1. Overview of Geophysical Research Research
Who: 7 PhDs
3 Post Docs
1 Res. Scientist
1 Visit. Prof.
GTS
What: Develop
Improved Seismic
Imaging Algorithms,
Solve Practical
Problems

2. Ideal Seismic Section c Given: d(x,y,t) = Lc Find: c = L-1 d Depth
velocity
Given: d(x,y,t) = Lc
Find: c = L-1 d c 2 1
Time
Depth

3. 5 oil+processing companies
Geophysical Research
Science
Earthquake
CSIM Consortium
5 oil+processing companies
Engineering
Exploration

4. Geophysical Research Productivity
7 PhDs awarded since 2009
2017: 6 internships in USA
and Aramco
5 KAUSTAramco employee
CSIM Consoritum: Aramco,
Chevron, TGS, CGG, TOTAL
Number Papers Published
In early years, through the joint project with CSIM on 3D Full Waveform Inversion (FWI) and the software it produced, Aramco has built the foundation for software development on FWI. Through constant communications and discussions with CSIM on industrial challenging problems, Aramco gained great benefits of generating innovative ideas and algorithms.  
Dr. Tong Fei (senior Aramco scientist)
Year
Abdullah AlTheyab: Best Student Presentation SEG FWI Workshop Bejing 2015
Sherif Hanafy: 2017 SEG Honorary Lecturer for Middle East & Africa
GTS: 2013 SEG Distinguished Lecturer & 2010 Kauffman Gold Medal

5. Our FWI Contributions Summarized
2017
Volume editor Yonghe Sun wrote: “The book is a comprehensive and up-to-date volume on geophysical inverse problems with applications to seismic imaging, migration velocity analysis, and full-waveform inversion….is a pioneer in many of the topics covered, including least-squares migration, migration deconvolution, traveltime-guided waveform inversion, and image-domain waveform inversion. The readers will benefit from the unique perspective of the author as both an outstanding researcher and educator.

6. 3D Multiscale Phase Inversion
of Seismic Data
Lei Fu
4 km
Given: d = Lc
Find: c = L-1 d

7. Seismic Inverse Problem
4 km
6 km
Given: dobs = Lc
Find: c
2 + w2c-2 )d = F (
Soln: min || Lc-dobs|| 2 Dd
1000x1000x500 FD model
10,000 traces/shot
50,000 simulations/iteration
20-50 iterations
100s terrabytes input data
 c(k+1) = c(k) + a[LTL]-1LT Dd

8. 3D SEG/EAGE Overthrust Model
a) True Model
b) Initial Model
nz=61;
ny=401;
nx=401;
dx=10;
Receiver #: 10000;
Shot #: 1600;
25 Hz Ricker
c) FWI Tomogram
d) MPI Tomogram

9. Vovle 3D OBC Data a) Initial Model b) FWI Tomogram c) MPI Tomogram
d) RTM Image Based on a)
e) RTM Image Based on b)
f) RTM Image Based on c)
Cable 2
Cable 2
Cable 2

10. Vovle 3D OBC Data RTM Image for Cable2 a) Initial b) FWI c) MPI 2 4 6
Depth (km) 2 4 6 12 6 12 6 12
X (km)
X (km)
X (km)

11. Vovle 3D OBC Data Magnified Views of RTM Image for Cable2
Zoom View of Box A
Zoom View of Box B
2.4
2.2
2.8
Depth (km)
Depth (km)
2.6
3.2
3.0 5 6 5 6 5 6 8 9 8 9 8 9
X (km)
X (km)

12. Two-anomaly Model a) True Model b) Initial Model c) FWI Tomogram
nz=41;
ny=81;
nx=81;
dx=10;
Receiver #: 1600;
Shot #: 400;
15 Hz Ricker
c) FWI Tomogram
d) MPI Tomogram

13. Problem: Local minima + Cycle Skipping in FWI
Motivation
Problem: Local minima + Cycle Skipping in FWI
Inaccurate initial velocity model;
Inaccurate amplitude modeling, attenuation medium (Q);
Elastic effects;
Noise;
Solution:
Multiscale phase inversion (MPI), less initial model and amplitude dependent, less affected by elastic effects, robust.

14. Vovle 3D OBC Data Survey Geometry Cable Location Shot Location Y (km)6
# 12
400 m
Shot Location
Y (km) 3
# 2
# 1 6 12
X (km)