Global surface wave inversion with model constraints‡

Practical applications of surface wave inversion demand reliable inverted shear‐wave profiles and a rigorous assessment of the uncertainty associated to the inverted parameters. As a matter of fact, the surface wave inverse problem is severely affected by solution non‐uniqueness: the degree of non‐uniqueness is closely related to the complexity of the observed dispersion pattern and to the experimental inaccuracies in dispersion measurements. Moreover, inversion pitfalls may be connected to specific problems such as inadequate model parametrization and incorrect identification of the surface wave modes. Consequently, it is essential to tune the inversion problem to the specific dataset under examination to avoid unnecessary computations and possible misinterpretations. In the heuristic inversion algorithm presented in this paper, different types of model constraints can be easily introduced to bias constructively the solution towards realistic estimates of the 1D shear‐wave profile. This approach merges the advantages of global inversion, like the extended exploration of the parameter space and a theoretically rigorous assessment of the uncertainties on the inverted parameters, with the practical approach of Lagrange multipliers, which is often used in deterministic inversion, which helps inversion to converge towards models with desired properties (e.g., ‘smooth’ or ‘minimum norm' models). In addition, two different forward kernels can be alternatively selected for direct‐problem computations: either the conventional modal inversion or, instead, the direct minimization of the secular function, which allows the interpreter to avoid mode identification. A rigorous uncertainty assessment of the model parameters is performed by posterior covariance analysis on the accepted solutions and the modal superposition associated to the inverted models is investigated by full‐waveform modelling. This way, the interpreter has several tools to address the more probable sources of inversion pitfalls within the framework of a rigorous and well‐tested global inversion algorithm. The effectiveness and the versatility of this approach, as well as the impact of the interpreter's choices on the final solution and on its posterior uncertainty, are illustrated using both synthetic and real data. In the latter case, the inverted shear velocity profiles are blind compared with borehole data.     

航空电磁拟三维模型空间约束反演

为了克服时间域航空电磁数据单点反演结果中常见的电阻率或层厚度横向突变造成数据难以解释的问题,通过引入双向约束实现航空电磁拟三维空间约束反演.除考虑沿测线方向相邻测点之间的横向约束外,同时还考虑了垂直测线方向测点在空间上的相互约束.为此,首先设计拟三维模型中固定层厚和可变层厚两种空间约束反演方案,然后通过在目标函数中引入沿测线和垂直测线方向上的模型参数约束矩阵,并使用L-BFGS算法使目标函数最小化,获得最优拟三维模型空间反演解.基于理论模型和实测数据反演,对单点反演与两种空间约束反演方案的有效性进行比较,证明本文空间约束反演算法对于噪声的压制效果好,反演的界面连续光滑,同时内存需求和反演时间少,是一种快速有效的反演策略.     

联合利用走时与波形反演技术研究地壳三维速度结构(Ⅱ)——应用

使用阻尼最小二乘法进行震源参数和地壳三维速度结构的走时联合反演．所用资料为Ｓ波和Ｐ波到时差,并用人工地震资料的二维解释结果作为三维速度模型的特定约束条件．为建立初始模型,又利用天然地震构成了准二维剖面．在走时反演基础上,利用遗传算法进行了几个地震事件的波形反演尝试,并对走时反演获得的地壳速度结构模型的局部进行了修正．以３４°～４２°Ｎ,９４°～１１２°Ｅ作为研究区域,在该区域中收集了１９８６年以来大量地震的Ｓ波和Ｐ波到时差资料,７条人工地震二维速度剖面资料和２个数字化地震台的几个地震的三分向记录资料．对这些资料进行了处理,最后得出了０～２５ｋｍ深度不同截面的速度分布,并对所得结果进行了分析．     

南海东北部地震海洋学联合调查与反演

通过地震海洋学联合调查航次,获得了同时采集的高分辨率多道地震数据与水文数据（XBT）联合调查数据.通过针对性处理获得了清晰的海洋水体反射地震剖面,捕捉到南海东北部次表层涡旋,证明了该调查方法的有效性.利用背景场约束反演方法得到海水的波阻抗、声速、密度、温度和盐度剖面,揭示了涡旋具有相对低速、高密、低温与高盐特征.处理和反演结果表明,在浅部和部分噪声较强和无约束反演区域,结果误差较大,可靠性较低.调查结果说明,获得较高信噪比的原始高分辨率地震资料采集技术方法、联合约束数据的获取、针对性的特殊处理方法和有效的反演方法是地震海洋学资料有效分析利用的重要保证.     

