Simulation Database Research Articles

Development and Validation of Machine-Learned Actuator Line Model for Hydrokinetic Turbine Rotor

Abstract This study develops and validates a machine-learned (ML) actuator line model (ALM) as a promising advancement in turbine/rotor modeling that can be applied for diverse engineering applications. The model alleviates two limitations of the standard ALM, namely, its reliance on the pre-defined lift and drag coefficient tables and its inability to account for flow unsteadiness. The ML-ALM model is trained using a blade-resolved simulation database of forces acting on blade elements for unsteady inflow conditions. The model is validated for solitary turbine performance and wake predictions against experimental data, and is verified for an inline turbine case for the performance and wake predictions of the downstream turbine against blade-resolved simulations. Its engineering applicability is demonstrated for an 8-turbine array farm simulation. The ML-ALM predicts turbine performance and wakes within 10% of blade-resolved results along with credible advection of tip vortical structures including breakdown and turbulent kinetic energy burst, using 92% less computational time than corresponding blade-resolved simulations.

Mapping neutron biological effectiveness for DNA damage induction as a function of incident energy and depth in a human sized phantom

We present new developments for an ab-initio model of the neutron relative biological effectiveness (RBE) in inducing specific classes of DNA damage. RBE is evaluated as a function of the incident neutron energy and of the depth inside a human-sized reference spherical phantom. The adopted mechanistic approach traces neutron RBE back to its origin, i.e. neutron physical interactions with biological tissues. To this aim, we combined the simulation of radiation transport through biological matter, performed with the Monte Carlo code PHITS, and the prediction of DNA damage using analytical formulas, which ground on a large database of biophysical radiation track structure simulations performed with the code PARTRAC. In particular, two classes of DNA damage were considered: sites and clusters of double-strand breaks (DSBs), which are known to be correlated with cell fate following radiation exposure. Within a coherent modelling framework, this approach tackles the variation of neutron RBE in a wide energy range, from thermal neutrons to neutrons of hundreds of GeV, and reproduces effects related to depth in the human-sized receptor, as well as to the receptor size itself. Besides providing a better mechanistic understanding of neutron biological effectiveness, the new model can support better-informed decisions for radiation protection: indeed, current neutron weighting (ICRP)/quality (U.S. NRC) factors might be insufficient for use in some radiation protection applications, because they do not account for depth. RBE predictions obtained with the reported model were successfully compared to the currently adopted radiation protection standards when the depth information is not relevant (at the shallowest depth in the phantom or for very high energy neutrons). However, our results demonstrate that great care is needed when applying weighting factors as a function of incident neutron energy only, not explicitly considering RBE variation in the target. Finally, to facilitate the use of our results, we propose look-up RBE tables, explicitly considering the depth variable, and an analytical representation of the maximal RBE vs. neutron energy.

Operation optimization for a CHP system using an integrated approach of ANN and simulation database

Operation optimization for a CHP system using an integrated approach of ANN and simulation database

OPRLM: A Web Tool and a Database for Positioning and Simulations of Proteins in Realistic Lipid Membranes

OPRLM: A Web Tool and a Database for Positioning and Simulations of Proteins in Realistic Lipid Membranes

Modeling of vibrational nonequilibrium effects on pressure Hessian tensor using physics-assisted deep neural networks

This study focuses on modeling the effect of vibrational nonequilibrium on the pressure Hessian tensor. The pressure Hessian tensor is one of the unclosed processes involved in the velocity gradient evolution equation. Accessing the velocity gradient tensor following a fluid particle is needed to understand the physics of various nonlinear turbulent processes. Our modeling strategy employs a combination of two different deep neural networks (DNNs) to model the magnitude and the directional aspects of the vibrational nonequilibrium tensor separately. Both the DNNs are optimized using two appropriate physics-assisted custom loss functions, comparing the desired features of the DNN predictions against the exact behavior observed in the direct numerical simulation (DNS) database. A detailed investigation of four different DNS databases of vibrationally excited decaying compressible turbulence with different initial Reynolds numbers, Mach numbers, and vibrational Damköhler numbers is performed to identify the appropriate normalized forms of input and output quantities for the two neural networks. The training of both the DNNs is done using the DNS data of merely one simulation. The trained models are then subjected to extensive evaluation against different DNS databases. Indeed, the new model captures many DNS features quite well. Such an extensive evaluation of the new model proves the generalizability of the model, at least in the range of parameters involved in our study.

