Earthquake control: impossible or adjacent possible?

Check out the talk by Ioannis Stefanou and Filippo Masi at the Data-Centric Engineering Institute of the University of Sydney

Earthquakes nucleate when large amounts of elastic energy are suddenly released due to abrupt sliding over seismic faults. Besides physical causes, this energy release can be also triggered by injecting large amounts of fluids in the earth’s crust. Indeed, recent events show that injections can reactivate existing seismogenic faults and induce/trigger important earthquakes.

However, one can see the problem of fluid injections from another perspective. The dependence of fault friction on fluid pressure can be used as an input for stabilizing it. New results based on the mathematical theory of control show that it is indeed possible to stabilize and restrict chaos in this kind of non-linear unstable frictional systems and assure slow frictional dissipation toward desirable global asymptotic equilibria of lower potential energy.

Our mathematical derivations show that earthquake robust control is possible, provided that fault friction is bounded. Friction is a complex phenomenon involving several spatio-temporal scales and complex dynamics. In spite of complexity, the finiteness of energy of any physical process influencing friction as well as existing laboratory and in-situ experiments, support the boundedness of friction. However, the quantification of these bounds is challenging as experiments are difficult to perform deep down in the earth. As a result, we have to rely also on numerical simulations and digital twins of seismic faults, which until know are computationally very demanding.

To this end, we develop a novel deep learning approach for modeling complex materials, by enforcing the laws of thermodynamics. The method is accurate, fast and scalable, and will allow, in the near future, to derive reliable quantitative estimations of the bounds of fault friction under the effect of multiphyics couplings and complex grain/particle dynamics. In parallel, it will enable to optimize our earthquake control strategies by performing large scale simulations of real fault systems in a virtual, data- and physics-driven environment, that fits on a laptop.

Presentation slides are available here

Does viscosity regularize mesh dependency and strain localization on a mathematical plane?

The answer is no! In https://authors.elsevier.com/a/1dux9AQEIviHm (https://doi.org/10.1016/j.cma.2021.114185) we prove mathematically and show numerically that viscoplasticity of Perzyna or consistency type in combination with inertia cannot prevent (a) strain localization on a mathematical plane and (b) mesh dependency in numerical calculations.

We show also the existence of emergent traveling plastic waves and we investigate some of their properties. These waves can be related to the Portevin–Le Chatelier effect and are also observed in numerical simulations of fault gouges under large shear. To be continued!

Complex 3D plot of dispersion functions derived after bifurcation analysis.

Low cost materials like paper towel can give useful insights about induced earthquakes – our new paper in Geophysical Research Letters

#ERCStG, #earthquakes, #geothermalenergy, @EuropeanResearchCouncil, @EcoleCentraleNantes

Videos of our experiments that show sequences of aseismic and seismic events due to simulated injections: Movie S1, Movie S2, Movie S3, Movie S4

Abstract: Earthquakes nucleate when large amounts of elastic energy, stored in the earth’s crust, are suddenly released due to abrupt sliding over a fault. Fluid injections can reactivate existing seismogenic faults and induce/trigger earthquakes by increasing fluid pressure. Here we develop an analogous experimental system of simultaneously loaded and wetted absorbent porous paper to quantify theoretically the process of wetting-induced earthquakes. This strategy allows us to gradually release the stored energy by provoking low intensity tremors. We identify the key parameters that control the outcome of the applied injection strategy, which include the initial stress state, fault segmentation, and segment-activation rate. Subsequent injections, initiated at high stress levels, can drive the system faster towards its instability point, nucleating a large earthquake. Starting at low stress levels, however, they can reduce the magnitude of the natural event by at least one unit.

For more: (PDF) Absorbent porous paper reveals how earthquakes could be mitigated (researchgate.net)

Thermodynamics-based ANN (TANN) for constitutive modeling – our new paper in JMPS

Thermodynamics-based Artificial Neural Networks for constitutive modeling – ScienceDirect

Abstract: Machine Learning methods and, in particular, Artificial Neural Networks (ANNs) have demonstrated promising capabilities in material constitutive modeling. One of the main drawbacks of such approaches is the lack of a rigorous frame based on the laws of physics. This may render physically inconsistent the predictions of a trained network, which can be even dangerous for real applications.

Here we propose a new class of data-driven, physics-based, neural networks for constitutive modeling of strain rate independent processes at the material point level, which we define as Thermodynamics-based Artificial Neural Networks (TANNs). The two basic principles of thermodynamics are encoded in the network’s architecture by taking advantage of automatic differentiation to compute the numerical derivatives of a network with respect to its inputs. In this way, derivatives of the free-energy, the dissipation rate and their relation with the stress and internal state variables are hardwired in the architecture of TANNs. Consequently, our approach does not have to identify the underlying pattern of thermodynamic laws during training, reducing the need of large data-sets. Moreover the training is more efficient and robust, and the predictions more accurate. Finally and more important, the predictions remain thermodynamically consistent, even for unseen data. Based on these features, TANNs are a starting point for data-driven, physics-based constitutive modeling with neural networks.

We demonstrate the wide applicability of TANNs for modeling elasto-plastic materials, using both hyper- and hypo-plasticity models. Strain hardening and softening are also considered for the hyper-plastic scenario. Detailed comparisons show that the predictions of TANNs outperform those of standard ANNs. Finally, we demonstrate that the implementation of the laws of thermodynamics confers to TANNs high robustness in the presence of noise in the training data, compared to standard approaches.

TANNs’ architecture is general, enabling applications to materials with different or more complex behavior, without any modification.

breakage_animation

Grain breakage, polydispersity and shear band thickness evolution in seismogenic faults using Cosserat continuum and breakage theory – our new JMPS paper @IEinav @NCollinsCraft

We study a granular layer sheared under constant volume which simulates the fast undrained deformation observed in seismogenic faults. We examine the effect of grain size polydispersity and grain breakage on the thickness of shear bands. By implementing the model using the finite element method we provide an explanation for the geological formation of double cataclastic shear bands.

https://www.sciencedirect.com/science/article/pii/S0022509620302106