Physics Information Research Articles

Research on optimization design of physics teaching objectives based on information fusion technology

Abstract Physics experiment is one of the key points and difficulties of physics teaching, and it has become a recognized bottleneck that restricts the quality of physics teaching. This paper firstly explores the integration of high school physics experimental teaching with information technology based on physics experimental teaching practice and physics curriculum standards through various ways and means. Secondly, it proposes constructing a collaborative physics experimental teaching design scheme based on information fusion technology, using three non-hierarchical multi-model fusion methods with Gaussian stochastic processes for modeling to achieve effective fusion of multiple non-hierarchical models of physics response information and physics teaching. Finally, a practical study in the form of a physics experimental teaching case is analyzed, and the teaching effect is measured to verify the effectiveness and feasibility of the design scheme. The results show that the physics teaching design based on information fusion technology effectively improves students’ experimental performance by 21.2%, learning interest by 36.4%, and learning participation by 19.8%. This study takes physics experimental teaching and information fusion technology as the entry point and improves the overall teaching quality of physics as the goal, which provides reference or reference teaching reform and practice in the context of information technology.

A Study on C-band Synthetic Aperture Radar Soil Moisture Estimation Based on Machine Learning Using Soil Physics, Topography, and Hydrological Information

In this study, we applied machine learning to estimate soil moisture levels in South Korea by harnessing data from the Sentinel-1 C-band synthetic aperture radar (SAR). Our approach incorporated not only the relationship between backscattering coefficients and soil moisture but also diverse physical characteristics. This encompassed topographic information, soil physics data, and antecedent precipitation which is a hydrological factor influencing the initial condition of soil moisture. We applied a variety of machine-learning techniques and conducted a comprehensive analysis to compare the performance of each model.

Multilevel method for predicting flow fields in radial turbines based on sparsity-promoting dynamic mode decomposition

PurposeThe purpose of this study is to propose a fast method for predicting flow fields with periodic behavior with verification in the context of a radial turbine to meet the urgent requirement to effectively capture the unsteady flow characteristics in turbomachinery. Aiming at meeting the urgent requirement to effectively capture the unsteady flow characteristics in turbomachinery, a fast method for predicting flow fields with periodic behavior is proposed here, with verification in the context of a radial turbine (RT).Design/methodology/approachSparsity-promoting dynamic mode decomposition is used to determine the dominant coherent structures of the unsteady flow for mode selection, and for flow-field prediction, the characteristic parameters including amplitude and frequency are predicted using one-dimensional Gaussian fitting with flow rate and two-dimensional triangulation-based cubic interpolation with both flow rate and rotation speed. The flow field can be rebuilt using the predicted characteristic parameters and the chosen model.FindingsUnder single flow-rate variation conditions, the turbine flow field can be recovered using the first seven modes and fitted amplitude modulus and frequency with less than 5% error in the pressure field and less than 9.7% error in the velocity field. For the operating conditions with concurrent flow-rate and rotation-speed fluctuations, the relative error in the anticipated pressure field is likewise within an acceptable range. Compared to traditional numerical simulations, the method requires a lot less time while maintaining the accuracy of the prediction.Research limitations/implicationsIt would be challenging and interesting work to extend the current method to nonlinear problems.Practical implicationsThe method presented herein provides an effective solution for the fast prediction of unsteady flow fields in the design of turbomachinery.Originality/valueA flow prediction method based on sparsity-promoting dynamic mode decomposition was proposed and applied into a RT to predict the flow field under various operating conditions (both rotation speed and flow rate change) with reasonable prediction accuracy. Compared with numerical calculations or experiments, the proposed method can greatly reduce time and resource consumption for flow field visualization at design stage. Most of the physics information of the unsteady flow was maintained by reconstructing the flow modes in the prediction method, which may contribute to a deeper understanding of physical mechanisms.

