磁共振探测技术(Magnetic Resonance Sounding,MRS)以其无损、定量、直接等优势,被广泛应用于地下水调查、水文环境评价以及灾害水源预警等领域.在实际应用中,强工频谐波和随机噪声等严重影响MRS信号的质量,导致后续水文参数解释不准确.针对这一问题,提出谐波建模和自相关相结合的方法进行消噪以及信号特征参数提取.首先构建工频谐波模型,针对建模算法严重依赖工频基频精度的问题,采用自适应扫描方式搜索方案,大幅提高搜索准确度和速度;其次推导了MRS信号自相关表达式,提出了自相关参数提取的非线性拟合方法.仿真数据结果表明,建模消噪方法有效消除了工频谐波,信噪比平均提升了17.03 dB;自相关处理后,信噪比进一步提升了16.10 dB,初始振幅和弛豫时间参数提取结果的准确度比处理前分别提高了3.8倍和2.8倍.通过不同信噪比和弛豫时间的重复实验,得到当噪声水平小于200 nV和弛豫时间大于200 ms时,自相关参数提取具有较高的稳定性.最后,通过野外实测数据处理实验,进一步验证了联合消噪和参数提取方法的有效性.%Magnetic Resonance Sounding (MRS) is a non-invasive geophysical technique providing the ability to quantitatively and directly detect the aquifer properties,which is commonly used in groundwater survey,hydrological assessment and advanced detection of a water source that may cause disastrous accident in the underground engineering.In field applications,high power line harmonic,ambient noise and other interferences seriously affect the quality of the detected MRS signal,resulting in decreasing the interpretation accuracy of hydrological parameters of the subsurface aquifers.To tackle this problem,a novel method based on adaptive harmonic modeling and autocorrelation is proposed to reduce the power line harmonic and ambient noise,and extract the characteristic parameters of the MRS signal Firstly,the power line and its harmonics are modeled based on the Fourier transform.To solve the problem that the modeling algorithm is heavily dependent on the estimation precision of the base frequency,an adaptive scanning method is adopted to improve both the accuracy and search speed.Secondly,the autoeorrelation expression of MRS signal is deduced,and a nonlinear fitting method for the parameter extraction from the autocorrelation sequence is proposed.The synthetic results show that the power line harmonics can be eliminated effectively and the signal-to-noise ratio (SNR) is improved by 17.03 dB after the adaptive harmonic modeling.After the autocorrelation processing,the SNR rises by 16.10 dB further.The accuracy of the initial amplitude and relaxation time parameters estimated from the autocorrelation of the MRS signal is 3.8 times and 2.8 times higher than that from the original MRS signal.The results of repeated experiments with different SNRs and relaxation times show that the autocorrelation parameter extraction method is particularly stable and reliable when the noise level is less than 200 nV and the relaxation time is over 200 ms.Finally,the validity of combined denoising and parameter extraction method is verified by the results of field measurement and borehole logging.
展开▼