A lateral model parameter correlation procedure for one-dimensional inverse modelling

We present a new, fast and versatile method, the lateral parameter correlation method, of invoking lateral smoothness in model sections of one-dimensional (1D) models. Modern, continuous electrical and electromagnetic methods are capable of recording very large data sets and except for a few cases, standard inversion methodology still relies on 1D models. In environments where the lateral rate of change of resistivity is small, 1D inversion can be justified but model sections of concatenated 1D models do not necessarily display the expected lateral smoothness.
The lateral parameter correlation method has three steps. First, all sounding data are inverted individually. Next, a laterally smooth version of each model parameter, one at a time, is found by solving a simple constrained inversion problem. Identity is postulated between the uncorrelated and correlated parameters and the equations are solved including a model covariance matrix. As a last step, all sounding data are inverted again to produce models that better fit the data, now subject to constraints by including the correlated parameter values as a priori values. Because the method separates the inversion from the correlation it is much faster than methods where the inversion and correlation are solved simultaneously, typically with a factor of 200–500.
Theoretical examples show that the method produces laterally smooth model sections where the main influence comes from the well-determined parameters in such a way that problems with equivalence and poor resolution are alleviated. A field example is presented, demonstrating the improved resolution obtained with the lateral parameter correlation method. The method is very flexible and is capable of coupling models from inversion of different data types and information from boreholes.     

各向异性TI介质qP反射波走时层析成像

地震走时层析成像是反演地层各向异性参数分布的有效方法,但是关于地震各向异性介质走时层析成像的研究并不多,其技术远远没有达到成熟的阶段.在野外数据采集时,地表反射波观测方式相对井间和垂直地震剖面观测方式的成本更低,利用qP反射波走时反演各向异性参数具有更加广泛的实用价值.本文实现的TI介质地震走时层析成像方法结合了TI介质反射波射线追踪算法、走时扰动方程和非线性共轭梯度算法,它可以对任意强度的TI介质模型进行反演,文中尝试利用qP反射波走时重建TI介质模型的参数图像.利用qP反射波对层状介质模型和块状异常体模型进行走时反演,由于qP波相速度对弹性模量参数和Thomsen参数的偏微分不同,所以可以分别反演弹性模量参数和Thomsen参数.数值模拟结果表明:利用qP反射波可以反演出TI介质模型的弹性模量参数与Thomsen参数,不同模型的走时迭代反演达到了较好的收敛效果,与各向同性介质走时反演结果相比较,各向异性介质走时反演结果具有较好的识别能力.     

Generalised Series Expansion (GSE) used in DC geoelectric–seismic joint inversion

Horizontally layered (1D) earth models are often assumed as a model estimate for the interpretation of geophysical data measured along 2D geological structures. In this process, the individual data sets are usually inverted independently, and it is considered only in a later phase of interpretation that these local (1D) models have common characteristic features. Taking account of these common attributes, instead of the successive independent interpretations, the lateral variations of geometrical and petrophysical parameters can be efficiently determined for the whole 2D structure by applying a series expansion. Using global basis functions, two advantages can be achieved: (i) choosing an appropriate number of basis functions helps us to restrict the complexity of the model; (ii) the integration of all the data sets measured along the profile gives rise to the application of simultaneous or joint inversion methods. This results in a decrease of the number of independent unknowns, a higher stability during the inversion and a more accurate and reliable parameter estimation.In this paper, a joint inversion algorithm is presented using DC geoelectric apparent resistivities and refraction seismic travel times measured along various layouts above a 2D geological model. To describe lateral variations series, expansions are used, and furthermore, to improve the often used approximation of a (locally) 1D forward modelling, the integral mean value of the horizontally changing model parameters (calculated along an appropriately defined interval) is introduced. We call the inversion procedure that combines series expansions and the concept of integral mean Generalised Series Expansion (GSE) inversion. The method was developed and tested for both the simultaneous (integrating data sets of one method or methods on the same physical basis) and the joint inversion (where data sets of methods on different physical bases are joined together), using synthetic and field data sets. It is also demonstrated that the equivalence problem inherent in the independent inversion of DC geoelectric data can efficiently be resolved by the use of the joint GSE inversion method in the cases of conductive and resistive equivalent geological models.     

