Physics-based Simulation Research Articles

Sensitivity analysis of a dual-continuum model system for integrated CO2 sequestration and geothermal extraction in a fractured reservoir

Depleted hydrocarbon reservoirs are considered as the most feasible option for CO2 geological sequestration and utilization. Most of the hydrocarbon reservoirs are naturally fractured. Simulation of fluid flow and heat transfer in these fractured formations remains a significant challenge in reservoir engineering. In this study, a dual-continuum model is developed to simulate integrated CO2 sequestration and CO2-circulated geothermal extraction in a fractured reservoir block in North Oman. The high-dimensional sensitivity of key parameters controlling CO2-brine flow and heat transfer in this matrix-fracture system is quantitatively evaluated by an efficient surrogate modeling approach. The surrogate models are constructed and validated based on a suite of physics-based model simulations. It is found that fracture permeability dominates the CO2 injectivity, storage, circulation and associated geothermal extraction. Response surface analysis shows that the flow area density between matrix-fracture and matrix block length controls the flux interaction between matrix and fracture formations. In contrast, the fracture aperture shows negligible influence in the dual-continuum modeling system. Particularly, sensitivity varying with locations on the response surface is analyzed for defined performance indicators.

Simplified site response analysis for regional seismic risk assessments

Seismic risk assessment on a regional scale requires an estimation of ground motion intensity and its spatial variation over large geographical areas. This task is particularly challenging in the presence of very soft soil deposits where significant amplification occurs at specific frequencies that are often referred to as local resonances. It is often not feasible to conduct a nonlinear site response analysis at thousands of sites of interest across a region such as an urban area due to the computational effort involved and the required detailed information on soil profiles at each site. Thus, in regional seismic risk analyses the hazard is typically represented by a field of response spectral ordinates estimated by ground motion models. This study proposes a simplified site response analysis procedure to improve the characterization of spectral ordinates to be used in regional seismic risk assessments. In the proposed procedure, reduced-order models are used to conduct site response analysis using very few parameters to transform response spectra at rock outcrop into response spectra at the surface of each of the sites. A particular emphasis is given to soft-soil sites characterized by low shear-wave velocities overlaid on rock or firm soils. A non-uniform continuous shear beam model with a parabolic variation of shear modulus along the depth and modal damping is used in combination with Random Vibration Theory. It is shown that the proposed approach provides better estimates than those of ground motion models or physics-based ground motion simulation models which often cannot capture the local resonances. Site-specific validation of the shear beam model and overall simplified site response analysis is performed at four soft-soil sites in San Francisco, California. Despite its simplicity, it is shown that the proposed procedure adequately captures the amplitudes and spectral shapes of the motions. It is also shown that in conjunction with ground motion models or physics-based models, the proposed simplified site response analysis procedure can be used to efficiently simulate the seismic hazard on a regional scale through an example of a regional seismic hazard analysis of 160 softer soil sites in Downtown San Francisco during the Loma Prieta earthquake.

Estimating ground motion intensities using simulation-based estimates of local crustal seismic response

Abstract. It is estimated that 2 billion people will move to cities in the next 30 years, many of which possess high seismic risk, underscoring the importance of reliable hazard assessments. Current ground motion models for these assessments typically rely on an extensive catalogue of events to derive empirical ground motion prediction equations (GMPEs), which are often unavailable in developing countries. Considering the challenge, we choose an alternative method utilizing physics-based (PB) ground motion simulations and develop a simplified decomposition of ground motion estimation by considering regional attenuation (Δ) and local site amplification (A), thereby exploring how much of the observed variability can be explained solely by wave propagation effects. We deterministically evaluate these parameters in a virtual city named Tomorrowville, located in a 3D-layered crustal velocity model containing sedimentary basins, using randomly oriented extended sources. Using these physics-based empirical parameters (Δ and A), we evaluate the intensities, particularly peak ground acceleration (PGA), of hypothetical future earthquakes. The results suggest that the estimation of PGA using the deterministic Δ−A decomposition exhibits a robust spatial correlation with the PGA obtained from simulations within Tomorrowville. This method exposes an order-of-magnitude spatial variability in PGA within Tomorrowville, primarily associated with the near-surface geology and largely independent of the seismic source. In conclusion, advances in PB simulations and improved crustal structure determination offer the potential to overcome the limitations of earthquake data availability to some extent, enabling prompt evaluation of ground motion intensities.

