Fraunhofer Institute for Industrial Mathematics ITWM

Seismic Reflection Method

Seismic measurement methods – especially the seismic reflection method – are very important for the exploration of crude oil and natural gas. Artificially excited sound waves are used for the examination of the subsurface. These are excited at several different points at the surface.

Seismic Imaging

Seismic, HPC, and AI for Oil, Gas, and Energy Research

For around 20 years, we have been a research partner for companies in the international oil and gas industry – thanks to our combined expertise in applied seismics, High Performance Computing (HPC) and, more recently, Machine Learning (ML). As part of contract research, we develop algorithms and methods for seismic data processing, seismic prestack imaging, and the interpretation of migrated data.

We deliver project results along the entire value chain: from the design and development of new algorithms through to the delivery of robust, production-ready software packages. Due to the large data volumes involved in our applications, we place strong emphasis on scalability for parallel computing architectures already during algorithm development. We implement parallelism at all levels – intra-node, inter-node and vectorization. Our software often benefits directly from HPC tools such as fast parallel file systems and parallelization libraries developed by our colleagues in other teams within our »High Performance Computing« department.

In addition to custom-built software packages, we also develop licensable products and carry out specialized data processing projects for companies using our proprietary solutions. Our daily work with customer data – often from recent surveys using modern acquisition concepts – enables us to identify real-world challenges early on and to initiate projects for improved data processing or to address them directly on customer request. In particular, our developments in the field of Machine Learning are driven and continuously advanced by these synergies.

Article »Research News«

Seismic Imaging Provides an Insight into the Earth’s Subsurface

The seismic reflection method involves sending acoustic waves thousands of meters under the Earth’s surface. The reflected signals provide scientists with information about the characteristics of the subsurface. This is how oil and gas deposits can be discovered. While these will eventually no longer be extracted as we move away from fossil fuels, accurate knowledge of rock structures is important for better understanding known reservoirs. The Earth’s subsurface could be a key place to store greenhouse gases. With improved machine learning algorithms and new high-performance computing concepts, researchers at the Fraunhofer Institute for Industrial Mathematics ITWM are now able to process seismic data in greater detail. The technology could also be useful for constructing wind farms.

Research News Article (fraunhofer.de)

Our Research Activities Focus on the Following Main Areas:

Seismic depth migration (»Generalized Radon Transform« / GRT)
Specialized Processing of marine OBN and land data
Deep neural networks for processing and interpretation of seismic data (DLseis)
Seismic methods for renewable energy

Example Projects and Methods

Boulder-Detection with Machine Learning

We developed a process that detects and locates even small boulders or other disturbing obstacles in the ground.

DLSeis

The project »Deep Learning for Large Seismic Applications« (DLseis) deals with basic research up to ready-to-use deep learning tools for seismic applications.

Seismic Prestack-Depth Migration GRT

»Generalized Radon Transform« (GRT) depth migration excellently demonstrates how the unified competence of the department HPC could be used for the successful development of a tailor-made product.

RTM

The method of Reverse Time Migration (RTM) stands out by high imaging quality even in case of high geological complexity. Solving the wave equation realistically allows the exact imaging of structures with strongly contrasting seismic velocities as they occur.