Piecewise‐continuous velocity inversion

We suggest a new method to determine the piecewise‐continuous vertical distribution of instantaneous velocities within sediment layers, using different order time‐domain effective velocities on their top and bottom points. We demonstrate our method using a synthetic model that consists of different compacted sediment layers characterized by monotonously increasing velocity, combined with hard rock layers, such as salt or basalt, characterized by constant fast velocities, and low velocity layers, such as gas pockets. We first show that, by using only the root‐mean‐square velocities and the corresponding vertical travel times (computed from the original instantaneous velocity in depth) as input for a Dix‐type inversion, many different vertical distributions of the instantaneous velocities can be obtained (inverted). Some geological constraints, such as limiting the values of the inverted vertical velocity gradients, should be applied in order to obtain more geologically plausible velocity profiles. In order to limit the non‐uniqueness of the inverted velocities, additional information should be added. We have derived three different inversion solutions that yield the correct instantaneous velocity, avoiding any a priori geological constraints. The additional data at the interface points contain either the average velocities (or depths) or the fourth‐order average velocities, or both. Practically, average velocities can be obtained from nearby wells, whereas the fourth‐order average velocity can be estimated from the quartic moveout term during velocity analysis. Along with the three different types of input, we consider two types of vertical velocity models within each interval: distribution with a constant velocity gradient and an exponential asymptotically bounded velocity model, which is in particular important for modelling thick layers. It has been shown that, in the case of thin intervals, both models lead to similar results. The method allows us to establish the instantaneous velocities at the top and bottom interfaces, where the velocity profile inside the intervals is given by either the linear or the exponential asymptotically bounded velocity models. Since the velocity parameters of each interval are independently inverted, discontinuities of the instantaneous velocity at the interfaces occur naturally. The improved accuracy of the inverted instantaneous velocities is particularly important for accurate time‐to‐depth conversion.     

基于阵列感应与自然电位联合反演地层水电阻率

原状地层水电阻率是重要的储层参数,也是进行精细储层评价的基础.基于泥浆侵入数值模拟与侵入过程中井周岩石物理特征分析,确定了薄膜电位的产生位置,针对储层高、低侵等不同侵入特征,提出了可适用于包括存在"低阻环"等不同侵入特征时储层电阻率分布的"五参数"地层模型,基于几何因子理论与有限元方法,建立了阵列感应与自然电位测井联合反演算法,实现了地层电阻率参数反演,重构了地层径向电阻率剖面,进而精确求取了地层水电阻率.通过对实际资料处理表明:反演算法稳定可靠,"五参数"模型能很好地表征储层电阻率分布形态,重构储层电阻率剖面,确定薄膜电位产生位置;基于阵列感应与自然电位的联合反演,能精确计算原状地层水电阻率,为储层评价与流体性质识别提供依据.     

Comparing the performances of four stochastic optimisation methods using analytic objective functions, 1D elastic full‐waveform inversion,and residual static computation

We compare the performances of four stochastic optimisation methods using four analytic objective functions and two highly non‐linear geophysical optimisation problems: one‐dimensional elastic full‐waveform inversion and residual static computation. The four methods we consider, namely, adaptive simulated annealing, genetic algorithm, neighbourhood algorithm, and particle swarm optimisation, are frequently employed for solving geophysical inverse problems. Because geophysical optimisations typically involve many unknown model parameters, we are particularly interested in comparing the performances of these stochastic methods as the number of unknown parameters increases. The four analytic functions we choose simulate common types of objective functions encountered in solving geophysical optimisations: a convex function, two multi‐minima functions that differ in the distribution of minima, and a nearly flat function. Similar to the analytic tests, the two seismic optimisation problems we analyse are characterised by very different objective functions. The first problem is a one‐dimensional elastic full‐waveform inversion, which is strongly ill‐conditioned and exhibits a nearly flat objective function, with a valley of minima extended along the density direction. The second problem is the residual static computation, which is characterised by a multi‐minima objective function produced by the so‐called cycle‐skipping phenomenon. According to the tests on the analytic functions and on the seismic data, genetic algorithm generally displays the best scaling with the number of parameters. It encounters problems only in the case of irregular distribution of minima, that is, when the global minimum is at the border of the search space and a number of important local minima are distant from the global minimum. The adaptive simulated annealing method is often the best‐performing method for low‐dimensional model spaces, but its performance worsens as the number of unknowns increases. The particle swarm optimisation is effective in finding the global minimum in the case of low‐dimensional model spaces with few local minima or in the case of a narrow flat valley. Finally, the neighbourhood algorithm method is competitive with the other methods only for low‐dimensional model spaces; its performance sensibly worsens in the case of multi‐minima objective functions.     

