리눅스 컨테이너와 버전 관리 시스템을 이용한 소프트웨어 연구 환경 구축

하완수 ( Wansoo Ha ) 한국지구물리·물리탐사학회 2021 지구물리와 물리탐사 Vol.24 No.2

With advancements in software technology, more scientists and engineers are employing computer software and programming tools for research. However, several issues can arise in software-based research: environment setting, reproducibility, and loss of source codes. This study investigates the use of Linux containers and version control systems to prevent these problems. Managing research projects using a cloud source-code repository and building a research environment in a Linux container can prevent the abovementioned problems and make research collaboration easier. For researchers with no experience with Linux containers, a repository of project template containing shell scripts for building and running containers has been released.

송신원 추정을 통합한 라플라스 영역 지도학습 기반 탄성파 역산

조준현,하완수,Jun Hyeon Jo,Wansoo Ha 한국지구물리·물리탐사학회 2025 물과 미래(한국수자원학회지) Vol.58 No.5

전통적인 전파형 역산의 한계를 극복하기 위해 제안된 딥러닝 기반 탄성파 역산 기술들은 인공 합성 자료들을 대상으로 우수한 역산 성능을 입증하였다. 그러나, 송신원 추정 문제를 다루는 딥러닝 기반 탄성파 역산 연구는 거의 찾아볼 수 없다. 특히 현장 탐사 자료의 성공적인 역산을 위해서는 정확한 송신 파형을 사용하는 것이 중요하다. 역산에 사용되는 송신원의 파형이 실제 송신원과 크게 다를 경우 역산 결과가 참 속도 모델에 수렴하지 못하거나 계산 비용이 증가할 수 있다. 따라서, 본 연구는 딥러닝 기반 탄성파 역산에서 송신 파형의 불확실성이 역산 정확도에 미치는 영향을 분석하기 위해 수행하였다. 특히, 훈련 자료에 민감한 지도학습 기반 탄성파 역산에 초점을 맞추었으며, 라플라스 영역에서의 송신원 추정을 전처리 단계로 통합하였다. 시간이나 주파수 영역과 달리 라플라스 영역에서 송신 파형은 진폭 정보만을 가지므로 송신원 추정이 보다 단순하고 계산 효율적이다. 송신원 추정 알고리즘은 기존 라플라스 영역 완전 파형 역산에서 사용하는 것과 동일한 알고리즘을 채택하였으며, 뉴턴법을 이용하여 송신원 매개변수를 반복적으로 갱신하였다. 또한, 라플라스 영역 딥러닝 기반 탄성파 역산을 위한 새로운 심층 신경망을 제안하여 선행 연구와 정량적으로 비교 분석하였다. 제안된 신경망을 이용해 라플라스 영역 역산에서 더욱 효율적이고 우수한 역산 성능을 달성하였으며, 두 가지 벤치마크 모델을 이용한 수치 예제를 통해 송신원 추정의 중요성과 제안된 접근법의 효과성을 입증하였다. Deep learning-based seismic inversion techniques have been proposed to overcome the limitations of conventional full-waveform inversion (FWI) and have demonstrated excellent inversion performance on synthetic datasets. However, studies addressing the source estimation challenges in deep learning-based seismic inversion remain scarce. Accurate estimation of the source wavelet is particularly critical for the successful inversion of field seismic data. If the wavelet used deviates significantly from the true source wavelet, the inversion may fail to converge to the true velocity model or incur increased computational costs. Therefore, this study investigates how uncertainties in source wavelet estimation affect the accuracy of the deep learning-based seismic inversion, with a particular focus on supervised learning approaches, which are inherently sensitive to training data. To address this issue, source estimation in the Laplace domain is incorporated as a preprocessing step. Unlike the time or frequency domains, the source wavelet in the Laplace domain retains only the amplitude information, simplifying the estimation process and improving computational efficiency. The source estimation algorithm adopts the same approach as used in conventional Laplace-domain FWI, iteratively updating the source parameters using the Newton method. Furthermore, we propose a novel deep neural network for deep-learning-based seismic inversion in the Laplace-domain and conduct a quantitative comparison with the previous study. The proposed network achieved improved efficiency and superior inversion performance in the Laplace-domain inversion. Numerical examples using two benchmark models demonstrate the importance of source wavelet estimation and the effectiveness of the proposed approach.

제온 파이 보조 프로세서를 이용한 3차원 주파수 영역 음향파 파동 전파 모델링 병렬화

류동현,조상훈,하완수,Ryu, Donghyun,Jo, Sang Hoon,Ha, Wansoo 한국지구물리·물리탐사학회 2017 지구물리와 물리탐사 Vol.20 No.3

3D seismic data processing methods such as full waveform inversion or reverse-time migration require 3D wave propagation modeling and heavy calculations. We compared efficiency and accuracy of a Xeon Phi coprocessor to those of a high-end server CPU using 3D frequency-domain wave propagation modeling. We adopted the OpenMP parallel programming to the time-domain finite difference algorithm by considering the characteristics of the Xeon Phi coprocessors. We applied the Fourier transform using a running-integration to obtain the frequency-domain wavefield. A numerical test on frequency-domain wavefield modeling was performed using the 3D SEG/EAGE salt velocity model. Consequently, we could obtain an accurate frequency-domain wavefield and attain a 1.44x speedup using the Xeon Phi coprocessor compared to the CPU.