Does the direct effect of friction increase continuously with absolute temperature?

Constitutive models of fault friction form the basis of physics-based simulations of seismic activity. A generally accepted framework for the slip-rate and state dependence of friction involves a thermally activated process, whereby the probability of slip along microasperities adheres to an Arrhenius law. This model, which has become widely adopted among experimentalists and theoreticians, predicts a continuous increase of the direct effect with absolute temperature, but is it observed experimentally? Leveraging comprehensive laboratory data across diverse hydrothermal, barometric, and lithological conditions, we demonstrate that, contrary to the classical view, the direct effect for a given deformation mechanism remains largely temperature-independent. Instead, the incremental shifts in the direct effect often coincide with the brittle to semi-brittle transition, across which distinct deformation mechanisms operate. These considerations challenge the validity of the classical model. Realistic constitutive laws for rock failure within the lithosphere must incorporate the contributions of multiple deformation mechanisms within active fault zones.

Analysis of postdisaster economy using high-resolution disaster and economy simulations.

We present an integrated framework that utilizes high-resolution seamless simulations of disasters and national economies for estimating the economic impacts of disasters. The framework consists of three components: a physics-based simulator to simulate the disaster and estimate the response of the infrastructure; a tool that estimates the losses suffered by the infrastructure based on its response; and an agent-based economic model (ABEM) that simulates the national economy considering the infrastructure damage and postdisaster decisions of the economic entities. The ABEM used in the framework has been implemented in a high-performance computing environment to simulate large economies at 1:1 scale. Furthermore, it has been calibrated to the Japanese economy using publicly available macroeconomic data and validated to the Japanese economy under the business-as-usual scenario. We demonstrate the integrated framework by simulating a potential Nankai-trough earthquake disaster and estimating its impacts on the Japanese economy. The seismic response of 1.8 million buildings of the Osaka-bay area has been estimated using a large-scale earthquake disaster simulator and corresponding repair costs are estimated using the Performance Assessment Calculation Tool. As per our estimates, repair costs amount to approximately 15 trillion Yen. Considering the investments made by impacted households and firms toward recovery, the postdisaster economy is simulated using the ABEM for 5 years under two recovery scenarios. Industrial production is expected to recover in three quarters whereas 10-13 quarters will be required to finish all the repairwork.

Reliability modeling for perception systems in autonomous vehicles: A recursive event-triggering point process approach

Ensuring the reliability of sensor-fusion-based perception systems is crucial for the safe deployment of autonomous vehicles. Such systems function through a sequence of interconnected stages, where errors in upstream stages may propagate to downstream stages and trigger additional errors. The cross-stage error propagation conceptually exists and makes errors in different stages, not independent, posing model challenges, estimation challenges, and data challenges for reliability modeling. The existing methods cannot be applied to address all these challenges. Thus, this paper presents a recursive event-triggering point process to explicitly consider the error propagation based on the simulated data. The data are simulated from a proposed error injection framework, which can generate various errors from a sequence of interconnected stages in a perception system. The latent and probabilistic error propagation information is incorporated into a modified expectation–maximization (EM) algorithm for parameter estimation. The numerical and physics-based simulation case studies demonstrate the prediction accuracy and interpretability of the proposed modeling methodology.