海底广角地震剖面反演方法对比——以南海礼乐滩OBS剖面为例

广角地震测线（OBS973-2）位于南海南部陆缘,其地壳深部构造是研究南海共轭扩张及形成演化的直接证据之一.本文采用2D射线追踪技术,结合与之重合的多道地震测线（NH973-2）时深转换结果,对OBS973-2测线重新进行了正、反演研究,得到了礼乐滩及邻近海区的精细地壳结构.与前人结果相比,本文基于正反演速度模型,把测线分为陆壳区（0~200 km）、洋陆过渡区（200~280 km）和洋盆区（280~370 km）.地壳结构在不同区域差异明显,陆壳区沉积层厚度横向差异大,且速度横向不均匀,地壳整体厚度大（约20 km）,有横向速度差;洋陆过渡区速度和厚度横向均匀,地壳减薄（约8 km）;洋壳区地壳厚度减薄至6 km.与以往研究相比,新的认识集中在两个方面,（1）在方法上,综合广角地震和多道地震数据,借助正演和反演方法,能够得到更多更可靠的地壳结构信息.（2）在地壳结构上,结合广角地震与多道地震,得到洋陆过渡区莫霍面向海减薄的形态及其埋深（约12~18 km,海平面为0 km）;进一步验证礼乐滩区域在洋陆过渡区没有明显的高速层,为非火山型陆缘,其共轭扩张点为中沙地块;陆壳区上地壳强烈的拉张作用在速度模型表现出横向速度异常和低速区,在多道地震剖面上表现为大量10~20 km的正断层.     

A Coastal Aquifer Study Using Magnetotelluric and Gravity Methods in Abo Zenema, Egypt

Magnetotelluric (MT) soundings and gravity methods were employed to study the deep freshwater aquifer in the area north of Abo Zenema city on the eastern side of the Gulf of Suez, Egypt. Seven MT sites and 48 gravity stations were surveyed along northeast–southwest profiles as close as possible to a line perpendicular to the coast of the Gulf of Suez. The MT survey was conducted using high and low frequencies to investigate shallow and deep areas, respectively. One-dimensional inversion was conducted using a heuristic inversion scheme of the Bostick algorithm. The MT data were also inverted with a 2-D smooth model inversion routine using the nonlinear conjugate gradient method to infer variation in vertical and lateral resistivity inside the Earth. A 100-Ohm-m homogeneous half-space initial model was used to invert the TE mode data only. Then, the inverted model obtained from the TE mode data was used as an initial model for inversion of the TM mode data. The inverted model thus obtained from the TM mode data inversion was used as an initial model for the inversion of the joint TE and TM responses. Two-dimensional (2-D) forward modeling of the gravity data was conducted using the 2-D polygon method of Talwani’s algorithm for an arbitrarily shaped body and was based on the subsurface information from the MT survey and the available information about the geological structure of the study area. This method enabled us to obtain the basement structure of the coastal aquifer in the study area. The results from the analysis and the interpretation of MT and gravity data were used to detect and delineate the groundwater coastal aquifer in the study area.     

Three-dimensional inversion of the single-profile magnetotelluric data

It is preferable to use the three-dimensional (3D) magnetotelluric inversion, which provides volumetric geoelectric models, to handle the array input data. However, the soundings are frequently conducted on the single profiles or on the profiles that are considerably spaced apart from each other. We explore the possibilities of the 3D inversion of such data by the example of a three-layer model containing three local inhomogeneities. We previously showed that the simple processing of the data and their 1D or 2D inversion enable reconstructing the background cross section and locating all the three inhomogeneities. In the present paper, we use this information for constructing several versions of the starting model and carrying out the smoothing 3D inversion of the data. The experiments show that if the background cross section is incorporated into the starting model, the final model provided by the inversion closely reproduces the real distribution of all geoelectric parameters. At the same time, if the starting model that hosts the inhomogeneities has the form of a homogeneous half-space, the inversion is not able to reconstruct an adequate final model.

