APro, developed in KAERI for the process-based total system performance assessment (TSPA) of deep geological disposal systems, performs finite element method (FEM)-based multiphysics analysis. In the FEM-based analysis, the mesh element quality influences the numerical solution accuracy, memory requirement, and computation time. Therefore, an appropriate mesh structure should be constructed before the mesh stability analysis to achieve an accurate and efficient process-based TSPA. A generic reference case of DECOVALEX-2023 Task F, which has been proposed for simulating stationary groundwater flow and time-dependent conservative transport of two tracers, was used in this study for mesh stability analysis. The relative differences in tracer concentration varying mesh structures were determined by comparing with the results for the finest mesh structure. For calculation efficiency, the memory requirements and computation time were compared. Based on the mesh stability analysis, an approach based on adaptive mesh refinement was developed to resolve the error in the early stage of the simulation time-period. It was observed that the relative difference in the tracer concentration significantly decreased with high calculation efficiency.

The most important thing in development of a process-based TSPA (Total System Performance Assessment) tool for large-scale disposal systems (like APro) is to use efficient numerical analysis methods for the large-scale problems. When analyzing the borehole in which the most diverse physical phenomena occur in connection with each other, the finest mesh in the system is applied to increase the analysis accuracy. Since thousands of such boreholes would be placed in the future disposal system, the numerical analysis for the system becomes significantly slower, or even impossible due to the memory problem in cases. In this study, we propose a tractable approach, so called global-local iterative analysis method, to solve the large-scale process-based TSPA problem numerically. The global-local iterative analysis method goes through the following process: 1) By applying a coarse mesh to the borehole area the size of the problem of global domain (entire disposal system) is reduced and the numerical analysis is performed for the global domain. 2) Solutions in previous step are used as a boundary condition of the problem of local domain (a unit space containing one borehole and little part of rock), the fine mesh is applied to the borehole area, and the numerical analysis is performed for each local domain. 3) Solutions in previous step are used as boundary conditions of boreholes in the problem of global domain and the numerical analysis is performed for the global domain. 4) steps 2) and 3) are repeated. The solution derived by the global-local iterative analysis method is expected to be closer to the solution derived by the numerical analysis of the global problem applying the fine mesh to boreholes. In addition, since local problems become independent problems the parallel computing can be introduced to increase calculation efficiency. This study analyzes the numerical error of the globallocal iterative analysis method and evaluates the number of iterations in which the solution satisfies the convergence criteria. And increasing computational efficiency from the parallel computing using HPC system is also analyzed.

APro, a modularized process-based total system performance assessment framework, was developed at the Korea Atomic Energy Research Institute (KAERI) to simulate radionuclide transport considering coupled thermal-hydraulic-mechanicalchemical processes occurring in a geological disposal system. For reactive transport simulation considering geochemical reactions, COMSOL and PHREEQC are coupled with MATLAB in APro using an operator splitting scheme. Conventionally, coupling is performed within a MATLAB interface so that COMSOL stops the calculation to deliver the solution to PHREEQC and restarts to continue the simulation after receiving the solution from PHREEQC at every time step. This is inefficient when the solution is frequently interchanged because restarting the simulation in COMSOL requires an unnecessary setup process. To overcome this issue, a coupling scheme that calls PHREEQC inside COMSOL was developed. In this technique, PHREEQC is called through the “MATLAB function” feature, and PHREEQC results are updated using the COMSOL “Pointwise Constraint” feature. For the one-dimensional advection-reaction-dispersion problem, the proposed coupling technique was verified by comparison with the conventional coupling technique, and it improved the computation time for all test cases. Specifically, the more frequent the link between COMSOL and PHREEQC, the more pronounced was the performance improvement using the proposed technique.

