Scholarlink

<p>This research presents a novel machine learning algorithm based on the Gaussian Mixture Model (GMM) to enhance the demodulation accuracy of the FBG sensor. The proposed method leverages the probabilistic clustering ability of GMM to distinguish multiple Bragg wavelength shifts under varying environmental conditions. The algorithm involves pre-processing raw FBG spectra, extracting key features, and applying GMM-based clustering to classify and predict wavelength shifts. The true peak mean, which is determined by updating the weight or mixing coefficient, is dependent on the highest height of the Gaussian signal. Calculating the strain and temperature applied to FBGs requires knowing the true peak and variations in peak wavelength, which may be found using the exact mean. Experimental validation demonstrates that the developed approach improves wavelength detection accuracy and noise robustness compared to other existing peak detection methods. The results indicate that the GMM-based machine learning technique is a promising solution for high-precision FBG sensors with good efficiency and accuracy with respect to the mean error as 0.031 pm and RMSE as 0.043 pm.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Developed machine learning algorithm for fiber Bragg grating sensor using Gaussian mixing method

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月20日

查看全文
  • Sunil Kumar,
  • Yogesh K M,
  • Sagar D

摘要

This research presents a novel machine learning algorithm based on the Gaussian Mixture Model (GMM) to enhance the demodulation accuracy of the FBG sensor. The proposed method leverages the probabilistic clustering ability of GMM to distinguish multiple Bragg wavelength shifts under varying environmental conditions. The algorithm involves pre-processing raw FBG spectra, extracting key features, and applying GMM-based clustering to classify and predict wavelength shifts. The true peak mean, which is determined by updating the weight or mixing coefficient, is dependent on the highest height of the Gaussian signal. Calculating the strain and temperature applied to FBGs requires knowing the true peak and variations in peak wavelength, which may be found using the exact mean. Experimental validation demonstrates that the developed approach improves wavelength detection accuracy and noise robustness compared to other existing peak detection methods. The results indicate that the GMM-based machine learning technique is a promising solution for high-precision FBG sensors with good efficiency and accuracy with respect to the mean error as 0.031 pm and RMSE as 0.043 pm.