Scholarlink

<p>Due to limitations in materials and mass production processes, triaxial fluxgate sensors generally suffer from measurement errors such as non-orthogonality, zero drift, and inconsistent scale factors (referred to as “steering errors”). These errors introduce noise ranging from tens to hundreds of nT in UAV aeromagnetic investigation, which cannot be effectively eliminated by conventional magnetic compensation or cutline leveling operations in severe cases, thereby restricting the application effectiveness of aeromagnetic investigation. To address this issue, a data-driven correction method combining “fluxgate quadratic features + machine learning” is proposed: Using the square of the optically pumped total field as the label, the triaxial outputs of the fluxgate are expanded into quadratic polynomial features to establish a machine learning model, achieving steering error correction. The 1:10,000 aeromagnetic experiment in Inner Mongolia shows that the accuracy of the fluxgate aeromagnetic data is improved from 193.62nT before correction to 6.42nT after correction, with a significant improvement effect. This method does not require a uniform magnetic field environment and can be trained during field aeromagnetic exploration. After correction, the fluxgate sensor can operate independently, significantly reducing system weight and the risk of damage in optical pumping. It provides support for the R&amp;D of low-cost, high-reliability UAV aeromagnetic systems.</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

Fluxgate Aeromagnetic Correction Based on Machine Learning-based and Data-Driven Approach

  • 期刊论文

  • 发表日期：

    2025年10月8日

查看全文
  • Li-zhong Xu,
  • Gang Qiu,
  • Nan Zhang,
  • Li-na Guo,
  • Hao Wang,
  • Hua Guo

摘要

Due to limitations in materials and mass production processes, triaxial fluxgate sensors generally suffer from measurement errors such as non-orthogonality, zero drift, and inconsistent scale factors (referred to as “steering errors”). These errors introduce noise ranging from tens to hundreds of nT in UAV aeromagnetic investigation, which cannot be effectively eliminated by conventional magnetic compensation or cutline leveling operations in severe cases, thereby restricting the application effectiveness of aeromagnetic investigation. To address this issue, a data-driven correction method combining “fluxgate quadratic features + machine learning” is proposed: Using the square of the optically pumped total field as the label, the triaxial outputs of the fluxgate are expanded into quadratic polynomial features to establish a machine learning model, achieving steering error correction. The 1:10,000 aeromagnetic experiment in Inner Mongolia shows that the accuracy of the fluxgate aeromagnetic data is improved from 193.62nT before correction to 6.42nT after correction, with a significant improvement effect. This method does not require a uniform magnetic field environment and can be trained during field aeromagnetic exploration. After correction, the fluxgate sensor can operate independently, significantly reducing system weight and the risk of damage in optical pumping. It provides support for the R&D of low-cost, high-reliability UAV aeromagnetic systems.