High-fidelity Model Simulations Research Articles

Reducing trainee mistakes. Better performance with changing to a high-fidelity simulation system?

Postpartum hemorrhage is a significant cause of both maternal morbidity and mortality worldwide and is increasing in incidence. This study aimed to assess improvement and identify shortcomings in trainee performance in different simulation systems in the management of postpartum hemorrhage. To perform a pilot study evaluating and comparing high- and low-fidelity simulation models, assessing improvement in repeated performance with high-fidelity mode and identifying mistakes made assessed using Objective Structured Assessment of Technical Skills and thereby exploring what aspects of emergency management of postpartum hemorrhage should be prioritized in teaching settings and assessing what simulation setup is most effective in achieving competence. This was a prospective randomized, single-blinded, single-institution trial in a population of 17 junior obstetrical trainees at the Charité University Hospital Obstetric Simulation Center in Berlin. Trainees were randomized into 2 groups, with either initial low-fidelity simulation or high-fidelity simulation, followed by repeated assessment of performance, using the high-fidelity model simulation system. Individual simulation sessions were video-recorded and transcribed, and the timing of interventions was documented. Strandardized Objective Structured Assessment of Technical Skills forms were used as a checklist for performance. There was a statistically significant general improvement in performance (P=.02; 24.7-27.2 of 31.0 points; average of 8.7%) in the second cycle of simulation assessment and a statistically significant training effect (P=.043; 24.4-28.4 of 31.0 points; average of 12.9%) in the group that underwent repeat simulation assessment from the initial low-fidelity system to the high-fidelity system compared with the group using the same high-fidelity setup (P=.276; 25.0-25.8 of 31.0; average of 2.4%). There was an improvement in the performance when trainees underwent a repeated cycle of simulation assessment changing from a low-fidelity system to a high-fidelity system. Simulation assessment can identify mistakes and learning gaps that are important for obstetrical trainees. This study found that trainees make the same mistakes, regardless of which simulation model was initially used.

Collisions between Drones and Rotorcraft: Modeling of the Crash Response of Battery Packs

Drones operating in an urban environment pose a potential collision threat to rotorcraft. In this paper, the battery pack of the DJI MAVIC 2 ZOOM is analyzed since the battery is considered to be the greatest threat due to its high weight and stiffness. Following a pyramid-type building block approach, a high-fidelity model simulation was developed for LSDYNA based on a wide range of experiments, ranging from quasi-static material tests to quasi-static component tests up to high-velocity impact experiments. The high-fidelity model allows the prediction of damage in potential collision scenarios between a high-speed rotorcraft and the battery pack of the drone. For the particular impact configuration analyzed within this paper, the drone battery does not cause catastrophic failure of the windshield of the rotorcraft.

Deep-learning-generalized data-space inversion and uncertainty quantification framework for accelerating geological CO2 plume migration monitoring

Efficient coupling of high-fidelity simulation models and history-monitored data to predict future migration behavior of CO2 plume is crucial for leakage risk management and assessment during the process of geological CO2 sequestration (GCS). However, the conventional model-based inverse modeling and forecasting procedure generally presents intensive computation-cost due to numerous high-fidelity model simulations and thus hinders its practical applications. We present a novel data-driven direct forecast and uncertainty quantification framework that hybridizes deep convolutional autoencoder and deep residual neural network to represent spatial distribution of subsurface fluid by latent space features and achieve predictions of future fluid dynamic behavior given history-monitored data, respectively. The linear-type data-space inversion approaches based on Bayesian theorem are generalized to non-linear model applications through deep-learning, i.e., referred to as DSI-DL. Compared to the conventional learning-based direct forecast approach (LDFA) that numerous training samples are required to guarantee predictability of the supervised machine-learning models, our proposed DSI-DL employs a data-augmentation strategy to automatically generate sufficient training samples from a relatively small ensemble of high-fidelity model simulations. The performance DSI-DL method is demonstrated on a synthetic 2D models and a practical 3D model for deep saline aquifer GCS applications. The deep-learning surrogate directly predicts the latent features of future spatial CO2 plumes given history-monitored CO2 plumes. Comparisons between LDFA and DSI-DL confirm that the proposed DSI-DL method obtains more robust and better results in predicting future CO2 plume migration, especially when solely a few number of large-scale high-fidelity model simulations is available at hand and large observation errors exist. The proposed deep-learning-based direct forecast approach contributes to an efficient computation framework to manage uncertainty of the long term process of GCS.