An overview of the STEP divertor design and the simple models driving the plasma exhaust scenario

Abstract This paper presents a comprehensive overview of the preliminary divertor design and plasma exhaust scenario for the reactor-class Spherical Tokamak for Energy Production (STEP) project. Due to the smaller size of the machine, with a major radius less than half that of most DEMO concepts, the current design features a double-null divertor geometry, comprising tightly baffled extended outer legs and shorter inner legs approaching an X-divertor. Leveraging a significant database of SOLPS-ITER simulations, the exhaust operational space is mapped out, offering valuable insights into the plasma exhaust dynamics. An approach involving the validation of simple, yet robust models capable of accurately predicting key exhaust parameters is detailed, thereby streamlining the design process. The simple models are used to simulate the entire plasma scenario from the plasma current ramp-up, through the burning phase, to the plasma current ramp-down. Notably, the findings suggest that pronounced detachment, with peak heat loads below engineering limits and electron temperatures below 5 eV, is achievable with a divertor neutral pressure between 10 - 15 Pa during
the burning phase, and pressures below 5 Pa during the ramp-up to maximise the auxiliary current-drive efficiency. Throughout the scenario, an Ar concentration of ~3% in the scrape-off layer (SOL) is required, in combination with a core radiation fraction of 70% driven by intrinsic emission and extrinsic injection of Xe seeded fuelling pellets. However, significant uncertainties remain regarding key parameters such as the SOL heat flux width, Ar screening, and plasma kinetic effects.

Nonlinear finite element model updating and fourier neural operator for digital twin of reinforced concrete containment vessel under seismic excitations

A series of unidirectional seismic excitation tests on a 1/20 scaled reinforced concrete containment vessel (RCCV) specimen was conducted using a 6-DOF shake table. This study presents the development of a digital twin to accurately capture the complex behavior of reinforced concrete structures under seismic effects. With the aid of finite element (FE) software Abaqus, the RCCV structure was initially modeled and analyzed to obtain preliminary structural acceleration and displacement results. To enhance the accuracy of these simulations, a nonlinear FE model updating framework was developed by integrating Abaqus software with Python. The model updating of RCCV was conducted subsequently based on the shake table measured data to obtain results with significantly improved accuracy. It is shown that the FE simulations and the nonlinear FE model updating algorithm are highly effective for the high-fidelity simulation of large and complex structures. Furthermore, a Fourier neural operator (FNO) model is used to learn the FE simulation database and predict the FE simulation results. It is observed that FNO has an excellent performance in predicting the structural responses with various material and damping parameters and accomplished speedup over 5000 times compared to the FE simulations. The findings provide a basis for the application of nonlinear model updating algorithms at the structure system level through the high-performance GPU-based parallel computing and underscore that the strategy of FE analysis combined with artificial intelligence (AI) has great potential in the field of digital twins for large and complex structures.

Anisotropy of Reynolds Stresses and Their Dissipation Rates in Lean H2-Air Premixed Flames in Different Combustion Regimes