A Physics-informed Neural Network for Pulsed Thermography- Based Defect Detection

The non-destructive testing technique of pulsed thermography has been widely used in materials to detect defects due to its low cost, large detection area, and rapid implementation. After conducting experiments, data processing is often necessary to reduce the influences of noise and non-uniform backgrounds and highlight defect information. However, during the analysis, physical information is usually ignored. In addition to thermographic data analysis, numerical simulations are also popular for analytical studies based on physical information, but they don’t fully utilize the experimental data. To address this, a new method was proposed using a physics-informed neural network (PINN) for thermographic data processing. PINN combines the prediction capabilities of deep neural networks with physical laws presented as partial differential equations and boundary conditions, allowing for both experimental data and physics information to be utilized in modelling. In pulsed thermography, the heat transfer is governed by Fourier's law of heat conduction in a three-dimensional system. However, there is a lack of temperature measurements in the depth direction. The proposed method solves this problem by using collocation points generated from Latin hypercube sampling. The PINN model provides a good estimation of the backgrounds in the thermograms, and the features of surface/subsurface defects are highlighted by subtracting the estimated backgrounds from the original thermograms. In the case study, the performance of the proposed method was found to be effective.

Prediction of supernova neutrino signals by detectors and its future challenges

Supernova neutrinos produced during a core collapse of a massive star, carries 99% of the energy produced during the violent phenomenon. These neutrinos are weakly interacting massive particles and can provide useful information for both particle physics (neutrino oscillations parameters) and astrophysics (explosion mechanism). This information can be used to explore physics beyond the standard model. Neutrinos escape from the supernova core hours before the light, so a neutrino signal providing information about supernova direction can enable early observation. The current generation of detectors, like, Super-Kamiokande (Super-K), Large Volume Detector (LVD), Borexino, Kamioka Liquid Scintillator Antineutrino Detector (KamLAND), and IceCube, as well as HALO, Daya Bay(reactor neutrino experiment) and NuMI Off-Axis νe Appearance experiment (NOvA), have the ability to detect only a few orders of magnitude of events and the next generation experiment like, Hyper-Kamiokande (Hyper-K), Deep Underground Neutrino Experiment (DUNE), and Jiangmen Underground Neutrino Observatory (JUNO) will have yet another order of magnitude in reach, as well as richer flavor sensitivity. This work will present a Monte Carlo based study using the SNOwGLoBES [1] package, which is used to estimate the event rate using folded fluxes, cross-sections, and detector smearing to determine mean expected neutrino interaction signals in multiple current and future detectors. A study is carried out for the calculation of core-collapse neutrino event rates in realistic detectors for different flux models, effects of different parameters on flux and its variation with time.

Predicting viscosity of ionic liquids - water mixtures by bridging UNIFAC modeling with interpretable machine learning

Predicting the viscosity of ionic liquids (ILs)-water mixtures precisely is considerable for diversity applications in chemical industries. In this work, interpretable machine learning models incorporate physics information developed to link the UNIversal quasi-chemical Functional group Activity Coefficient (UNIFAC) models as new predictive tools for the viscosity of ILs-water mixtures. Machine learning algorithms including Adaptive Boosting (AdaBoost), Categorical Boosting (CatBoost), Extreme Gradient Boosting (XGBoost), K-Nearest-Neighbour Regression (KNNR) and Random Forest (RF) are compared, with the Catboost model achieving the best accuracies (RMSE = 0.0084, MSE = 0.0001, MAE = 0.0020, R2 = 0.9941). We demonstrate that the UNIFAC model and the Stokes-Einstein relation assisted to reduce the feature dimensionality, and improve the predictive power of the viscosity of ILs-water mixtures. The describe of the Shapley’s additive explanations (SHAP) and Partial dependence plots (PDP) are giveing expression to the features importance of the UNIFAC model. Our work created a new paradigm to hyperlink the machine learning models with ILs-related properties, the UNIFAC model contains prior physical information to optimize the features of the experiment, and fill the gap of the input features and the viscosity of ILs-water mixtures.

Recovering the Forcing Function in Systems with One Degree of Freedom Using ANN and Physics Information

In engineering design, oftentimes a system’s dynamic response is known or can be measured, but the source generating these responses is not known. The mathematical problem where the focus is on inferring the source terms of the governing equations from the set of observations is known as an inverse source problem (ISP). ISPs are traditionally solved by optimization techniques with regularization, but in the past few years, there has been a lot of interest in approaching these problems from a deep-learning viewpoint. In this paper, we propose a deep learning approach—infused with physics information—to recover the forcing function (source term) of systems with one degree of freedom from the response data. We test our architecture first to recover smooth forcing functions, and later functions involving abruptly changing gradient and jump discontinuities in the case of a linear system. Finally, we recover the harmonic, the sum of two harmonics, and the gaussian function, in the case of a non-linear system. The results obtained are promising and demonstrate the efficacy of this approach in recovering the forcing functions from the data.