Various linear system solvers with multi-physics analysis schemes are compared focusing on the near-field region considering thermal-hydraulic-chemical (THC) coupled multi-physics phenomena. APro, developed at KAERI for total system performance assessment (TSPA), performs a finite element analysis with COMSOL, for which the various combinations of linear system solvers and multi-physics analysis schemes should to be compared. The KBS-3 type disposal system proposed by Sweden is set as the target system and the near-field region, which accounts for most of the computational burden is considered. For comparison of numerical analysis methods, the computing time and memory requirement are the main concerns and thus the simulation time is set up to one year. With a single deposition hole problem, PARDISO and GMRESSSOR are selected as representative direct and iterative solvers respectively. The performance of representative linear system solvers is then examined through a problem with an increasing number of deposition holes and the GMRES-SSOR solver with a segregated scheme shows the best performance with respect to the computing time and memory requirement. The results of the comparative analysis are expected to provide a good guideline to choose better numerical analysis methods for TSPA.

APro, a modularized framework of the process-based total system performance assessment, has been developed by KAERI to simulate the radionuclide transport in geological disposal system considering multi-physics phenomena. However, the target problem including more than 10,000 boreholes and over 100,000 years of simulation time is computationally challenging to deal with numerical solvers provided by COMSOL Multiphysics constituting APro. To alleviate the computational burden, machine learning (ML) techniques have been studied to develop a surrogate model replacing the heavy computation part. In recent studies, attempts have been made to integrate the knowledge of physics and numerical methods into the ML model for partial differential equations (PDEs). Unlike conventional ML approaches solely relying on data-driven method, the integration can help to make the ML model more specialized for solving PDEs. The hybrid neural network (NN) solver method is one of the strategies to develop more efficient PDE solver by interleaving NN with numerical solvers like finite element method (FEM). The hybrid NN model on the premise of numerical solver is easier to train and more stable than the purely data-driven model. For example, one previous study has used the hybrid NN model as a corrector for an incomplete numerical solver for the advection-diffusion problem. In every time step of simulation, NN corrects the error of incomplete solution obtained by a relaxed numerical solver with coarse meshing. The simulation in the next time step starts from the corrected solution, so NN interacts with the numerical solver iteratively. If the corrector is successfully trained, the incomplete but fast solver with corrector can provide reliable results comparable to the original massive solver. This study adopts the hybrid concept to develop a surrogate model for the near-field region, which is the heavy computation part in the simulation of geological disposal system. Various incomplete models such as coarse meshing or emptying the borehole domain are studied to construct a hybrid NN solver. This study also covers how to embed the hybrid NN in COMSOL Multiphysics to train and use it during the simulation.

Domain decomposition method (DDM) has been widely employed for the numerical analysis of large-scale problems due to its applicability to parallel computing. DDM divides the modeling domain into a set of subdomains and obtains the entire solution iteratively until the values of each subdomain which are shared with other subdomains, such as boundary values, are converged. Therefore, in general, DDM is a memory-efficient iterative algorithm with inherent parallelism on the geometric level. APro, the process-based total system performance assessment model, aims for simulating the radionuclide transport considering coupled multi-physics phenomena occurring in large-scale geological disposal system, which are inevitably accompanied by huge memory burden. Therefore, DDM is applicable for the large-scale problem of APro and its performance in parallel computing needs to be examined. The DDM solvers provided by COMSOL which constitute APro can be classified into two methods. One is the overlapping Schwarz method that each subdomain overlaps its neighboring domains and the other is the Schur complement method that subdomains are non-overlapping and separated by boundary domains. For the Schwarz method, the additive, hybrid, multiplicative and symmetric methods can be selected according to the solution update scheme. And for the Schur method, the additive and multiplicative ordering options can be chosen for solving Schur complement system. In this study, the calculation efficiency of the DDM solvers in COMSOL and the applicability to the cluster environment were examined. In aspect of efficiency, the memory requirements with different number of subdomains and calculation schemes were compared in a single node. Then, the memory requirements with increasing number of disposal tunnels and deposition holes were investigated in multiple nodes. As a result, on the cluster environment, with the help of distributed memory architecture which enables efficient memory usage, the applicability of DDM solvers to the large-scale problem of APro was confirmed.

