Scholarlink

<p>To address the issue of errors caused by ground vibrations in high-precision absolute gravity measurements, a vibration error correction method based on an Adam algorithm-optimized BP neural network is proposed. By constructing a vibration error model and analyzing the influence mechanism of vibration modes on the reconstruction error of falling body trajectories, the strong correlation between time errors and vibration signals is verified. A nonlinear relationship prediction model is proposed using a BP neural network to establish the relationship between vibration signals and time coordinate errors, thereby correcting time coordinate errors and calculating gravitational acceleration. The Adam algorithm was employed to replace the traditional stochastic gradient descent (SGD) algorithm for optimizing the backpropagation neural network, effectively enhancing the model’s convergence speed and prediction accuracy. Simulation results and practical applications demonstrate that this method effectively separates vibration interference components from interference signals. In field absolute gravity observation experiments, the measurement accuracies at the national gravity benchmark point (H03), the experimental office (H09), and the mountainous site (H20) reached 1.51, 1.30, and 3.01 µGal respectively.</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

Vibration error correction in absolute gravity measurement using BP neural network

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月18日

查看全文
  • Yongzhuo Niu,
  • Qiong Wu,
  • Yang Zhang,
  • Zilu Li

摘要

To address the issue of errors caused by ground vibrations in high-precision absolute gravity measurements, a vibration error correction method based on an Adam algorithm-optimized BP neural network is proposed. By constructing a vibration error model and analyzing the influence mechanism of vibration modes on the reconstruction error of falling body trajectories, the strong correlation between time errors and vibration signals is verified. A nonlinear relationship prediction model is proposed using a BP neural network to establish the relationship between vibration signals and time coordinate errors, thereby correcting time coordinate errors and calculating gravitational acceleration. The Adam algorithm was employed to replace the traditional stochastic gradient descent (SGD) algorithm for optimizing the backpropagation neural network, effectively enhancing the model’s convergence speed and prediction accuracy. Simulation results and practical applications demonstrate that this method effectively separates vibration interference components from interference signals. In field absolute gravity observation experiments, the measurement accuracies at the national gravity benchmark point (H03), the experimental office (H09), and the mountainous site (H20) reached 1.51, 1.30, and 3.01 µGal respectively.