EESD special issue: AI and data‐driven methods in earthquake engineering – (Part 1)

EESD special issue: AI and data‐driven methods in earthquake engineering – (Part 1)

Investigating steady unconfined groundwater flow using Physics Informed Neural Networks

A deep learning technique called Physics Informed Neural Networks (PINNs) is adapted to study steady groundwater flow in unconfined aquifers. This technique utilizes information from underlying physics represented in the form of partial differential equations (PDEs) alongside data obtained from physical observations. In this work, we consider the Dupuit–Boussinesq equation, which is based on the Dupuit–Forchheimer approximation, as well as a recent, more complete model derived by Di Nucci (2018) as underlying models. We then train PINNs on data obtained from steady-state analytical solutions and laboratory based experiments.Using PINNs, we predict phreatic surface profiles given different input flow conditions and recover estimates for the hydraulic conductivity from the experimental observations. We show that PINNs can eliminate the inherent inability of the Dupuit–Boussinesq equation to predict flows with seepage faces. Moreover, the inclusion of physics information from the Di Nucci and Dupuit–Boussinesq models constrains the solution space and produces better predictions than training on data alone. PINNs based predictions are robust and show a little effect from added noise in the training data. Furthermore, we compare the PINNs solutions obtained via the Di Nucci and Dupuit–Boussinesq flow models to examine the effects of higher order flow terms that are included in the Di Nucci formulation but are neglected by the Dupuit–Boussinesq approximation. Lastly, we discuss the effectiveness of using PINNs for examining groundwater flow.

Deep learning-driven nonlinear reduced-order models for predicting wave-structure interaction

The feasibility of using a Long Short-Term Memory (LSTM) network-driven Non-Intrusive Reduced Order Model (NIROM) to predict the dynamics of a two-dimensional box floating and interacting with surface water waves is assessed in this study. The ground for these wave-structure interactions (WSI) problems, namely box displacements and hydrodynamic forces arising from wave interaction corresponding to a particular surface wave profile, are computed using a single-phase Smoothed Particle Hydrodynamics (SPH). The dimensionality of the system is first reduced using the Discrete Empirical Interpolation Method (DEIM) and the LSTM is applied to the reduced system resulting in a DEIM-LSTM network for developing a surrogate for prediction. This is further enhanced by incorporating the physics information into the loss function resulting in a physics-informed LSTM (LSTM-PINN) for predicting the rigid body dynamics of box motion. The performance of these networks in predicting the dynamics of the floating box is compared as a proof-of-concept demonstration.

Building thermal modeling and model predictive control with physically consistent deep learning for decarbonization and energy optimization

Being a primary contributor to global energy consumption and energy-related carbon emissions, the building and building construction sectors are a crucial player in the decarbonization and energy efficiency efforts. This article proposes a novel physically consistent deep learning (PCDL) approach for building thermal modeling and assesses its potential for optimizing building energy efficiency and indoor thermal comfort through model predictive control (MPC). The PCDL model considers physical relationships between system inputs and outputs and is applied to predict the indoor thermal climate. With the strict guarantee of satisfying the laws of physics, the proposed PCDL model has better generalization ability than other machine learning approaches. Subsequently, the PCDL model is integrated into an MPC controller to optimize building energy consumption and indoor thermal comfort. The proposed approach is tested on Carpenter Hall, a multi-zone office building located in Cornell University Campus, through simulations. Based on the simulation results, the PCDL model demonstrate a much better generalization ability for yielding physically-feasible predictions compared to the long short-term memory (LSTM) model. Compared to an On/Off controller, a state space model-based MPC controller, and an LSTM-based MPC controller, the proposed PCDL-based MPC reduces by 5.8%, 4.5%, and 8.9% energy consumption and improves by 55%, 59%, and 64% indoor thermal comfort, respectively, and therefore enhances the building decarbonization progress. The results emphasize the importance of considering physics information in data-driven models and highlight the advantages of the proposed PCDL-based MPC controller.