Rapid Estimation of Climate-Air Quality Interactions in Integrated Assessment Using a Response Surface Model.

Air quality and climate change are substantial and linked sustainability challenges, and there is a need for improved tools to assess the implications of addressing these challenges together. Due to the high computational cost of accurately assessing these challenges, integrated assessment models (IAMs) used in policy development often use global- or regional-scale marginal response factors to calculate air quality impacts of climate scenarios. We bridge the gap between IAMs and high-fidelity simulation by developing a computationally efficient approach to quantify how combined climate and air quality interventions affect air quality outcomes, including capturing spatial heterogeneity and complex atmospheric chemistry. We fit individual response surfaces to high-fidelity model simulation output for 1525 locations worldwide under a variety of perturbation scenarios. Our approach captures known differences in atmospheric chemical regimes and can be straightforwardly implemented in IAMs, enabling researchers to rapidly estimate how air quality in different locations and related equity-based metrics will respond to large-scale changes in emission policy. We find that the sensitivity of air quality to climate change and air pollutant emission reductions differs in sign and magnitude by region, suggesting that calculations of "co-benefits" of climate policy that do not account for the existence of simultaneous air quality interventions can lead to inaccurate conclusions. Although reductions in global mean temperature are effective in improving air quality in many locations and sometimes yield compounding benefits, we show that the air quality impact of climate policy depends on air quality precursor emission stringency. Our approach can be extended to include results from higher-resolution modeling and also to incorporate other interventions toward sustainable development that interact with climate action and have spatially distributed equity dimensions.

Hierarchical Surrogate-Assisted Evolutionary Algorithm for Integrated Multi-Objective Optimization of Well Placement and Hydraulic Fracture Parameters in Unconventional Shale Gas Reservoir

Integrated optimization of well placement and hydraulic fracture parameters in naturally fractured shale gas reservoirs is of significance to enhance unconventional hydrocarbon energy resources in the oil and gas industry. However, the optimization task usually presents intensive computation-cost due to numerous high-fidelity model simulations, particularly for field-scale application. We present an efficient multi-objective optimization framework supported by a novel hierarchical surrogate-assisted evolutionary algorithm and multi-fidelity modeling technology. In the proposed framework, both the net present value (NPV) and cumulative gas production (CGP) are regarded as the bi-objective functions that need to be optimized. The hierarchical surrogate-assisted evolutionary algorithm employs a novel multi-fidelity particle-swarm optimization of a global–local hybridization searching strategy where the low-fidelity surrogate model is capable of exploring the populations globally, while the high-fidelity models update the current populations and thus generate the next generations locally. The multi-layer perception is chosen as a surrogate model in this study. The performance of our proposed hierarchical surrogate-assisted global optimization approach is verified to optimize the well placement and hydraulic fracture parameters on a hydraulically fractured shale gas reservoir. The proposed surrogate model can obtain both the NPV and CPG with satisfactory accuracy with only 500 training samples. The surrogate model significantly contributes to the convergent performance of multi-objective optimization algorithm.

Multifidelity Heating Prediction of Adaptable, Deployable Entry Placement Technology Vehicles

The objective of this work was to investigate a multifidelity modeling approach to accurately and efficiently predict the laminar convective, turbulent convective, and radiative heating on adaptable, deployable entry placement technology vehicles in Mars entry. A previously developed cokriging-based multifidelity modeling approach was used to model the heat fluxes at several surface locations along the vehicle, including the rib sections. The laminar convective heat flux multifidelity model was found to have a mean convective heat rate error of approximately 3% when compared with high-fidelity computational fluid dynamics (CFD) simulations. The turbulent convective heat flux multifidelity model was found to have a mean convective heat rate error of approximately 8% when compared with high-fidelity CFD simulations. The radiative heat flux multifidelity model was found to have a mean convective heat rate error of approximately 10% when compared with high-fidelity CFD simulations. Compared with a single-fidelity model, the multifidelity model required, at most, one-third the number of high-fidelity model evaluations to obtain the same accuracy level. The computational cost of evaluating the multifidelity model was approximately six orders of magnitude less than one high-fidelity model simulation.