The interrelation between Reynolds stresses and their dissipation rate tensors for different Karlovitz number values was analysed using a direct numerical simulation (DNS) database of turbulent statistically planar premixed H2-air flames with an equivalence ratio of 0.7. It was found that a significant enhancement of Reynolds stresses and dissipation rates takes place as a result of turbulence generation due to thermal expansion for small and moderate Karlovitz number values. However, both Reynolds stresses and dissipation rates decrease monotonically within the flame brush for large Karlovitz number values, as the flame-generated turbulence becomes overridden by the strong isotropic turbulence. Although there are similarities between the anisotropies of Reynolds stress and its dissipation rate tensors within the flame brush, the anisotropy tensors of these quantities are found to be non-linearly related. The predictions of three different models for the dissipation rate tensor were compared to the results computed from DNS data. It was found that the model relying upon isotropy and a linear dependence between the Reynolds stress and its dissipation rates does not correctly capture the turbulence characteristics within the flame brush for small and moderate Karlovitz number values. In contrast, the models that incorporate the dependence of the invariants of the anisotropy tensor of Reynolds stresses were found to capture the components of dissipation rate tensor for all Karlovitz number conditions.

A hybrid load prediction method of office buildings based on physical simulation database and LightGBM algorithm

Building load prediction plays an important role in building energy savings and mechanical and electrical system optimization control. The dynamic energy consumption can be accurately calculated using the traditional physical energy simulation method that entails a complex setup and verification process due to the numerous input parameters required. It is also difficult to change the physical model once it has been determined. The data-mining method is fast in calculation and simple to use, but its prediction accuracy is limited by historical data quality. It is difficult to predict the load of new buildings without historical data. To solve these problems, this study proposes a hybrid building load prediction method for office buildings. The proposed method uses EnergyPlus to generate a building load database that includes than 25.14 million data cases, covering 35 types of building geometry and 2870 building examples. Based on the above database, the LightGBM algorithm was selected to extract feature variables that affect the load and build a load prediction model. The training results show that there are 24 key feature variables for office building load prediction. The hourly MAPE of the cooling load prediction model is 6.95 % and RMSE is 4.31 W/m2, and the hourly MAPE of the heating load prediction model is 7.09 % and RMSE is 11.64 W/m2 compared with EnergyPlus model. Two actual office buildings are selected as case studies to validate the model prediction accuracy. Results show that comparing predicted results with measured data, the hourly cooling load of the MAPE is 12.42 %. Coparing predicted results with actual heating load, daily MAPE is 7.97 %.

THE INFLUENCE OF OBSERVATIONAL ERRORS IN GAIA DR3 ON THE RECONSTRUCTION OF GLOBULAR CLUSTER ORBITS ON A COSMOLOGICAL TIMESCALE

In recent years, the emerging field of astronomy focused on the history of galaxy formation, known as Galactic Archaeology, has been gaining popularity. Globular clusters have been involved in many key processes occurring in the Milky Way, making their study, particularly the reconstruction of their orbits, significantly important. The Gaia DR3 catalog provides parameters for 165 globular clusters, such as proper motions, radial velocity, and heliocentric distance, with certain accuracy. Therefore, it is important to examine the influence of measurement errors in these parameters on the initial data when converting to the Galactocentric coordinate system and, consequently, on the shape of the orbits. We integrated the orbits of globular clusters 10 billion years lookback. For physical justification during the integration, we used the external dynamic potential with the individual number 411321 from the cosmological simulation database IllustrisTNG-100, which best reproduces the potential of the Milky Way. The integration was performed using the parallel N-body code φ-GPU, based on a fourth-order Hermite scheme with hierarchical individual block timesteps. A total of 1,000 randomizations of the initial data were created considering a normal distribution of errors, and the influence of errors on the scatter of initial velocities and on the shape of the orbits was examined. The parameters with the largest relative errors are proper motions and radial velocity, while the smallest errors are in heliocentric distance. It was found that 85% of the globular clusters have relative errors in all parameters of no more than 10%, and 5.4% have errors of no more than 1%. Investigating the influence of measurement errors for clusters with different magnitudes of relative errors, we concluded that for most globular clusters, the influence of measurement errors on the shape of the orbits is not significant. Consequently, it is possible to reconstruct the orbits with high accuracy for these clusters. Since the reconstruction of globular cluster orbits involves cosmological timescales, accounting for measurement errors is an important aspect of the preparatory procedure before the main integration.