Contribution of machine learning and physics-based sound simulations for the optimization of brass instruments

Sound simulations by physical model are interesting to transcribe the physics underlying the functioning of a musical instrument. These simulations make it possible to create virtual sounds with a mode of operation representative of the interaction musician/instrument (driving a sound by the causes that create it). The work consists in studying the contribution of machine learning (ML) methods in the optimization of a musical instrument. The application concerns the brass instruments (trumpets). From a training set of virtual instruments, ML methods are trained to model the intonation (represented by the equivalent fundamental pitch) and the ease of emission (represented by the threshold pressure obtained by linear stability analysis) of sounds, according to the modal parameters of the input impedance. With the ML models, an optimization with genetic algorithms is next carried out to improve the intonation and the ease of emission. The last stage of the approach consists in the optimization of the bore of the instrument (optimization of the shape of the leadpipe). The results show that the approach gives promising results: optimal instruments are actually efficient when re-simulated with the physical model. The manufacturing of an optimal instrument is in progress to confirm the validity of the approach

Alchemical Transformations and Beyond: Recent Advances and Real-World Applications of Free Energy Calculations in Drug Discovery.

Computational methods constitute efficient strategies for screening and optimizing potential drug molecules. A critical factor in this process is the binding affinity between candidate molecules and targets, quantified as binding free energy. Among various estimation methods, alchemical transformation methods stand out for their theoretical rigor. Despite challenges in force field accuracy and sampling efficiency, advancements in algorithms, software, and hardware have increased the application of free energy perturbation (FEP) calculations in the pharmaceutical industry. Here, we review the practical applications of FEP in drug discovery projects since 2018, covering both ligand-centric and residue-centric transformations. We show that relative binding free energy calculations have steadily achieved chemical accuracy in real-world applications. In addition, we discuss alternative physics-based simulation methods and the incorporation of deep learning into free energy calculations.

SAM-RL: Sensing-aware model-based reinforcement learning via differentiable physics-based simulation and rendering

SAM-RL: Sensing-aware model-based reinforcement learning via differentiable physics-based simulation and rendering

Deforming Patient-Specific Models of Vascular Anatomies to Represent Stent Implantation via Extended Position Based Dynamics.

Angioplasty with stent placement is a widely used treatment strategy for patients with stenotic blood vessels. However, it is often challenging to predict the outcomes of this procedure for individual patients. Image-based computational fluid dynamics (CFD) is a powerful technique for making these predictions. To perform CFD analysis of a stented vessel, a virtual model of the vessel must first be created. This model is typically made by manipulating two-dimensional contours of the vessel in its pre-stent state to reflect its post-stent shape. However, improper contour-editing can cause invalid geometric artifacts in the resulting mesh that then distort the subsequent CFD predictions. To address this limitation, we have developed a novel shape-editing method that deforms surface meshes of stenosed vessels to create stented models. Our method uses physics-based simulations via Extended Position Based Dynamics to guide these deformations. We embed an inflating stent inside a vessel and apply collision-generated forces to deform the vessel and expand its cross-section. We demonstrate that this technique is feasible and applicable for a wide range of vascular anatomies, while yielding clinically compatible results. We also illustrate the ability to parametrically vary the stented shape and create models allowing CFD analyses. Our stenting method will help clinicians predict the hemodynamic results of stenting interventions and adapt treatments to achieve target outcomes for patients. It will also enable generation of synthetic data for data-intensive applications, such as machine learning, to support cardiovascular research endeavors.

Evaluation of directionality in physics-based ground motion simulations of strike-slip earthquakes

Recent advances in engineering seismology and computing power have led to the development of realistic physics-based ground motion simulations. The use of these simulations for engineering applications requires that the ground motions exhibit characteristics consistent with those of recorded ground motions. While several studies have evaluated the intensity, frequency content, duration, and other characteristics of ground motion simulations for their possible use in engineering applications, few have focused on directionality. This article evaluates directionality in the response of single-degree-of-freedom oscillators when subjected to CyberShake Study 15.12 simulated ground motions from strike-slip earthquakes, by carefully comparing it to the directionality in recorded ground motions having the same style of faulting. Physics-based ground motion simulations at 334 stations from 5 different rupture variations in two large-magnitude earthquake scenarios on the Elsinore fault are evaluated. The orientation of maximum oscillator response and its spatial distribution are studied for each rupture. The orientation of maximum oscillator response is found to occur systematically close to the epicentral transverse orientation at all rupture distances, consistent with recent findings for ground motions recorded during strike-slip earthquakes. In addition, the orientation-specific spectral accelerations, when computed as a function of angular distance from the epicentral transverse orientation, are found to exhibit a variation consistent with the overall trend observed in records from the next-generation attenuation (NGA)-West2 database. However, the level of polarization at short periods in the simulated hybrid broadband waveforms used in this study is larger than that in recorded ground motions.

