Semi-supervised learning for acoustic impedance inversion

Alfarraj, Motaz; AlRegib, Ghassan

doi:10.1190/segam2019-3215902.1

Cited by 78 publications

(30 citation statements)

References 26 publications

(1 reference statement)

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…The estimated models are utilized by forward modeling to simulate a seismic gather, and the misfit between the input and simulated data is the loss objective for training the DL network. (b) The semi-supervised learning approach following [42], in which a recurrent network predicts acoustic impedance from pre-stack seismic data. The estimated impedance model is utilized by forward modeling to simulate seismic data.…”

Section: A Physics-guided Architecturesmentioning

confidence: 99%

“…A semi-supervised learning physics-guided architecture was proposed by Alfarraj and Alragib [42] for impedance inversion. They proposed to represent the DL loss objective function as a weighted sum of two loss functions:…”

Section: A Physics-guided Architecturesmentioning

confidence: 99%

“…The advantage of this architecture, depicted in figure 15(b), is the ability to learn from partially labeled data: unsupervised learning from unlabeled data is boosted by supervised learning from labeled data, and since the training set maybe imbalanced (having more unlabeled data then labeled, or vice versa) the relative contribution of the loss functions is controlled by the weight parameters α, β. While the forward modeling in [34] is free of learnable parameters, the architecture of [42] allows to learn parameters of the forward modeling, which provides additional flexibility in case of uncertainty regarding some parameters of the forward physical model.…”

Section: A Physics-guided Architecturesmentioning

confidence: 99%

See 2 more Smart Citations

Deep Learning for Seismic Inverse Problems: Toward the Acceleration of Geophysical Analysis Workflows

Adler¹,

Araya‐Polo²,

Poggio³

2021

IEEE Signal Process. Mag.

View full text Add to dashboard Cite

Seismic inversion is a fundamental tool in geophysical analysis, providing a window into the Earth. In particular, it enables the reconstruction of large scale subsurface earth models for hydrocarbon exploration, mining, earthquakes analysis, shallow hazard assessment and other geophysical tasks. This article provides a comprehensive and timely overview of emerging datadriven Deep Learning (DL) solutions to seismic inverse problems, including velocity, impedance, reflectivity model building and seismic bandwidth extension. The article reviews seismic wave propagation and signal acquisition principles using large scale sensor arrays in offshore and inland exploration. In addition, the relations between the seismic forward and inverse problems are established, and prominent model-based solutions including seismic tomography, Full-Waveform Inversion (FWI) and multiscale FWI are explained. A primer to DL principles is presented, including empirical risk minimization, stochastic gradient-based learning, Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN) and regularization. The article then presents DL-based solutions addressing full seismic inversion as well as DL-assisted FWI solutions including Encoder-Decoder architectures, CNN-based networks and RNN-based networks with Long Short Term Memory and Gated Recurrent Units. Seismic signals features extraction methods such as semblance velocity analysis and seismic spectrograms are discussed, as well as batch vs. sequential data processing architectures. The article further presents advanced solutions based on Generative Adversarial Networks , and Physics-guided DL architectures that incorporate numerical geophysical modeling into the DL training loop, enabling unsupervised and semi-supervised learning.

show abstract

Section: A Physics-guided Architecturesmentioning

confidence: 99%

Section: A Physics-guided Architecturesmentioning

confidence: 99%

Section: A Physics-guided Architecturesmentioning

confidence: 99%

See 1 more Smart Citation

Deep Learning for Seismic Inverse Problems: Toward the Acceleration of Geophysical Analysis Workflows

Adler¹,

Araya‐Polo²,

Poggio³

2021

IEEE Signal Process. Mag.

View full text Add to dashboard Cite

show abstract

“…One tricky problem is that supervised learning needs abundant labeled data for training. To alleviate this problem, Alfarraj and AlRejib propose semi-supervised learning for seismic impedance inversion to infer acoustic impedance (AI) and elastic impedance (EI) [2,3]. It only needs a few logs as labels and is less likely to get overfitted.…”

Section: Introductionmentioning

confidence: 99%

Semi-supervised Impedance Inversion by Bayesian Neural Network Based on 2-d CNN Pre-training

Ge¹,

Wang²,

Zheng³

2021

Preprint

View full text Add to dashboard Cite

Seismic impedance inversion can be performed with a semi-supervised learning algorithm, which only needs a few logs as labels and is less likely to get overfitted. However, classical semi-supervised learning algorithm usually leads to artifacts on the predicted impedance image. In this artical, we improve the semi-supervised learning from two aspects. First, by replacing 1-d convolutional neural network (CNN) layers in deep learning structure with 2-d CNN layers and 2-d maxpooling layers, the prediction accuracy is improved. Second, prediction uncertainty can also be estimated by embedding the network into a Bayesian inference framework. Local reparameterization trick is used during forward propagation of the network to reduce sampling cost. Tests with Marmousi2 model and SEAM model validate the feasibility of the proposed strategy.

show abstract

“…Based on the limited measurement data sets using limited instrumentation, a data-driven inverse problem approach could pave way to better modelling of pressure-velocity (and inverse) coupling. Data-driven methods involving supervised machine learning techniques and neural network are being used for seismic studies [8][9][10] and show promising research avenues. Based on the published literature, the current focus is on understanding the rock properties based on the acoustic sampling through impedance measurements.…”

mentioning

confidence: 99%

An Overview of Acoustic Impedance Measurement Techniques and Future Prospects

et al. 2021

View full text Add to dashboard Cite

In order to progress in the area of aeroacoustics, experimental measurements are necessary. Not only are they required for engineering applications in acoustics and noise engineering, but also they are necessary for developing models of acoustic phenomenon around us. One measurement of particular importance is acoustic impedance. Acoustic Impedance is the measure of opposition of acoustical flow due to the acoustic pressure. It indicates how much sound pressure is generated by the vibration of molecules of a particular acoustic medium at a given frequency and can be a characteristic of the medium.The aim of the present paper is to give a synthetic overview of the literature on impedance measurements and to discuss the advantage and disadvantage of each measurement technique. In this work, we investigate the three main categories of impedance measurement techniques, namely reverberation chamber techniques, impedance tube techniques, and far-field techniques. Theoretical principles for each technique are provided along with a discussion on historical development and recent advancements for each technique.

show abstract

Semi-supervised learning for acoustic impedance inversion

Cited by 78 publications

References 26 publications

Deep Learning for Seismic Inverse Problems: Toward the Acceleration of Geophysical Analysis Workflows

Deep Learning for Seismic Inverse Problems: Toward the Acceleration of Geophysical Analysis Workflows

Semi-supervised Impedance Inversion by Bayesian Neural Network Based on 2-d CNN Pre-training

An Overview of Acoustic Impedance Measurement Techniques and Future Prospects

Contact Info

Product

Resources

About