Multifidelity Turbulent Heating Prediction of Hypersonic Inflatable Aerodynamic Decelerators with Surface Scalloping

The objective of this work was to investigate a multifidelity modeling approach to accurately and efficiently predict the turbulent convective heating on hypersonic inflatable aerodynamic decelerators with both smooth and scalloped walls. A previously developed co-kriging-based multifidelity modeling approach was used to model the turbulent and laminar convective heat fluxes on smooth wall vehicles. The smooth wall turbulent and laminar multifidelity heating models were then combined to create a multifidelity model of the augmented turbulent heat flux on scalloped vehicles. The smooth wall turbulent heat-flux multifidelity model was found to have a mean convective heat rate error of approximately 7% when compared to high-fidelity computational fluid dynamics (CFD) simulations. The scalloped augmented turbulent heat-flux multifidelity model was found to have a mean convective heat rate error of approximately 10% when compared to high-fidelity CFD simulations. Compared to a single-fidelity model, the multifidelity model required approximately one-quarter the number of high-fidelity model evaluations to obtain the same accuracy level. The computational cost of evaluating the multifidelity model was approximately five orders of magnitude less than one high-fidelity model simulation.

Conditioning of deep-learning surrogate models to image data with application to reservoir characterization

Imaging-type monitoring techniques are used in monitoring dynamic processes in many domains, including medicine, engineering, and geophysics. This paper aims to propose an efficient workflow for application of such data for the conditioning of simulation models. Such applications are very common in e.g. the geosciences, where large-scale simulation models and measured data are used to monitor the state of e.g. energy and water systems, predict their future behavior and optimize actions to achieve desired behavior of the system. In order to reduce the high computational cost and complexity of data assimilation workflows for high-dimensional parameter estimation, a residual-in-residual dense block extension of the U-Net convolutional network architecture is proposed, to predict time-evolving features in high-dimensional grids. The network is trained using high-fidelity model simulations. We present two examples of application of the trained network as a surrogate within an iterative ensemble-based workflow to estimate the static parameters of geological reservoirs based on binary-type image data, which represent fluid facies as obtained from time-lapse seismic surveys. The differences between binary images are parameterized in terms of distances between the fluid-facies boundaries, or fronts. We discuss the impact of the choice of network architecture, loss function, and number of training samples on the accuracy of results and on overall computational cost. From comparisons with conventional workflows based entirely on high-fidelity simulation models, we conclude that the proposed surrogate-supported hybrid workflow is able to deliver results with an accuracy equal to or better than the conventional workflow, and at significantly lower cost. Cost reductions are shown to increase with the number of samples of the uncertain parameter fields. The hybrid workflow is generic and should be applicable in addressing inverse problems in many geophysical applications as well as other engineering domains.

Multifidelity Modeling for Efficient Aerothermal Prediction of Deployable Entry Vehicles

The objective of this work was to investigate a multifidelity modeling approach to accurately and efficiently predict the aerothermal response of a large-diameter deployable hypersonic entry vehicle in Mars entry. A cokriging-based multifidelity modeling approach was developed that used several refinements including lower–upper decomposition for parallelization, distance weighted root-mean-square error adaptive sampling, and surface distribution parameterization using Hicks–Henne bump functions. Several computational tools of varying fidelity were investigated to model the surface convective heat flux, shear stress, pressure, and radiative heat flux in the multifidelity modeling process. The multifidelity model was found to have a mean convective heat rate error of 4.6%, a mean pressure force error of 0.81%, a mean shear force error of 2.86%, and a mean radiative heat rate error of 11.1% when compared to high-fidelity computational fluid dynamics simulations. Compared to a single-fidelity model, the multifidelity model required approximately one-half the number of high-fidelity model evaluations to obtain the same accuracy level. The computational cost of constructing and evaluating the multifidelity model were approximately one and five orders of magnitude less, respectively, than one high-fidelity model simulation.