An end-to-end framework for fire following earthquake simulation at regional scale: A case study on the 2024 Japan Noto Peninsula earthquake

Fire following earthquakes remains a significant and threatening hazard to communities in urban regions. Addressing this critical issue requires an effective simulation method that is suitable for widespread adoption. This paper presents a systematic end-to-end framework for simulating fire following earthquakes at a regional scale by integrating models and methods available in existing literature. Relevant theories and models are summarized and incorporated into the calculation workflow. The framework encompasses three key features: (1) GIS-based regional building data management, (2) physics-based fire simulation, and (3) high-fidelity visualization. The implementation of the framework is presented in a sequential and detailed manner, with a focus on its practicality and replicability. Each component of the workflow is designed to be adaptable, allowing for easy customization and improvement to accommodate unique site characteristics or the integration of new models through the introduction of sub-modules. To demonstrate the practicality of the proposed framework, a case study based on the recent 2024 Noto Peninsula earthquake in Japan is presented. The simulation results correspond well with the observed actual damage at the fire following earthquake disaster site, with an accuracy of 87.8 %, which validates the framework's accuracy and reliability in simulating fire following earthquakes.

Buffer Screening of Protein Formulations Using a Coarse-Grained Protocol Based on Medicinal Chemistry Interactions.

In drug and vaccine development, the designed protein formulation should be highly stable against the temperature, pH, buffer, excipients, and other environmental settings. Similarly, in a sensing unit, one needs to know how strongly two biomolecules bind to guide the design of the biorecognition unit accordingly. Typically, the community performs a series of experiments to thoroughly examine the parameter space, the so-called design-of-experiment (DoE) method, to identify the optimal formulation conditions. Unfortunately, extensive physical testing entails high costs, repeatability issues, and a lack of in-depth knowledge of the underlying mechanisms that affect the final outcome. To address these challenges, we developed a physics-based simulation protocol for buffer screening of protein formulations. We are introducing a coarse-grained molecular simulation protocol that consists of six different interactions. The so-called medicinal chemistry interactions (electrostatics, hydrophobicity, hydrogen bonding propensity, disulfide bonding, and water-water) are based on the physical nature of the protein's amino acid and the partitioning/polarity of any other chemical constituent. The protocol is applied in immunoglobulin-based monoclonal antibodies. We have analyzed the protein behavior as a function of acidity (pH) to discover the isoelectric point by solving the Poisson-Boltzmann equation in a mesoscale grid. To identify the conditions under which the protein oligomerizes in a given buffer, pH, temperature, and ionic strength, we are performing dissipative particle dynamics (DPD) simulations. The protocol allows researchers to reach the high time/space scales required to study protein formulations in their full complexity. Combined with the disruptive protein folding artificial intelligence (AI) algorithms that have been recently developed, the protocol creates a powerful digital framework for cultivating advanced pharmaceutical and biological applications.

A data-driven dual-optimization hybrid machine learning model for predicting carbon dioxide trapping efficiency in saline aquifers: Application in carbon capture and storage

Geological carbon sequestration (GCS) is a promising method to reduce global carbon emissions. To better understand the CO2 trapping mechanism, the CO2 sequestration process was simulated through a physics-based reservoir simulator, and the residual trapping (RTI) and solubility trapping (STI) efficiencies were analyzed. Through the experiments, a database containing 4543 data points with reservoir physical parameters and rock properties was established. In this study, a kernel extreme learning machine (KELM) model based on variational mode decomposition (VMD) and dung beetle optimization optimizer (DBO) was proposed to perform the prediction task. Through the VMD, the original data sequences were decomposed and combined, and the obtained subsequences were executed modeling prediction and reconstruction respectively. The reliability of the prediction model was confirmed to be greater than 97.71% through domain analysis. In addition, it was applied to a tight gas reservoir in China in to confirm its validity in field applications. Comparative results indicated a high agreement between the prediction results and the simulation results, suggesting that the proposed method can complement numerical simulations and help users understand the application of machine learning in carbon capture and storage (CCS).