APro, developed by KAERI as a process-based total system performance assessment model, can simulate the radionuclide transport affected by thermal, hydraulic, mechanical and geochemical changes that may occurs in the engineering and natural barriers of a geological disposal system. APro targets a large-scale and heterogeneous 3D system that includes more than 10,000 boreholes located about 500 m underground and hundreds of fractures of different sizes distributed within an area of several km2. Simulating transport and reaction phenomena for such a system through the global implicit approach (GIA) may require considerable computational resources or be intractable in some cases. Therefore, APro adopts the sequential non-iterative approach (SNIA), one of the operator splitting (OS) methods, to separate the mass transport and reaction phenomena into independent problems. By using SNIA, the parallel computation performance in APro with multiple cores is expected to be improved. In this study, the effect of SNIA on the parallel computation performance was analyzed through a simple 1D reactive transport problem. Without SNIA, finite difference equations, discretized from the partial differential equations (PDEs) describing the reactive transport problem, have to be solved at once because all dependent variables are nonlinearly and spatially interconnected through reaction and mass transport terms. When the reaction and mass transport terms are separated through SNIA, the mass transport problem can be converted into independent linear equations for each chemical and the efficient linear system solver can be applied to each linear equation. In particular, since the reaction problem is changed to independent nonlinear equations for each node, the parallel computation performance can be greatly improved. To verify this, the 1D reactive transport problem was implemented in MATLAB, and SNIA and GIA were applied to solve the problem. As a result, there was no significant difference in results between SNIA and GIA for proper spatial and temporal discretization, which verified the accuracy of SNIA. In order to see the parallel computation performance, the calculation times for SNIA and GIA with increasing number of cores were measured and compared. As the number of cores increased, the SNIA calculation speed became faster than that of GIA, which verified that SNIA could improve parallel computation performance in APro. In the future, the effect of SNIA on the parallel computation performance will be verified for the numerical analysis of large-scale geological disposal systems.

2017년 국내 육성 사과 ‘아리수’ 품종에서 노린재류에 의한 피해와 유사한 반점 증상이 확인되었다. 사과 과실 전면에 발생하는 이 반점 증상은 크기가 2∼4 mm이며 진한 갈색, 짙은 고동색으로 반점 중심부가 약간 함몰되는 형태를 가진다. 이 연구는 이러한 반점증상의 원인이 노린재에 의한 피해인지 여부를 알기 위해 수행하였다. 노린재류는 접종 전 채집하였고, 6월 초순부터 ‘아리수’ 사과에 약 15일 간격으로 썩덩나무노린재와 갈색날개노린재를 각각 접종하였다. ‘아리수’ 사과의 수확기인 9월 초순까지 총 5회(6월 상순, 7월 상순, 7월 중순, 8월 상순, 8월 중순) 접종하였으며, 접종후 접종시기별로 피해 양상을 확인하였다. 6월 상순에 노린재에 의해 가해 받은 과실은 비대하지 못하거나 낙과하였다. 다른 시기에 노린재에 의해 가해 받은 접종 1~2일후 흡즙구멍을 확인할 수 있었고, 5일 후부터는 가해 받은 부위 주변이 붉게 착색되는 halo 증상이 나타났다. 노린재류에 의한 ‘아리수’ 사과의 피해양상은, 기 보고된 ‘후지’ 품종에서의 피해양상과 거의 유사하였다. 실험결과 2017년 ‘아리수’ 사과에 발생한 반점증상은 노린재에 의한 것이 아님을 확인할 수 있었다.

This paper proposes an integrated positioning system to localize a moving object in the shadow-area that exists in the water tank. The new water tank for underwater robots is constructed to evaluate the navigation performance of underwater vehicles. Several sensors are integrated in the water tank to provide the position information of the underwater vehicles. However there are some areas where the vehicle localization becomes very poor since the very limited sensors such as sonar and depth sensors are effective in underwater environment. Also there are many disturbances at sonar data. To reduce these disturbances, an extended Kalman filter has been adopted in this research. To localize the underwater vehicles under the hostile situations, a SVR (Support Vector Regression) has been systematically applied for estimating the position stochastically. To demonstrate the performance of the proposed algorithm (an extended Kalman filter + SVR analysis), a new UI (User Interface) has been developed.