Three-dimensional reconstruction of laser-direct-drive inertial confinement fusion hot-spot plasma from x-ray diagnostics on the OMEGA laser facility (invited).

A deep-learning convolutional neural network (CNN) is used to infer, from x-ray images along multiple lines of sight, the low-mode shape of the hot-spot emission of deuterium-tritium (DT) laser-direct-drive cryogenic implosions on OMEGA. The motivation of this approach is to develop a physics-informed 3-D reconstruction technique that can be performed within minutes to facilitate the use of the results to inform changes to the initial target and laser conditions for the subsequent implosion. The CNN is trained on a 3D radiation-hydrodynamic simulation database to relate 2D x-ray images to 3D emissivity at stagnation. The CNN accounts for the lack of an absolute spatial reference and the different bands of photon energies in the x-ray images. While previous work [O. M. Mannion etal., Phys. Plasmas 28, 042701 (2021) and A. Lees etal., Phys. Rev. Lett. 127, 105001 (2021)] studied the effect of mode-1 asymmetries on implosion performance using nuclear diagnostics, this work focuses on the effect of mode 2 inferred from x-ray diagnostics on implosion performance. A current analysis of 19 DT cryogenic implosions indicates there is an upper limit of ∼20% reduction in the neutron yield caused by an ℓ = 2 amplitude for ℓ2/ℓ0 ≤ 0.32. These conclusions are supported by 2D simulations.

Design of a polarization-insensitive broadband achromatic metalens for mid-wave infrared detector.

Metalenses, boasting outstanding focusing efficiency and high-resolution imaging capabilities, have generated widespread usage in fields such as integrated optics, achromatic imaging, and optical holography. In this study, we have developed a broadband achromatic metalens within the detection range from 3 to 5 µm, and it has a numerical aperture (NA) of 0.71 with a remarkable maximum focusing efficiency of 63.8% at the focal plane within the specified bandwidth. We have further delved into the dispersion control mechanism that combines the geometric and transmission phases and optimized the constructed phase response simulation database using the particle swarm optimization (PSO) algorithm, ensuring a precise phase matching between the actual wavefront and the ideal focusing wavefront. This metalens with its ability to expand the array size has the potential to create a compact infrared imager, which holds significant importance in achieving efficient detection and integration within infrared detectors.

Evaluation of spatial resolution effects in rough wall-bounded turbulence

This study investigates the impact of insufficient spanwise spatial resolution on the measurement accuracy of streamwise velocity fluctuations over rough walls. We use a direct numerical simulation (DNS) database of turbulent open-channel flow over three-dimensional sinusoidal roughness with varied wavelengths and roughness heights. Employing a triple decomposition, we investigate both the attenuation of the turbulent fluctuations (about the local mean), u′\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$u^\\prime$$\\end{document} and the dispersive stresses (roughness-induced fluctuations of the time-averaged mean about the global mean), U~\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$${\ ilde{U}}$$\\end{document}. A boxcar filter on DNS data is applied to investigate the effects of spanwise spatial filtering on these quantities. Our analysis reveals the significance of two key length-scale ratios for velocity measurements over rough walls: the wire length relative to the spatially and temporally plane-averaged Kolmogorov scale at the roughness crest (l/⟨η⟩k\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$l/\\langle \\eta \\rangle _k$$\\end{document}), and the wire length relative to the roughness spanwise wavelength (l/Λy\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$l/\\Lambda _y$$\\end{document}). We observe that maintaining l/⟨η⟩k\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$l/\\langle \\eta \\rangle _k$$\\end{document} constant while increasing l/Λy\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$l/\\Lambda _y$$\\end{document} attenuates the variance of U~\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$${\ ilde{U}}$$\\end{document} and u′\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$u^\\prime$$\\end{document} within the roughness sublayer. When fixing l/Λy\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$l/\\Lambda _y$$\\end{document}, an increase in l/⟨η⟩k\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$l/\\langle \\eta \\rangle _k$$\\end{document} influences the turbulent fluctuations across all wall-normal locations. These findings highlight the necessity of considering both length scales when evaluating spanwise spatial resolution in turbulence measurements over rough walls.