     

Three-dimensional inversion of the single-profile magnetotelluric data

It is preferable to use the three-dimensional (3D) magnetotelluric inversion, which provides volumetric geoelectric models, to handle the array input data. However, the soundings are frequently conducted on the single profiles or on the profiles that are considerably spaced apart from each other. We explore the possibilities of the 3D inversion of such data by the example of a three-layer model containing three local inhomogeneities. We previously showed that the simple processing of the data and their 1D or 2D inversion enable reconstructing the background cross section and locating all the three inhomogeneities. In the present paper, we use this information for constructing several versions of the starting model and carrying out the smoothing 3D inversion of the data. The experiments show that if the background cross section is incorporated into the starting model, the final model provided by the inversion closely reproduces the real distribution of all geoelectric parameters. At the same time, if the starting model that hosts the inhomogeneities has the form of a homogeneous half-space, the inversion is not able to reconstruct an adequate final model.     

Appraisal of equivalence and suppression problems in 1D EM and DC measurements using global optimization and joint inversion*

Global optimization with very fast simulated annealing (VFSA) in association with joint inversion is performed for 1D earth structures. The inherent problems of equivalence and suppression in electromagnetic (EM) and direct current (DC) resistivity methods are studied. Synthetic phase data from multifrequency sounding using a horizontal coplanar coil system and synthetic apparent resistivity data from Schlumberger DC resistivity measurements are inverted individually and jointly over different types of layered earth structures. Noisy data are also inverted. The study reveals that global optimization of individual data sets cannot solve inherent equivalence or suppression problems. Joint inversion of EM and DC measurements can overcome the problem of equivalence very well. However, a suppression problem cannot be solved even after combination of data sets. This study reveals that the K-type earth structure is easiest to resolve while the A-type is the most difficult. We also conclude that the equivalence associated with a thin resistive layer can be resolved better than that for a thin conducting layer.     

基于柯西分布约束和快速近端目标函数优化的三维重力反演方法

优化算法的选取在很大程度上影响着三维重力反演的计算效率,从而制约着三维重力反演的实用性.在复杂地质构造背景下,不同岩性单元之间可能会发生物性突变,产生尖锐边界.为此,本文提出了一种新的基于柯西分布约束和快速近端目标函数（Fast Proximal Objective Function,FPOF）优化的三维重力反演方法.FPOF优化方法的一个突出特点是在每一步迭代过程中逐一计算剖分网格内的未知密度参数,因此,有较低的计算复杂度和较高的计算效率.此外,目标函数中柯西范数（Cauchy norm）的引入会对反演结果施加稀疏性,有助于产生块状效果.理论模型测试表明,本文方法不仅能产生更加聚焦的反演效果,而且反演所需的时间也比传统的共轭梯度优化方法少.最后将本文方法应用于我国西部某地区实际重力数据,反演结果与已知的地质信息有较好的一致性.     

Solutions of the inherent problem of the equivalence in direct current resistivity and electromagnetic methods through global optimization and joint inversion by successive refinement of model space

The problem of equivalence in direct current (DC) resistivity and electromagnetic methods for a thin resistive and conducting layer is well‐known. Attempts have been made in the past to resolve this problem through joint inversion. However, equivalence still remains an unresolved problem. In the present study, an effort is made to reduce non‐uniqueness due to equivalence using global optimization and joint inversion by successive refinement of the model space. A number of solutions derived for DC resistivity data using very fast simulated annealing global inversion that fits the observations equally well, follow the equivalence principle and show a definite trend. For a thin conductive layer, the quotient between resistivity and thickness is constant, while for a resistive one, the product between these magnitudes is constant. Three approaches to obtain very fast simulated annealing solutions are tested. In the first one, layer resistivities and thicknesses are optimized in a linear domain. In the second, layer resistivities are optimized in the logarithmic domain and thicknesses in the linear domain. Lastly, both layer resistivities and thicknesses are optimized in the logarithmic domain. Only model data from the mean models, corresponding to very fast simulated annealing solutions obtained for approach three, always fit the observations. The mean model defined by multiple very fast simulated annealing solutions shows extremely large uncertainty (almost 100%) in the final solution after inversion of individual DC resistivity or electromagnetic (EM) data sets. Uncertainty associated with the intermediate resistive and conducting layers after global optimization and joint inversion is still large. In order to reduce the large uncertainty associated with the intermediate layer, global optimization is performed over several iterations by reducing and redefining the search limits of model parameters according to the uncertainty in the solution. The new minimum and maximum limits are obtained from the uncertainty in the previous iteration. Though the misfit error reduces in the solution after successive refinement of the model space in individual inversion, it is observed that the mean model drifts away from the actual model. However, successive refinement of the model space using global optimization and joint inversion reduces uncertainty to a very low level in 4–5 iterations. This approach works very well in resolving the problem of equivalence for resistive as well as for conducting layers. The efficacy of the approach has been demonstrated using DC resistivity and EM data, however, it can be applied to any geophysical data to solve the inherent ambiguities in the interpretations.     

