• Geophysics and Geophysical Exploration
• /
• v.26 no.3
• /
• pp.114-125
• /
• 2023

The physics-informed neural network (PINN) has been proposed to overcome the limitations of various numerical methods used to solve partial differential equations (PDEs) and the drawbacks of purely data-driven machine learning. The PINN directly applies PDEs to the construction of the loss function, introducing physical constraints to machine learning training. This technique can also be applied to wave equation modeling. However, to solve the wave equation using the PINN, second-order differentiations with respect to input data must be performed during neural network training, and the resulting wavefields contain complex dynamical phenomena, requiring careful strategies. This tutorial elucidates the fundamental concepts of the PINN and discusses considerations for wave equation modeling using the PINN approach. These considerations include spatial coordinate normalization, the selection of activation functions, and strategies for incorporating physics loss. Our experimental results demonstrated that normalizing the spatial coordinates of the training data leads to a more accurate reflection of initial conditions in neural network training for wave equation modeling. Furthermore, the characteristics of various functions were compared to select an appropriate activation function for wavefield prediction using neural networks. These comparisons focused on their differentiation with respect to input data and their convergence properties. Finally, the results of two scenarios for incorporating physics loss into the loss function during neural network training were compared. Through numerical experiments, a curriculum-based learning strategy, applying physics loss after the initial training steps, was more effective than utilizing physics loss from the early training steps. In addition, the effectiveness of the PINN technique was confirmed by comparing these results with those of training without any use of physics loss.

• Current Optics and Photonics
• /
• v.1 no.2
• /
• pp.143-149
• /
• 2017

LED arrays with pixel numbers of $3{\times}3$, $4{\times}4$, and $5{\times}5$ have been studied in this paper in order to enhance the optical output power and decrease heat dissipation of an AlGaInP-based light emitting diode display device (pixel size of $280{\times}280{\mu}m$) fabricated by micro-opto-electro-mechanical systems. Simulation results showed that the thermal resistances of the $3{\times}3$, $4{\times}4$, $5{\times}5$ arrays were $52^{\circ}C/W$, $69.7^{\circ}C/W$, and $84.3^{\circ}C/W$. The junction temperature was calculated by the peak wavelength shift method, which showed that the maximum value appears at the center pixel due to thermal crosstalk from neighboring pixels. The central temperature would be minimized with $40{\mu}m$ pixel pitch and $150{\mu}m$ substrate thickness as calculated by thermal modeling using finite element analysis. The modeling can be used to optimize parameters of highly integrated AlGaInP-based LED arrays fabricated by micro-opto-electro-mechanical systems technology.

• Nuclear Engineering and Technology
• /
• v.51 no.4
• /
• pp.968-976
• /
• 2019

This paper presents a number of verification case studies for a recently developed sensitivity/uncertainty code package. The code package, ROMUSE (Reduced Order Modeling based Uncertainty/Sensitivity Estimator) is an effort to provide an analysis tool to be used in conjunction with reactor core simulators, in particular the Virtual Environment for Reactor Applications (VERA) core simulator. ROMUSE has been written in C++ and is currently capable of performing various types of parameter perturbations and associated sensitivity analysis, uncertainty quantification, surrogate model construction and subspace analysis. The current version 2.0 has the capability to interface with the Design Analysis Kit for Optimization and Terascale Applications (DAKOTA) code, which gives ROMUSE access to the various algorithms implemented within DAKOTA, most importantly model calibration. The verification study is performed via two basic problems and two reactor physics models. The first problem is used to verify the ROMUSE single physics gradient-based range finding algorithm capability using an abstract quadratic model. The second problem is the Brusselator problem, which is a coupled problem representative of multi-physics problems. This problem is used to test the capability of constructing surrogates via ROMUSE-DAKOTA. Finally, light water reactor pin cell and sodium-cooled fast reactor fuel assembly problems are simulated via SCALE 6.1 to test ROMUSE capability for uncertainty quantification and sensitivity analysis purposes.

• Nuclear Engineering and Technology
• /
• v.55 no.9
• /
• pp.3409-3416
• /
• 2023

Licensing the next-generation of nuclear reactor designs requires extensive use of Modeling and Simulation (M&S) to investigate system response to many operational conditions, identify possible accidental scenarios and predict their evolution to undesirable consequences that are to be prevented or mitigated via the deployment of adequate safety barriers. Deep Learning (DL) and Artificial Intelligence (AI) can support M&S computationally by providing surrogates of the complex multi-physics high-fidelity models used for design. However, DL and AI are, generally, low-fidelity 'black-box' models that do not assure any structure based on physical laws and constraints, and may, thus, lack interpretability and accuracy of the results. This poses limitations on their credibility and doubts about their adoption for the safety assessment and licensing of novel reactor designs. In this regard, Physics Informed Neural Networks (PINNs) are receiving growing attention for their ability to integrate fundamental physics laws and domain knowledge in the neural networks, thus assuring credible generalization capabilities and credible predictions. This paper presents the use of PINNs as surrogate models for accidental scenarios simulation in Nuclear Power Plants (NPPs). A case study of a Loss of Heat Sink (LOHS) accidental scenario in a Nuclear Battery (NB), a unique class of transportable, plug-and-play microreactors, is considered. A PINN is developed and compared with a Deep Neural Network (DNN). The results show the advantages of PINNs in providing accurate solutions, avoiding overfitting, underfitting and intrinsically ensuring physics-consistent results.