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).

Experimental study on the spray characteristics of high-pressure liquid ammonia under different ambient conditions

As a promising carbon-free fuel, ammonia is expected to be widely applied in internal combustion engines. However, the physical properties of ammonia are quite different from those of conventional fuels, which leads to different spray characteristics. In this paper, the ammonia spray under injection pressure as high as 80 MPa was visualized by the diffused back-illumination imaging method, and the liquid ammonia spray characteristics under different ambient pressures and ambient temperatures were analyzed. The liquid spray penetration length, cone angle and tip velocity calculated from the spray images provide a reference database for numerical simulation. The results show that the development characteristics of liquid ammonia spray are significantly different under flare flash boiling, transitional flash boiling and non-flash boiling conditions. Flash boiling (especially flare flash boiling) inhibits the initial liquid spray penetration. The spray tip velocity increases first and then gradually decreases under flash boiling conditions. Ammonia spray has obvious radial expansion at the initial stage of flare flash boiling and the spray contour under flare flash boiling conditions is noticeably distorted, forming an obvious dilute region. With the increase of ambient pressure, the intensity of flash boiling reduces, the dilute region gradually disappears, and the distortion of the spray contour gradually weakens. Under non-flash boiling conditions, ammonia spray presents a dense and regular shape; there is no spray acceleration but a sharp decrease in spray tip velocity at the initial stage, and then the spray penetrates forward at a similar velocity. The spray penetration velocity decreases significantly with the increase of ambient pressure. The increase in ambient temperature accelerates the vaporization of ammonia spray. The liquid spray penetration length decreases and maintains with only slight fluctuations during the injection process as the ambient temperature increases from 300 K to 600 K because the vaporization rate and penetration velocity reach a balance.

Digital Resources Construction and Education Research for the Course of Material Research and Testing Methods

The course of Materials Research and Testing Methods aims to help students master the skills of testing and analyzing the composition, structure, property and performance of materials. To improve the teaching quality of this course and help the students understand well about the content, we introduced the distinctive teaching methods such as applying case teaching, developing virtual simulation software and database of cutting-edge instruments, carrying out instrument training, building platform of instrument popularization and national first-class course, guiding research projects and scientific competitions. Therefore, the course has its own distinctive materials that are collected to build the digital resources. How the teaching quality of this course improved by these methods will also be discussed. Furthermore, the construction of the digital resources can help students realize the new learning mode of learning at any time, flexible learning and long-term learning, broadening the channels and ways for students to study the course.

A novel subgrid-scale stress model considering the influence of combustion on turbulence: A priori and a posteriori assessment

Most classical turbulence models were proposed and developed based on non-reacting flows without considering the effects of combustion on turbulence. The flow structure in turbulent combustion is more complex, creating challenges to the applicability of traditional turbulence models. Given this, a novel flame surface and k-equation-based gradient model (FKGM) considering combustion effects is proposed for the modeling of the subgrid-scale (SGS) stress in large eddy simulation (LES). The SGS stress is calculated based on the SGS kinetic energy (kSGS) and normalized velocity gradient. The velocity gradient incorporates first-order gradients at multiple stencil locations and considers the anisotropy of the stress near the flame surface. The FKGM model is first validated by the a priori analysis based on the direct numerical simulation (DNS) database of a premixed swirling flame. The closure terms of the kSGS equation are well validated, and the predicted SGS stress using the FKGM model achieves good agreement with the filtered DNS results. In the a posteriori LES study, the FKGM model demonstrates better performance compared with the traditional dynamic Smagorinsky model and velocity gradient model, providing more accurate predictions for mean and fluctuation velocities. The error analysis for SGS kinetic energy is further conducted by comparing the LES results with the DNS data, and the results indicate that the underestimation of the generation term of the kSGS equation is the main source of error.