Combination of physics‐based and data‐driven modeling for nonlinear structural seismic response prediction through deep residual learning

AbstractDespite great progress in seeking accurate numerical approximator to nonlinear structural seismic response prediction using deep learning approaches, tedious training process and large volume of structural response data under earthquakes for training and validation are often prohibitively accessible. In our methodology, the main innovation can be seen in the incorporation of deep neural networks (DNNs) into a classical numerical integration method by using a hybridized integration time‐stepper. In this way, the linear physics information of the structure and the obscure nonlinear dynamics are smoothly combined. We propose to use residual network (ResNet) to learn time‐stepping schemes specifically for the nonlinear state variables of the system. Our Physics‐DNN hybridized integration (PDHI) time‐stepping scheme provides important advantages over current pure data‐driven approaches, including (i) a flexible framework incorporating known time‐invariant physics information, (ii) requirement of structural seismic response data being circumvented by simple short bursts of trajectories collected from underlying nonlinear components, and (iii) efficiency in training and validation process. Besides, our results indicate that a simple feedforward or convolutional architecture outperforms recurrent networks to fulfill the requirement of prediction accuracy as well as long‐range memory in structural dynamic analysis. Several numerical and experimental examples are presented to demonstrate the performance of the method.

Interpretable sparse identification of a bistable nonlinear energy sink

Bistable nonlinear energy sinks have received great interest due to their efficient broad-band targeted energy transfer over a wide range of input energy levels. The precise identification of bistable nonlinear stiffness force is of significance to predict and enhance the system performance of the vibration energy absorption. However, the nonlinear stiffness force in nonlinear energy sink structures with local bistability is difficult to measure and identify because of snap-through characteristics. Inspired by physics-informed data-driven regression in machine learning, an interpretable sparse identification method is proposed to determine the stiffness force of a bistable nonlinear energy sink. The restoring force surface is constructed on bistable nonlinear energy sink equations and the nonlinear stiffness force trajectory is intercepted by assuming two quasi-zero velocity planes. Furthermore, the candidate functions in the sparse regression algorithm can be physically informed by conducting the least-squares parameter fitting of the intercepted nonlinear stiffness force trajectories. Numerical investigations demonstrate that the proposed method not only gives physics information but also improves the accuracy by 0.48%, 3.26% and 22.21% under the noise level of 30 dB, 20 dB, and 10 dB, respectively. Moreover, the reconstructed dynamic response has a good agreement with the theory. Experimental measurements are performed on a magnetically coupled bistable nonlinear energy sink. Results show that the accuracy improves by 4.52% and 11.76% compared to restoring force surface and Hilbert transform-based methods, respectively.

SenseNet: A Physics-Informed Deep Learning Model for Shape Sensing

Shape sensing is an emerging technique for the reconstruction of deformed shapes using data from a discrete network of strain sensors. The prominence is due to its suitability in promising applications such as structural health monitoring in multiple engineering fields and shape capturing in the medical field. In this work, a physics-informed deep learning model, named SenseNet, was developed for shape sensing applications. Unlike existing neural network approaches for shape sensing, SenseNet incorporates the knowledge of the physics of the problem, so its performance does not rely on the choices of the training data. Compared with numerical physics-based approaches, SenseNet is a mesh-free method, and therefore it offers convenience to problems with complex geometries. SenseNet is composed of two parts: a neural network to predict displacements at the given input coordinates, and a physics part to compute the loss using a function incorporated with physics information. The prior knowledge considered in the loss function includes the boundary conditions and physics relations such as the strain–displacement relation, material constitutive equation, and the governing equation obtained from the law of balance of linear momentum. SenseNet was validated with finite-element solutions for cases with nonlinear displacement fields and stress fields using bending and fixed tension tests, respectively, in both two and three dimensions. A study of the sensor density effects illustrated the fact that the accuracy of the model can be improved using a larger amount of strain data. Because general three dimensional governing equations are incorporated in the model, it was found that SenseNet is capable of reconstructing deformations in volumes with reasonable accuracy using just the surface strain data. Hence, unlike most existing models, SenseNet is not specialized for certain types of elements, and can be extended universally for even thick-body applications.

Differentiable physics-enabled closure modeling for Burgers’ turbulence

Data-driven turbulence modeling is experiencing a surge in interest following algorithmic and hardware developments in the data sciences. We discuss an approach using the differentiable physics paradigm that combines known physics with machine learning to develop closure models for Burgers’ turbulence. We consider the one-dimensional Burgers system as a prototypical test problem for modeling the unresolved terms in advection-dominated turbulence problems. We train a series of models that incorporate varying degrees of physical assumptions on an a posteriori loss function to test the efficacy of models across a range of system parameters, including viscosity, time, and grid resolution. We find that constraining models with inductive biases in the form of partial differential equations that contain known physics or existing closure approaches produces highly data-efficient, accurate, and generalizable models, outperforming state-of-the-art baselines. Addition of structure in the form of physics information also brings a level of interpretability to the models, potentially offering a stepping stone to the future of closure modeling.

