structure, and each up-sampling layer is combined with a corresponding layer of the encoder through jump connection; and training the electromagnetic interference picture reconstruction model by using the training set, and reconstructing the in-situ electric heating metal sheet sample image receiving electromagnetic interference by using the trained model. High-quality images are provided for in-situ dynamic observation, and researches such as material failure mechanism analysis and phase change process monitoring are assisted.
Need to check novelty before this filing date? Find Prior Art