Site characterization using full waveform inversion

We discuss recent progress in the full-waveform-based imaging of probed soils, with geotechnical site characterization applications in mind. The primary goal is the reconstruction of the material profile of near-surface, arbitrarily heterogeneous formations, in terms of the formation's spatially distributed elastic properties, using elastic waves as the probing agent.We describe first the formulation and numerical resolution of the underlying time-dependent inverse medium problem; we report briefly on numerical experiments using synthetic data and artificial target soil profiles. These demonstrate robust reconstruction. We then report extensively on the details of a field experiment, whose records we subsequently used to drive the inversion algorithms in order to characterize the site where the field experiment took place. Lastly, we compare the inverted site profile with profiles obtained using the Spectral-Analysis-of-Surface-Waves (SASW) method, in an attempt to compare our methodology against a widely used concurrent inversion approach. We also compare the inverted profile at select locations with the results of independently performed CPT tests.Overall, whether exercised by synthetic or by physical data, the full waveform inversion method we discuss herein appears quite promising for the robust subsurface imaging of near-surface deposits in support of geotechnical site characterization investigations.     

Multistep inversion workflow for 3D long‐offset damped elastic waves in the Fourier domain

We present a new workflow for imaging damped three‐dimensional elastic wavefields in the Fourier domain. The workflow employs a multiscale imaging approach, in which offset lengths are laddered, where frequency content and damping of the data are changed cyclically. Thus, the inversion process is launched using short‐offset and low‐frequency data to recover the long spatial wavelength of the image at a shallow depth. Increasing frequency and offset length leads to the recovery of the fine‐scale features of the model at greater depths. For the fixed offset, we employ (in the imaging process) a few discrete frequencies with a set of Laplace damping parameters. The forward problem is solved with a finite‐difference frequency‐domain method based on a massively parallel iterative solver. The inversion code is based upon the solution of a least squares optimisation problem and is solved using a nonlinear gradient method. It is fully parallelised for distributed memory computational platforms. Our full‐waveform inversion workflow is applied to the 3D Marmousi‐2 and SEG/EAGE Salt models with long‐offset data. The maximum inverted frequencies are 6 Hz for the Marmousi model and 2 Hz for the SEG/EAGE Salt model. The detailed structures are imaged successfully up to the depth approximately equal to one‐third of the maximum offset length at a resolution consistent with the inverted frequencies.     

基于主成分的时间域航空电磁数据神经网络反演仿真研究(英文)

传统上,时间域航空电磁数据通过拟合迭代反演计算得到大地模型,然而,由于航空电磁数据道间的较强相关性,导致病态反演,并引起超定问题;同时电磁数据的相关性使其与模型参数的映射关系复杂,增加了反演的复杂度。采用主成分分析法将航空电磁数据变换为正交的较少数量的主成分,不仅降低了数据道间的相关性,减小了数据量,同时压制了数据的不相关噪声。本文利用人工神经网络(ANN)逼近主成分与大地模型参数间的映射关系,避免了传统反演算法中雅克比矩阵的复杂计算。层状模型的主成分神经网络与数据神经网络的反演结果对比显示,主成分神经网络反演方法网络结构简单,训练步数少,反演结果好,特别是对于含噪数据。准二维模型的主成分ANN、数据ANN以及Zhody方法的反演结果显示了主成分神经网络具有更接近真实模型的反演效果,进一步证明了主成分神经网络反演方法适合海量航空电磁探测数据反演。