Classifacation of modern superconductors and their theory based on through experiments conducted on Cuprates

In this article, the physical nature of the phenomenon of superconductivity, related theoretical information and concepts of modern physics regarding the classification of superconductors are covered. The practical importance of the laws of superconductivity physics determined in cuprate superconductors is explained on the basis of the latest scientific innovations.

Guest Editorial: Solar Energy Utilization

Guest Editorial: Solar Energy Utilization

Physics-Informed AI Surrogates for Day-Ahead Wind Power Probabilistic Forecasting with Incomplete Data for Smart Grid in Smart Cities

Due to the high inherent uncertainty of renewable energy, probabilistic day-ahead wind power forecasting is crucial for modeling and controlling the uncertainty of renewable energy smart grids in smart cities. However, the accuracy and reliability of high-resolution day-ahead wind power forecasting are constrained by unreliable local weather prediction and incomplete power generation data. This article proposes a physics-informed artificial intelligence (AI) surrogates method to augment the incomplete dataset and quantify its uncertainty to improve wind power forecasting performance. The incomplete dataset, built with numerical weather prediction data, historical wind power generation, and weather factors data, is augmented based on generative adversarial networks. After augmentation, the enriched data is then fed into a multiple AI surrogates model constructed by two extreme learning machine networks to train the forecasting model for wind power. Therefore, the forecasting models’ accuracy and generalization ability are improved by mining the implicit physics information from the incomplete dataset. An incomplete dataset gathered from a wind farm in North China, containing only 15 days of weather and wind power generation data with missing points caused by occasional shutdowns, is utilized to verify the proposed method’s performance. Compared with other probabilistic forecasting methods, the proposed method shows better accuracy and probabilistic performance on the same incomplete dataset, which highlights its potential for more flexible and sensitive maintenance of smart grids in smart cities.

Physics-Driven Deep Learning for Computational Magnetic Resonance Imaging: Combining physics and machine learning for improved medical imaging.

Physics-driven deep learning methods have emerged as a powerful tool for computational magnetic resonance imaging (MRI) problems, pushing reconstruction performance to new limits. This article provides an overview of the recent developments in incorporating physics information into learning-based MRI reconstruction. We consider inverse problems with both linear and non-linear forward models for computational MRI, and review the classical approaches for solving these. We then focus on physics-driven deep learning approaches, covering physics-driven loss functions, plug-and-play methods, generative models, and unrolled networks. We highlight domain-specific challenges such as real- and complex-valued building blocks of neural networks, and translational applications in MRI with linear and non-linear forward models. Finally, we discuss common issues and open challenges, and draw connections to the importance of physics-driven learning when combined with other downstream tasks in the medical imaging pipeline.

Measuring unsteady drag of the flow around a sphere based on time series displacement measurements using physics-informed neural networks

Surface drag increases energy usage and emissions in numerous transport and engineering applications, with surface treatment being one technique aimed at reducing surface skin friction and hence, surface drag. One approach to investigate the performance of surface treatments is based on the measurement of the terminal velocities of settling spheres with and without surface treatment. This relatively affordable method is limited to the terminal velocity region, as the analysis of the data during the accelerating phase of the sphere requires the differentiation of the experimental data, which results in significant noise and drag measurements with high uncertainty. This paper compares two neural network approaches to determine the temporal derivatives of experimental data. Specifically, the ensemble averaged experimental data and physics information is used to define a model for a settling sphere, which is used to test the performance of neural networks trained on noisy displacement data using an approach based on the governing differential equation and a direct approach, with varying loss term weights, measurement error levels and temporal resolutions. The results show that the neural network trained directly on the experimental displacement data is able to reduce the uncertainty of the temporal derivatives to a level similar to that of the differential equation approach when initial conditions are included in the training. Finally, the two approaches are applied to experimental data to determine the unsteady drag. This paper demonstrates that neural networks can be used to determine the temporal derivatives of experimental data with relatively low uncertainty, particularly where the system physics is at least partially known.