A Comparison of Evaluation Methodologies of the Fractal Dimension of Premixed Turbulent Flames in 2D and 3D Using Direct Numerical Simulation Data

A Direct Numerical Simulation (DNS) database of statistically planar flames ranging from the wrinkled flamelets to the thin reaction zones regime and DNS data for a Bunsen premixed flame representing the wrinkled flamelets regime have been utilised to evaluate the fractal dimensions of flame surfaces using the filtering dimension method, the box-counting algorithm and the correlation dimension approach. The fractal dimension evaluated based on the fully resolved three-dimensional data has been found to be reasonably approximated by adding unity to the equivalent fractal dimension evaluated based on two-dimensional projections irrespective of the methodology of extracting fractal dimension. This indicates that the flame surface can be approximated as a self-similar fractal surface for the range of Karlovitz and Damköhler numbers considered here. While all methods, provide results identical to each other for benchmark problems, it has been found that the fractal dimension evaluation based on box-counting method provides almost identical results as that obtained using the filtering dimension method for both three and two dimensions, while the fractal dimensions based on the correlation dimension tend to be slightly smaller. The findings of the current analysis have the potential to be used to reliably estimate the actual fractal dimension in 3D based on experimentally obtained 2D binarised reaction progress variable field. The inner cut-off scales estimated based on all three methodologies yield comparable results in terms of order of magnitude with the box-counting method predicting a smaller value of inner cut-off scale in comparison to other methods. The execution times for fractal dimension extraction based on filtering dimension and box-counting methodologies are found to be comparable but the correlation dimension method is found to be considerably faster than the two alternative approaches and provides results consistent with theoretical bounds in all cases.

Extremely high wall pressure events in shock wave and turbulent boundary layer interactions using DNS data

This study investigates high-amplitude Extreme Wall Pressure fluctuation Events (EWPEs) in Shock wave/Turbulent Boundary Layer Interactions (STBLIs) through the conditional sampling of direct numerical simulation databases. The aim is to evaluate the effect of STBLIs and their strength on the statistical properties and associated turbulent structures of EWPEs using the conditional-averaging and clustering method. The temporal statistical results show that the occurrence probability and contribution ratio of EWPEs decrease downstream of strong STBLI, but their duration and interval time increase. Regarding two-dimensional wall pressure structures, the large population of small-scale structures becomes more elongated, but strong interactions induce a greater number of large-scale structures. The pairing of wall pressure events with a higher occurrence probability is verified by the joint probability density functions. Conditional analysis reveals that, as the interaction strength increases, the ejection motion associated with positive events occurs farther downstream and the spanwise vortex core locating above negative events is lifted up along the wall-normal direction. Moreover, analysis associates the paired wall pressure events with the sweep, ejection, and swirl motions in STBLIs, where hairpin eddies play an important role in the formation of positive–negative paired wall pressure structures.

“Why Should I Trust You?”: Exploring Interpretability in Machine Learning Approaches for Indirect SHM

Currently, machine learning (ML) methods are widely adopted in structural health monitoring (SHM), yet they are still mostly black boxes. On the other hand, given the significant responsibility associated with SHM, understanding the rationale behind it is critically important. In some cases, even experienced experts have difficulties finding evidence related to structural integrity within complex structural signals. Thus, solely relying on these black-box SHM systems carries inherent risks. Trustworthiness is the key for decision-makers when planning to act on predictions or deciding whether to deploy a new model. This understanding can also offer insights about the models, transforming untrustworthy models or predictions into reliable ones. The indirect SHM method using passing vehicles, an emerging technique in the past two decades, offers a rapid and cost-effective solution for bridge monitoring. Its signal components are affected by factors such as vehicle dynamics and road roughness, making them more complex than those in the direct method. Although ML methods have shown promising results in this domain, their results require further explanation. In this work, SHAP (SHapley Additive exPlanation) tools are utilized to interpret the result prediction of ML methods in indirect SHM. The trustworthiness of models is demonstrated through simulation databases: deciding whether a prediction should be trusted, choosing between models, and determining why a classifier should not be trusted.