표제지
요약문
목차
1. 서론 11
2. 이론 17
2.1. 수동 탄성파 탐사 17
2.1.1. 수동 탄성파 탐사 자료 취득 17
2.1.2. 탄성파 간섭법 18
2.2. 표면파 분산 특성 23
2.2.1. 표면파와 고유 값 문제 23
2.2.2. 분산관계 31
2.3. MASW 34
2.3.1. 자료취득 34
2.3.2. 분산 곡선 구성 41
2.3.3. S 파 속도 역산 44
2.4. CNN 기반 기계학습 46
2.4.1. 기계학습 46
2.4.2. CNN+MLP 모델 47
3. 수치예제 50
3.1. 합성자료를 이용한 기계학습 모델 훈련 50
3.1.1. 합성 데이터 생성 50
3.1.2. 기계학습 모델 훈련 55
3.2. 현장자료 적용 64
3.2.1. 지중환경연구 테스트 베드 64
3.2.2. 지중환경연구 테스트 베드 현장자료 적용 결과 68
3.2.3. 포항자료 72
3.2.4. 포항 현장 자료 적용 결과 74
4. 결론 및 토의 76
5. 참고문헌 79
Abstract 88
Figure 1. Example of 1D interferometry: Two green functions recorded at locations A and B. 19
Figure 2. Comparing seismic computation of seismic interferometry by modelling results with same velocity model; (a) passive seismic data, (d) active seismic data, (b) apply seismic interferometry to... 22
Figure 3. Applying spatial windows to compose 2D section; (a) results of gaussian function application to each CSG for location 120 m, (b)... 39
Figure 4. Machine learning algorithm diagram 49
Figure 5. FEM(Finite Element Modeling) for synthetic data with (a) S wave velocity models, resulting (b) seismograms 52
Figure 6. Examples of synthetic dispersion spectrum. 54
Figure 7. Loss function according to epoch: training loss(yellow) and validation loss(red) 56
Figure 8. Comparison inversion results from machine learning application(orange) with true models(blue) 58
Figure 9. Comparison of result varying with smoothing filter parameter; (a) estimated S wave velocity without filtering, (b) over-filtered... 60
Figure 10. Average relative error of (a) velocity estimated by machine learning with training data and (b) bilateral-filtered estimated velocity 63
Figure 11. Information of test bed: (a) Geography and (b) simplified acquisition line 65
Figure 12. Processing of (a) active seismic real data (data acquisition, 2D spectral windows application and phase-shift), and... 67
Figure 13. (a) Boring logs obtained from middle of the survey line (b) Estimated S wave velocity model from active seismic data and (b) Estimated S wave velocity model from passive seismic data 69
Figure 14. Comparison of average S wave velocity along the location of result estimated by active data (blue) and 1D S wave velocity... 70
Figure 15. Processing of (a) passive seismic real data (data acquisition, cross-coherence computation, and phase-shift), and (b) active... 73
Figure 16. S wave velocity estimated by passive data (orange) and S wave velocity estimated by passive data (blue). 75