Geometric physically-based simulation of microfinishing: tool design impact on workpiece shape

Geometric physically-based simulation of microfinishing: tool design impact on workpiece shape

Uncertainty propagation from crustal geologies to rock-site ground motion with a Fourier Neural Operator

Seismic hazard analyses in the area of a nuclear installation must account for a large number of uncertainties, including limited geological knowledge. Combining the accuracy of physics-based simulations with the expressivity of deep neural networks can help to quantify the influence of crustal geological uncertainties on surface ground motion. This work uses a Factorised Fourier Neural Operator (F-FNO) to learn the relationship between 3D crustal heterogeneous geologies and time-dependent wavefields at the Earth’s surface up to a 5 Hz maximal frequency. The F-FNO is pretrained on a generic database and then, fine-tuned with only 250 samples targeted to the Le Teil region (South-Eastern France). The F-FNO correctly predicts the wave arrival times and the main wavefronts. As quantified by Goodness-Of-Fit (GOF) criteria, 83% of predictions have excellent phase GOF and 75% have very good envelope GOF. The Peak Ground Velocity and Pseudo-Spectral Acceleration are also accurate, especially for geologies with low-to-moderate heterogeneities. Thanks to the F-FNO speed, ground motion distributions can be easily computed and provide safety margins compared to 1D simulations. These results show that the F-FNO is an efficient surrogate model to quantify the range of ground motion a nuclear installation could face in the presence of geological uncertainties.

A novel biomechanical model of the mouse forelimb predicts muscle activity in optimal control simulations of reaching movements.

Mice are key model organisms in genetics, neuroscience and motor systems physiology. Fine motor control tasks performed by mice have become widely used in assaying neural and biophysical motor system mechanisms, including lever or joystick manipulation, and reach-to-grasp tasks (Becker et al., 2019; Bollu et al., 2019; Conner at al., 2021). Although fine motor tasks provide useful insights into behaviors which require complex multi-joint motor control, there is no previously developed physiological biomechanical model of the adult mouse forelimb available for estimating kinematics (including joint angles, joint velocities, fiber lengths and fiber velocities) nor muscle activity or kinetics (including forces and moments) during these behaviors. Here we have developed a musculoskeletal model based on high-resolution imaging and reconstruction of the mouse forelimb that includes muscles spanning the neck, trunk, shoulder, and limbs using anatomical data. Physics-based optimal control simulations of the forelimb model were used to estimate in vivo muscle activity present when constrained to the tracked kinematics during mouse reaching movements. The activity of a subset of muscles was recorded via electromyography and used as the ground truth to assess the accuracy of the muscle patterning in simulation. We found that the synthesized muscle patterning in the forelimb model had a strong resemblance to empirical muscle patterning, suggesting that our model has utility in providing a realistic set of estimated muscle excitations over time when provided with a kinematic template. The strength of the resemblance between empirical muscle activity and optimal control predictions increases as mice performance improves throughout learning of the reaching task. Our computational tools are available as open-source in the OpenSim physics and modeling platform (Seth et al., 2018). Our model can enhance research into limb control across broad research topics and can inform analyses of motor learning, muscle synergies, neural patterning, and behavioral research that would otherwise be inaccessible.