• Geophysics and Geophysical Exploration
• /
• v.26 no.4
• /
• pp.268-274
• /
• 2023

Finite source inversion is performed with a Green's function matrix and geodetic coseismic displacement. Conventionally, the Green's function matrix is constructed using the Okada model (Okada, 1985). However, for more realistic earthquake simulations, recent research has widely adopted the physics-based model, which can consider various material properties such as elasticity, viscoelasticity, and elastoplasticity. We used the physics-based software PyLith, which is suitable for earthquake modeling. However, the PyLith does not provide a mesh generator, which makes it difficult to perform finite source inversions that require numerous subfaults and observation points within the model. Therefore, in this study, we developed CPInterface (COMSOL-PyLith Interface) to improve the convenience of finite source inversion by combining the processes of creating a numerical model including sub-faults and observation points, simulating earthquake modeling, and constructing a Green's function matrix. CPInterface combines the grid generator of COMSOL with PyLith to generate the Green's function matrix automatically. CPInterface controls model and fault information with simple parameters. In addition, elastic subsurface anomalies and GPS observations can be placed flexibly in the model. CPInterface is expected to enhance the accessibility of physics-based finite source inversions by automatically generating the Green's function matrix.

• Current Optics and Photonics
• /
• v.4 no.6
• /
• pp.509-515
• /
• 2020

A practical single photon source for fiber-based quantum information processing is still lacking. As a possible 1.55-㎛ quantum-dot single photon source, an InGaAsP/InP-air-aperture micropillar cavity is investigated in terms of fabrication tolerance. By properly modeling the processing uncertainty in layer thickness, layer diameter, surface roughness and the cavity shape distortion, the fabrication imperfection effects on the cavity quality are simulated using a finite-difference time-domain method. It turns out that, the cavity quality is not significantly changing with the processing precision, indicating the robustness against the imperfection of the fabrication processing. Under thickness error of ±2 nm, diameter uncertainty of ±2%, surface roughness of ±2.5 nm, and sidewall inclination of 0.5°, which are all readily available in current material and device fabrication techniques, the cavity quality remains good enough to form highly efficient and coherent 1.55-㎛ single photon sources. It is thus implied that a quantum dot contained InGaAsP/InP-air-aperture micropillar cavity is prospectively a practical candidate for single photon sources applied in a fiber-based quantum information network.

• 한국HCI학회:학술대회논문집
• /
• 2007.02a
• /
• pp.612-615
• /
• 2007

This research studies the Virtual Reality simulation of Newton's physics law on rigid body type of objects for physics learning. With network support, collaborative interaction is enabled so that people from different places can interact with the same set of objects in Collaborative Virtual Environment. The taxonomy of the interaction in different levels of collaboration is described as: distinct objects and same object, in which there are same object - sequentially, same object - concurrently - same attribute, and same object - concurrently - distinct attributes. The case studies are the interaction of users in two cases: destroying and creating a set of arranged rigid bodies. We identify a specific type of application for contents authoring with modeling systems integrated with real-time physics and implemented in VR system. In our application called Virtual Dollhouse, users can observe physics law while constructing a dollhouse using existing building blocks, under gravity effects.

• International Journal of CAD/CAM
• /
• v.9 no.1
• /
• pp.93-102
• /
• 2010

Shaping realistic hairstyles for digital characters is a difficult, long and tedious task. The lack of appropriate interaction metaphors enabling efficient and simple, yet accurate hair modeling further aggravates the situation. This paper presents 3D interaction metaphors for modeling virtual hair using haptic interfaces. We discuss user tasks, ergonomic aspects, as well as haptics-based styling and fine-tuning tools on an experimental prototype. In order to achieve faster haptic rates with respect to the hair simulation and obtain a transparent rendering, we adapt our simulation models to comply with the specific requirements of haptic hairstyling actions and decouple the simulation of the hair strand dynamics from the haptic rendering while relying on the same physiochemical hair constants. Besides the direct use of the discussed interaction metaphors in the 3D modeling area, the presented results have further application potential in hair modeling facilities for the entertainment industry and the cosmetic sciences.

• Nuclear Engineering and Technology
• /
• v.56 no.2
• /
• pp.728-744
• /
• 2024

The presence of hydrogen absorbed by zirconium-based cladding materials during reactor operation can trigger degradation mechanisms and endanger the rod integrity. Ensuring the durability of the rods in extended time-frames like dry storage requires anticipating hydrogen behavior using numerical modeling. In this context, the present paper describes a hydrogen post-processing tool for Falcon - HYPE, a PSI's in-house tool able to calculate hydrogen uptake, transport, thermochemistry, reorientation of hydrides and hydrogen-related failure criteria. The tool extracts all necessary data from a Falcon output file; therefore, it can be considered loosely coupled to Falcon. HYPE has been successfully validated against experimental data and applied to reactor operation and interim storage scenarios to present its capabilities.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.59 no.11
• /
• pp.2012-2015
• /
• 2010

In this paper, the mathematical models of the superheater and the desuperheater are derived based on the fundamental laws of physics, mass and energy balance. The parameters of the models are developed for the 500[MW] thermal power plant using the actual data. The simulated model outputs are well matched with the actual ones. It is expected that the proposed models are useful for the temperature controller design of the thermal power plant.