Scholarlink

<p>Aeromagnetic surveying is an efficient and practical geophysical exploration method that has been widely applied in resource prospecting and environmental monitoring. The quality of aeromagnetic data directly determines the data’s application effect. Aeromagnetic compensation plays a pivotal role in removing the magnetic interference from the aircraft platform. However, the traditional Tolles–Lawson (T–L) model compensation method is susceptible to the influence of multicollinearity and does not account for time-varying fields generated by onboard electronics. In addition, the existing machine learning methods often converge to a local optimum and degrade model performance during hyperparameter optimization. To address these issues, this study proposes a genetic algorithm-optimized XGBoost model (GA–XGBoost) with an optimal set of input features including three-component magnetic fields (M), attitude angles (A), and position features (P) to construct a nonlinear aeromagnetic compensation model. Based on the simulated figure of merit (FOM) flight aeromagnetic data with and without noise and the real FOM flight aeromagnetic data collected by Sander Geophysics Ltd. (SGL) near Ottawa, Ontario, Canada, the GA–XGBoost model with the input feature MPA exhibits the best compensation performance in complex magnetic interference environments, compared with the compensation results of T–L, back propagation neural network (BPNN), 1D convolutional neural network (1DCNN), and XGBoost models. Finally, we use the GA–XGBoost model on real survey lines collected by SGL. The root mean square error (RMSE) of the crossover points of the measured total magnetic field data compensated by the GA–XGBoost model is comparable to the compensated result by SGL, marginally better than that of the T–L and XGBoost model. Both the simulated and real measured aeromagnetic data compensation results have verified that the GA–XGBoost model with input feature MPA is an efficient method for aeromagnetic compensation.</p> Graphical abstract <p></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

An aeromagnetic interference compensation method by combining the XGBoost and genetic algorithm: a case study from Ottawa, Canada

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月27日

查看全文
  • Liangfeng Mo,
  • Yuan Yuan,
  • Zijin Li,
  • Wenbo Jin,
  • Da Li,
  • Zhongshan Jiang,
  • Bin Wu,
  • Guochao Wu

摘要

Aeromagnetic surveying is an efficient and practical geophysical exploration method that has been widely applied in resource prospecting and environmental monitoring. The quality of aeromagnetic data directly determines the data’s application effect. Aeromagnetic compensation plays a pivotal role in removing the magnetic interference from the aircraft platform. However, the traditional Tolles–Lawson (T–L) model compensation method is susceptible to the influence of multicollinearity and does not account for time-varying fields generated by onboard electronics. In addition, the existing machine learning methods often converge to a local optimum and degrade model performance during hyperparameter optimization. To address these issues, this study proposes a genetic algorithm-optimized XGBoost model (GA–XGBoost) with an optimal set of input features including three-component magnetic fields (M), attitude angles (A), and position features (P) to construct a nonlinear aeromagnetic compensation model. Based on the simulated figure of merit (FOM) flight aeromagnetic data with and without noise and the real FOM flight aeromagnetic data collected by Sander Geophysics Ltd. (SGL) near Ottawa, Ontario, Canada, the GA–XGBoost model with the input feature MPA exhibits the best compensation performance in complex magnetic interference environments, compared with the compensation results of T–L, back propagation neural network (BPNN), 1D convolutional neural network (1DCNN), and XGBoost models. Finally, we use the GA–XGBoost model on real survey lines collected by SGL. The root mean square error (RMSE) of the crossover points of the measured total magnetic field data compensated by the GA–XGBoost model is comparable to the compensated result by SGL, marginally better than that of the T–L and XGBoost model. Both the simulated and real measured aeromagnetic data compensation results have verified that the GA–XGBoost model with input feature MPA is an efficient method for aeromagnetic compensation.

Graphical abstract