Physics-Constrained Deep Learning of Geomechanical Logs

Chen, Yuntian; Zhang, Dongxiao

doi:10.1109/tgrs.2020.2973171

Cited by 61 publications

(14 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To solve this problem, Zhang et al (2018) proposed a cascaded LSTM (C‐LSTM) based on the LSTM. Some researchers have attempted to introduce formation information into LSTM to generate geomechanical logs (Chen & Zhang, 2020). Although LSTM‐based models can generate well logs, they have poor prediction accuracy on wells with rare patterns.…”

Section: Introductionmentioning

confidence: 99%

Well Log Generation via Ensemble Long Short‐Term Memory (EnLSTM) Network

Chen

Zhang

2020

Geophysical Research Letters

Self Cite

View full text Add to dashboard Cite

In this study, we propose an ensemble long short-term memory (EnLSTM) network, which can be trained on a small data set and process sequential data. The EnLSTM is built by combining the ensemble neural network and the cascaded LSTM network to leverage their complementary strengths. Two perturbation methods are applied to resolve the issues of overconvergence and disturbance compensation. The EnLSTM is compared with commonly used models on a published data set and proven to be the state-of-the-art model in generating well logs. In the case study, 12 well logs that cannot be measured while drilling are generated based on the logs available in the drilling process. The EnLSTM is capable of reducing cost and saving time in practice. Plain Language Summary A novel neural network, called EnLSTM, is proposed by combining the ensemble neural network, which has good performance on small-data problems, and the cascaded long short-term memory network, which is effective at processing sequential data. The EnLSTM's capability of processing sequential data based on a small data set is especially suitable for generating synthetic well logs. In addition, two perturbation methods are used to ensure that the EnLSTM can be fully trained in practice. In the experiments, the EnLSTM achieved the current best results on a published well log data set, and its application value is verified in a case study.

show abstract

Section: Introductionmentioning

confidence: 99%

Well Log Generation via Ensemble Long Short‐Term Memory (EnLSTM) Network

Chen

Zhang

2020

Geophysical Research Letters

Self Cite

View full text Add to dashboard Cite

show abstract

“…Moreover, the network architectures can be designed considering the focused problems. This means we can mimic the underlying physical phenomenon using a DNN [ 31 ] or use it describe some ambiguous physical laws [ 16 , 26 , 29 ], and even combine some physics-guided computations in a NN [ 27 ]. Current PIDL studies are summarized in Table 1 .…”

Section: Introductionmentioning

confidence: 99%

Multi-End Physics-Informed Deep Learning for Seismic Response Estimation

Sun

Yang

et al. 2022

Sensors

View full text Add to dashboard Cite

As a structural health monitoring (SHM) system can hardly measure all the needed responses, estimating the target response from the measured responses has become an important task. Deep neural networks (NNs) have a strong nonlinear mapping ability, and they are widely used in response reconstruction works. The mapping relation among different responses is learned by a NN given a large training set. In some cases, however, especially for rare events such as earthquakes, it is difficult to obtain a large training dataset. This paper used a convolution NN to reconstruct structure response under rare events with small datasets, and the main innovations include two aspects. Firstly, we proposed a multi-end autoencoder architecture with skip connections, which compresses the parameter space, to estimate the unmeasured responses. It extracts the shared patterns in the encoder and reconstructs different types of target responses in varied branches of the decoder. Secondly, the physics-based loss function, derived from the dynamic equilibrium equation, was adopted to guide the training direction and suppress the overfitting effect. The proposed NN takes the acceleration at limited positions as input. The output is the displacement, velocity, and acceleration responses at all positions. Two numerical studies validated that the proposed framework applies to both linear and nonlinear systems. The physics-informed NN had a higher performance than the ordinary NN with small datasets, especially when the training data contained noise.

show abstract

“…However, they can generally only deal with very specific types of constraints, complicated further by the recurrent or feedback nature of the networks. [30][31][32] In this work we provide a generalizable, statistical physics based approach to add a variety of constraints to LSTMs. To achieve this, we use ideas of path sampling combined with LSTM, facilitated through the principle of Maximum Caliber.…”

Section: Introductionmentioning

confidence: 99%

Path sampling of recurrent neural networks by incorporating known physics

Tsai¹,

Fields²,

Tiwary³

2022

Preprint

View full text Add to dashboard Cite

Recurrent neural networks have seen widespread use in modeling dynamical systems in varied domains such as weather prediction, text prediction and several others. Often one wishes to supplement the experimentally observed dynamics with prior knowledge or intuition about the system. While the recurrent nature of these networks allows them to model arbitrarily long memories in the time series used in training, it makes it harder to impose prior knowledge or intuition through generic constraints. In this work, we present a path sampling approach based on principle of Maximum Caliber that allows us to include generic thermodynamic or kinetic constraints into recurrent neural networks. We show the method here for a widely used type of recurrent neural network known as long short-term memory network in the context of supplementing time series collecting from all-atom molecular dynamics. We demonstrate the power of the formalism for different applications. Our method can be easily generalized to other generative artificial intelligence models and to generic time series in different areas of physical and social sciences, where one wishes to supplement limited data with intuition or theory based corrections.

show abstract

Physics-Constrained Deep Learning of Geomechanical Logs

Cited by 61 publications

References 34 publications

Well Log Generation via Ensemble Long Short‐Term Memory (EnLSTM) Network

Well Log Generation via Ensemble Long Short‐Term Memory (EnLSTM) Network

Multi-End Physics-Informed Deep Learning for Seismic Response Estimation

Path sampling of recurrent neural networks by incorporating known physics

Contact Info

Product

Resources

About