심층 신경망을 이용한 탄성파 속도 모델 구축 사례 분석

조준현 ( Jun Hyeon Jo ),하완수 ( Wansoo Ha ) 한국지구물리·물리탐사학회 2021 지구물리와 물리탐사 Vol.24 No.2

Velocity model building is an essential procedure in seismic data processing. Conventional techniques, such as traveltime tomography or velocity analysis take longer computational time to predict a single velocity model and the quality of the inversion results is highly dependent on human expertise. Full-waveform inversions also depend on an accurate initial model. Recently, deep neural network techniques are gaining widespread acceptance due to an increase in their integration to solving complex and nonlinear problems. This study investigated cases of seismic velocity model building using deep neural network techniques by classifying items according to the neural networks used in each study. We also included cases of generating training synthetic velocity models. Deep neural networks automatically optimize model parameters by training neural networks from large amounts of data. Thus, less human interaction is involved in the quality of the inversion results compared to that of conventional techniques and the computational cost of predicting a single velocity model after training is negligible. Additionally, unlike full-waveform inversions, the initial velocity model is not required. Several studies have demonstrated that deep neural network techniques achieve outstanding performance not only in computational cost but also in inversion results. Based on the research results, we analyzed and discussed the characteristics of deep neural network techniques for building velocity models.

Fault development pattern analysis with structural seismic attribute analysis of Exmouth sub-basin, Australia

우주환(Juhwan Woo),장성형(Seonghyung Jang),하완수(Wansoo Ha) 대한지질학회 2021 지질학회지 Vol.57 No.6

그래픽 프로세서를 이용한 시간 영역 3차원 파동 전파 모델링과 메모리 관리

김아름 ( Ahreum Kim ),류동현 ( Donghyun Ryu ),하완수 ( Wansoo Ha ) 한국지구물리·물리탐사학회 2016 지구물리와 물리탐사 Vol.19 No.3

We used graphics processing units for an efficient time-domain 3D wave propagation modeling. Since graphics processing units are designed for massively parallel processes, we need to optimize the calculation and memory management to fully exploit graphics processing units. We focused on the memory management and examined the performance of programs with respect to the memory management methods. We also tested the effects of memory transfer on the performance of the program by varying the order of finite difference equation and the size of velocity models. The results show that the memory transfer takes a larger portion of the running time than that of the finite difference calculation in programs transferring whole 3D wavefield.

OpenACC와 GPU를 이용한 3차원 파동 전파 모델링

김아름 ( Ahreum Kim ),이종우 ( Jongwoo Lee ),하완수 ( Wansoo Ha ) 한국지구물리·물리탐사학회 2017 지구물리와 물리탐사 Vol.20 No.2

We calculated 3D frequency- and Laplace-domain wavefields using time-domain modeling and Fourier transform or Laplace transform. We adopted OpenACC and GPU for an efficient parallel calculation. The OpenACC makes it easy to use GPU accelerators by adding directives in conventional C, C++, and Fortran programming languages. Accordingly, one doesn`t have to learn new GPGPU programming languages such as CUDA or OpenCL to use GPU. An OpenACC program allocates GPU memory, transfers data between the host CPU and GPU devices and performs GPU operations automatically or following user-defined directives. We compared performance of 3D wave propagation modeling programs using OpenACC and GPU to that using single-core CPU through numerical tests. Results using a homogeneous model and the SEG/EAGE salt model show that the OpenACC programs are approximately 53 and 30 times faster than those using single-core CPU.

코어레이와 MPI를 이용한 병렬 파동 전파 모델링과 거꿀 참반사 보정 성능 비교

류동현 ( Donghyun Ryu ),김아름 ( Ahreum Kim ),하완수 ( Wansoo Ha ) 한국지구물리·물리탐사학회 2016 지구물리와 물리탐사 Vol.19 No.3

Coarray is a parallel processing technique introduced in the Fortran 2008 standard. Coarray can implement parallel processing using simple syntax. In this research, we examined applicability of Coarray to seismic parallel processing by comparing performance of seismic data processing programs using Coarray and MPI. We compared calculation time using seismic wave propagation modeling and one to one communication time using domain decomposition technique. We also compared performance of parallel reverse-time migration programs using Coarray and MPI. Test results show that the computing speed of Coarray method is similar to that of MPI. On the other hand, MPI has superior communication speed to that of Coarray.

그래픽 프로세서를 이용한 고차 유한 차분식 기반 수중채널모델 연구

배호석(Ho Seuk Bae),김원기(Won-Ki Kim),손수욱(Su-Uk Son),하완수(Wansoo Ha) 한국시뮬레이션학회 2021 한국시뮬레이션학회 논문지 Vol.30 No.1