The repositories store code that was written to cover the topics below. Some example results as animated gifs are used to illustrate the outputs.
One example of streamlines flowing over wind turbines in the Dublin Mountains. Other simulations and ML examples are given below.
Data Science: Exploiting high performance computing and data science techniques to process and image large
datasets to image the subsurface within the business constraints for maximising commercial value.
Examples include, i) using generative adversarial networks to translate complex geophysical images
into human interpretable products, ii) using genetic algorithms as optimisation tools within the
workflow. iii) Understanding the value of information arguments for assessing the business delivery of
workflows relative to the impact and cost across the E&P workflow. The ultimate objective is to use data to make
better business decisons through better de-risking or using quantifiable measures.
Predictive Analytic dashboard for sensor data streaming from an FPSO
Generative adversial neural networks deployed to clean seismic data for analysis
Neural operator simulating elastic seismic wave propagation in complex volcanic envirnoments. The inversion results can be interpreted as the network learning the inverse Green’s function for the elastic wave operator. The network can also be trained to generate the Green’s function for a particular model by using the models as the input and the wavefield (or seismic traces) as the output.
The top panel shows the results from a numerical simulation directly solving the equations with a sample from the unseen layered model data. The left-hand side is for a snapshot after 1 second and the right-hand panel is after 3 seconds. The middle panel shows the results from the FNO network using the layered dataset to train.
The bottom panels show the difference. The overall match is very good though the network can struggle to capture the smaller amplitude scattering. This failure to accurately predict the scattering is greater when using very heterogeneous training models.
SIR Model of a generic pandemic across the Irish Population - worse case scenario with large Ro and no social distancing measures.
Geophysical Imaging: The most recent tools for imaging are Reverse Time Migration (RTM) and Full Wavefield Inversion (FWI). RTM imaging has shown to be hugely beneficial in imaging in complex geological regions, e.g. subsalt, whilst FWI has been used to develop high resolution velocity models which leads to better seismic images, can be fed into quantitative interpretation, used to image through gas clouds and used in hazard prediction.
Iterative conjugate gradient inversion to optimise data
Computational Seismology: Developing and implementing numerical methods to solve the elastic, visco-elastic and poroelastic wave equations in complex Earth models. This requires a strong background in seismic wave propagation and numerical methods along with programming on high performance computing hardware.
A vertical fault rupturing releasing energy which propagates as seismic waves
Geophysical fluid dynamics: Investigating single and multi-phase flow in complex porous structures including contaminant and reactive transport. Combining numerical models of subsurface fluid flow processes with static and dynamic rock deformation schemes enabling the simulation of fluid-rock interactions in a wide variety of geophysical applications.
Dam break and flow from one side to the other within a sealed box
Re=20,000 flow around a cube with turbulance using a cumulatant LBM method (Geier et al. 2015)
Re=20,000 air-flow around an old university building cumulatant LBM method (Geier et al. 2015)
Gas bubbles flowing
Gas bubbles flowing
Hydrological modelling of a tracer through a fault zone using ADE solution
Volcano Seismology: Researching the seismic source mechanism in volcanic environments and seismic wave propagation in such media. Location of long-period volcano signals through novel high performance computing algorithms.
Ubinas volcano, Peru
Yasur volcano
Non-linear elastic deformation in low velocity shallow volcanic layers