Broadband Ground-Motion Simulations with Machine-Learning-Based High-Frequency Waves from Fourier Neural Operators

ABSTRACT Seismic hazards analysis relies on accurate estimation of expected ground motions for potential future earthquakes. However, obtaining realistic and robust ground-motion estimates for specific combinations of earthquake magnitudes, source-to-site distances, and site conditions is still challenging due to the limited empirical data. Seismic hazard analysis also benefits from the simulation of ground-motion time histories, whereby physics-based simulations provide reliable time histories but are restricted to a lower frequency for computational reasons and missing information on small-scale earthquake-source and Earth-structure properties that govern high-frequency (HF) seismic waves. In this study, we use densely recorded acceleration broadband (BB) waveforms to develop a machine-learning (ML) model for estimating HF ground-motion time histories from their low-frequency (LF) counterparts based on Fourier Neural Operators (FNOs) and Generative Adversarial Networks (GANs). Our approach involves two separate FNO models to estimate the time and frequency properties of ground motions. In the time domain, we establish a relationship between normalized low-pass filtered and BB waveforms, whereas in the frequency domain, the HF spectrum is trained based on the LF spectrum. These are then combined to generate BB ground motions. We also consider seismological and site-specific factors during the training process to enhance the accuracy of the predictions. We train and validate our models using ground-motion data recorded over a 20 yr period at 18 stations in the Ibaraki province, Japan, considering earthquakes in the magnitude range M 4–7. Based on goodness-of-fit measures, we demonstrate that our simulated time series closely matches recorded observations. To address the ground-motion variability, we employ a conditioned GAN approach. Finally, we compare our results with several alternative approaches for ground-motion simulation (stochastic, hybrid, and ML-based) to highlight the advantages and improvements of our method.

Synthetic ground motions in heterogeneous geologies from various sources: the HEMEWS-3D database

Abstract. The ever-improving performances of physics-based simulations and the rapid developments of deep learning are offering new perspectives to study earthquake-induced ground motion. Due to the large amount of data required to train deep neural networks, applications have so far been limited to recorded data or two-dimensional (2D) simulations. To bridge the gap between deep learning and high-fidelity numerical simulations, this work introduces a new database of physics-based earthquake simulations. The HEterogeneous Materials and Elastic Waves with Source variability in 3D (HEMEWS-3D) database comprises 30 000 simulations of elastic wave propagation in 3D geological domains. Each domain is parametrized by a different geological model built from a random arrangement of layers augmented by random fields that represent heterogeneities. Elastic waves originate from a randomly located pointwise source parametrized by a random moment tensor. For each simulation, ground motion is synthesized at the surface by a grid of virtual sensors. The high frequency of waveforms (fmax⁡=5 Hz) allows for extensive analyses of surface ground motion. Existing and foreseen applications range from statistical analyses of the ground motion variability and machine learning methods on geological models to deep-learning-based predictions of ground motion that depend on 3D heterogeneous geologies and source properties. Data are available at https://doi.org/10.57745/LAI6YU (Lehmann, 2023).

Optimal Molecular Design: Generative Active Learning Combining REINVENT with Precise Binding Free Energy Ranking Simulations.

Active learning (AL) is a specific instance of sequential experimental design and uses machine learning to intelligently choose the next data point or batch of molecular structures to be evaluated. In this sense, it closely mimics the iterative design-make-test-analysis cycle of laboratory experiments to find optimized compounds for a given design task. Here, we describe an AL protocol which combines generative molecular AI, using REINVENT, and physics-based absolute binding free energy molecular dynamics simulation, using ESMACS, to discover new ligands for two different target proteins, 3CLpro and TNKS2. We have deployed our generative active learning (GAL) protocol on Frontier, the world's only exa-scale machine. We show that the protocol can find higher-scoring molecules compared to the baseline, a surrogate ML docking model for 3CLpro and compounds with experimentally determined binding affinities for TNKS2. The ligands found are also chemically diverse and occupy a different chemical space than the baseline. We vary the batch sizes that are put forward for free energy assessment in each GAL cycle to assess the impact on their efficiency on the GAL protocol and recommend their optimal values in different scenarios. Overall, we demonstrate a powerful capability of the combination of physics-based and AI methods which yields effective chemical space sampling at an unprecedented scale and is of immediate and direct relevance to modern, data-driven drug discovery.