摘要
We propose a novel method that combines gray system theory and robust M-estimation method to suppress the interference in controlled-source electromagnetic data. We estimate the standard deviation of the data using a gray model because of the weak dependence of the gray system on data distribution and size. We combine the proposed and threshold method to identify and eliminate outliers. Robust M-estimation is applied to suppress the effect of the outliers and improve the accuracy. We treat the M-estimators of the preserved data as the true data. We use our method to reject the outliers in simulated signals containing noise to verify the feasibility of our proposed method. The processed values are observed to be approximate to the expected values with high accuracy. The maximum relative error is 3.6676%, whereas the minimum is 0.0251%. In processing field data, we observe that the proposed method eliminates outliers, minimizes the root-mean-square error, and improves the reliability of controlled-source electromagnetic data in follow-up processing and interpretation.
针对强干扰导致人工源频率域电磁法测量结果出现偏倚的问题,研究将灰色系统理论和稳健M-估计综合用于数据处理的新方法。由于灰色系统理论对数据分布类型及数量依赖程度低,本文通过灰色建模求解测量数据的标准差,结合阈值法识别、剔除异常值。采用稳健M-估计估算测量结果,以压制异常值影响,提高处理结果的精度。最后将保留数据的M-估计值视为真实值。为验证所述方法处理效果,仿真生成含噪信号,剔除其有效频率频谱数据中的异常值,处理结果可高度逼近理论值,其最大相对误差为3.6676%,最小仅为0.0251%。然后对野外实测数据进行处理,并通过均方根误差、电位差频谱数据和视电阻率曲线形态对处理效果进行评价。研究结果表明:本文方法可以在保留可信数据的同时有效剔除异常值,均方根误差大幅下降,为人工源电磁法资料后续处理解释提供可靠的数据。
基金
supported by the National Natural Science Foundation of China(No.41227803)
the State High-Tech Development Plan of China(No.2014AA06A602)
the Fundamental Research Funds for the Central Universities of Central South University(No.2017557)