Data-driven modelling of nonlinear spatio-temporal fluid flows using a deep convolutional generative adversarial network

Deep learning techniques for fluid flow modelling have gained significant attention in recent years. Advanced deep learning techniques achieve great progress in rapidly predicting fluid flows without prior knowledge of the underlying physical relationships. However, most of existing researches focused mainly on either sequence learning or spatial learning, rarely on both spatial and temporal dynamics of fluid flows (Reichstein et al., 2019). In this work, an Artificial Intelligence (AI) fluid model based on a general deep convolutional generative adversarial network (DCGAN) has been developed for predicting spatio-temporal flow distributions. In deep convolutional networks, the high-dimensional flows can be converted into the low-dimensional “latent” representations. The complex features of flow dynamics can be captured by the adversarial networks. The above DCGAN fluid model enables us to provide reasonable predictive accuracy of flow fields while maintaining a high computational efficiency. The performance of the DCGAN is illustrated for two test cases of Hokkaido tsunami with different incoming waves along the coastal line. It is demonstrated that the results from the DCGAN are comparable with those from the original high fidelity model (Fluidity). The spatio-temporal flow features have been represented as the flow evolves, especially, the wave phases and flow peaks can be captured accurately. In addition, the results illustrate that the online CPU cost is reduced by five orders of magnitude compared to the original high fidelity model simulations. The promising results show that the DCGAN can provide rapid and reliable spatio-temporal prediction for nonlinear fluid flows.

Explicit Nonlinear Model Predictive Control for Electric Vehicle Traction Control

This paper presents a traction control (TC) system for electric vehicles with in-wheel motors, based on explicit nonlinear model predictive control. The feedback law, available beforehand, is described in detail, together with its variation for different plant conditions. The explicit controller is implemented on a rapid control prototyping unit, which proves the real-time capability of the strategy, with computing times on the order of microseconds. These are significantly lower than the required time step for a TC application. Hence, the explicit model predictive controller can run at the same frequency as a simple TC system based on proportional integral (PI) technology. High-fidelity model simulations provide: 1) a performance comparison of the proposed explicit nonlinear model predictive controller (NMPC) with a benchmark PI-based traction controller with gain scheduling and anti-windup features, and 2) a performance comparison among two explicit and one implicit NMPCs based on different internal models, with and without consideration of transient tire behavior and load transfers. Experimental test results on an electric vehicle demonstrator are shown for one of the explicit NMPC formulations.

Unsupervised machine learning based on non-negative tensor factorization for analyzing reactive-mixing

Analysis of reactive-diffusion simulations representing complex mixing processes requires a large number of independent model runs. For each high-fidelity model simulation, the model inputs are varied and the predicted mixing behavior is represented by temporal and spatial changes in species concentration. It is then required to discern how the model inputs (such as diffusivity, dispersion, anisotropy, and velocity field properties) impact the mixing process. This task is challenging and typically involves interpretation of large model outputs representing temporal and spatial changes of species concentration within the model domain. However, the task can be automated and substantially simplified by applying Machine Learning (ML) methods. In this paper, we present an application of an unsupervised ML method (called NTFk) using Non-negative Tensor Factorization (NTF) coupled with a custom clustering procedure based on k-means to reveal the temporal and spatial features in product concentrations. An attractive and unique aspect of the proposed ML method is that it ensures the extracted features are non-negative, which is important to obtain a meaningful deconstruction of the mixing processes. The ML methodology is applied to a large set of high-resolution finite-element model simulations representing anisotropic reaction-diffusion processes in perturbed vortex-based velocity fields. The applied finite-element method ensures that spatial and temporal species concentration are always non-negative, even in the case of high anisotropic contrasts. The simulated reaction is a fast irreversible bimolecular reaction A+B→C, where species A and B react to form species C. The reactive-diffusion model input parameters that control mixing include properties of the velocity field (such as vortex structures), anisotropic dispersion, and molecular diffusion. We demonstrate the applicability of the ML feature extraction method to produce a meaningful deconstruction of model outputs to discriminate between different physical processes impacting the reactants, their mixing, and the spatial distribution of the product C. The presented ML analysis allowed us to identify additive temporal and spatial features that characterize mixing behavior. The application of the proposed NTFk approach is not limited to reactive-mixing. NTFk can be readily applied to any observed or simulated datasets that can be represented as tensors (multi-dimensional arrays) and have separable latent signatures or features.

Multifidelity Analysis for Predicting Rare Events in Stochastic Computational Models of Complex Biological Systems.

Rare events such as genetic mutations or cell-cell interactions are important contributors to dynamics in complex biological systems, eg, in drug-resistant infections. Computational approaches can help analyze rare events that are difficult to study experimentally. However, analyzing the frequency and dynamics of rare events in computational models can also be challenging due to high computational resource demands, especially for high-fidelity stochastic computational models. To facilitate analysis of rare events in complex biological systems, we present a multifidelity analysis approach that uses medium-fidelity analysis (Monte Carlo simulations) and/or low-fidelity analysis (Markov chain models) to analyze high-fidelity stochastic model results. Medium-fidelity analysis can produce large numbers of possible rare event trajectories for a single high-fidelity model simulation. This allows prediction of both rare event dynamics and probability distributions at much lower frequencies than high-fidelity models. Low-fidelity analysis can calculate probability distributions for rare events over time for any frequency by updating the probabilities of the rare event state space after each discrete event of the high-fidelity model. To validate the approach, we apply multifidelity analysis to a high-fidelity model of tuberculosis disease. We validate the method against high-fidelity model results and illustrate the application of multifidelity analysis in predicting rare event trajectories, performing sensitivity analyses and extrapolating predictions to very low frequencies in complex systems. We believe that our approach will complement ongoing efforts to enable accurate prediction of rare event dynamics in high-fidelity computational models.

D-RM Builder: A software tool for generating fast and accurate nonlinear dynamic reduced models from high-fidelity models

Abstract Dynamic reduced models (D-RMs) derived from rigorous models are highly desired for speeding up dynamic simulations. A useful software tool named D-RM Builder was developed to automatically generate data-driven D-RMs from high-fidelity dynamic models. It allows a user to configure input/output variables, sample input space and generate sequences of step changes, launch high-fidelity model simulations, fit simulation results to a D-RM, and finally visualize and validate the D-RM. The Decoupled A-B Net (DABNet) nonlinear system identification model was used as the main D-RM type and was enhanced to model nonlinear multiple input and multiple output dynamic systems with options for double-pole formulation to handle fast/slow time scales and pole value optimization. The D-RM Builder tool has been successfully used to generate D-RMs for a highly nonlinear pH neutralization reactor system and a two-time-scale bubbling fluidized bed adsorber-reactor for CO2 capture.

Evaluation of multiple reduced-order models to enhance confidence in global sensitivity analyses

Variance-based global sensitivity analysis (e.g., the Sobol’ sensitivity index) can be used to identify the important parameters over the entire parameter space. However, one often cannot afford the computational costs of sampling-based approaches in combination with expensive high-fidelity forward models. Reduced-order models (ROM) can substantially accelerate calculation of these sensitivities. However, it is usually difficult to determine what type of ROM should be used and how accurately the ROM represents the high-fidelity model (HFM) results. In this paper, we propose to concurrently use multiple ROMs as a way to assess the robustness of the model-reduction method. Two sets of HFM simulations are needed, one set for building ROMs and the other for validating ROMs. Our goal is to keep the total number of HFM simulations to a minimum. Ideally some of the HFM simulations in the first set can be shared by different ROMs. Based on validation results, the ROMs can be combined with different schemes. We demonstrate that we can achieve the goal by using four different ROMs and still considerably save computational time compared to using traditional HFM simulation for calculating sensitivity indices. We apply the approach to an example problem of a large-scale geological carbon dioxide storage system, in which the objective is to calculate a sensitivity index to identify important parameters. For this problem, the locally best ROM provides better estimates than the weighted average from all ROMs.

Variable-fidelity aerodynamic shape optimisation of single-element airfoils at high-lift conditions

A computationally efficient simulation-driven variable-fidelity design methodology for single-element airfoils at high-lift conditions is described. Direct optimisation of an accurate but computationally expensive high-fidelity simulation model based on the Reynolds-averaged Navier-Stokes equations and a one equation turbulence model is replaced by iterative updating and re-optimisation of a cheap surrogate. The surrogate model exploits a low-fidelity model (a polynomial approximation of the high-fidelity model data) and appropriate correction that aim at aligning its corresponding airfoil surface pressure distribution and skin friction distribution with that of the high-fidelity model using a shape-preserving response prediction technique. The test cases include lift coefficient maximisation of single-element airfoils at high-lift conditions, subject to a constraint on the drag coefficient. Over 83% reduction in the number of high-fidelity model simulations is demonstrated when compared to high-fidelity model optimisation using a pattern-search algorithm.