R,D&I Projects
LTrace Geosciences is a leader in research, development, and innovation for the energy sector. Since 2018, we have specialized in transforming complex academic research into practical, high-impact commercial solutions.
Our core expertise spans a wide range of fields, including Artificial Intelligence, Digital Rock Physics, History Matching, and Quantitative Seismic Interpretation. In every project, our focus is clear: to bridge the gap between theory and application. We engineer advanced methods into robust, usable products with a high Technology and Commercial Readiness Level (TRL/CRL), delivering tangible value and a competitive edge to our clients.
Partner with us to turn your geological challenges into technological opportunities.
Portfolio
Sensitivity Analysis and Automation of the Generation of Relative Permeability Results in Digital Rocks
Status: Ongoing
This project aims to create new methods for simulating two-phase fluid flow in digital representations of rock. The primary focus is on analyzing how changes in relative permeability affect the simulation outcomes. These new techniques, combined with artificial intelligence and high-performance computing workflows, will be integrated into the GeoSlicer software to automate repetitive data analysis. The ultimate goal is to efficiently apply these advanced methodologies to a large volume of rock samples from Brazil’s Pre-salt layer.
Multiscale Analysis of Digital Rock Images for Reservoir Characterization
Status: Ongoing
This project aims to overcome the challenge of characterizing highly heterogeneous carbonate reservoir rocks by creating large, “meter-sized” 3D digital models. By integrating rock images from various scales to capture the complex pore structures of pre-salt formations, the project seeks to estimate rock properties at a more representative scale. The ultimate goal is to improve the accuracy of reservoir simulations and production forecasts.
Deep Learning for Building and Updating Reservoir Models
Status: Ongoing
Computational intelligence techniques, focusing on deep neural networks (deep learning), seeking to improve the quality and computational performance of constructing and the updating of geological models.
Multiscale PNM – Module for extracting multiscale pore and connection models for multiphase flow analysis
Status: Complete
Development of methods for extracting pore and multiscale connection models that enable the simulation of multiphase flow at the plug scale. In this project, the results were delivered under conditions sufficient for field testing and validation. These conditions include a graphical interface in the company’s software and high-performance computing processing. The ultimate goal was to enable deployment in a production environment for immediate value generation.
Smart Segmenter, a deep learning segmentation library for the GeoSlicer integrated digital rocks platform
Status: Complete
The project’s main objective was to increase the level of automation in oil industry image segmentation within a geological context. To this end, methodologies highlighted in the literature were selected, including models with convolutional neural networks and deep learning. These methodologies were added to a catalog of segmenters within the GeoSlicer digital rock tool. This enabled the use of the best method for batch segmentation of multiple images.
Large Image Visualization and Processing
Status: Complete
Design of a computational framework aimed at scaling digital rock analysis tools for application to images on the order of 10^13 elements using distributed processing on clusters, including monitoring of multiple simultaneous tasks and in-situ visualization. The objective was to scale the software developed by the proponent in a previous Cooperation Agreement and validated by geoscientists in the oil and gas industry, accelerating digital rock workflows.
Integrated platform for image analysis and simulation of Digital Rock
Status: Complete
The objective of this project was to research and develop methodologies and tools in digital rock physics for integration into the GeoSlicer software. The proposed developments included machine learning and cloud computing techniques. The specific objectives proposed met the growing demand for inputs in this field and enabled the analysis and processing of large sample volumes.
Deep learning for 4D seismic assimilation into reservoir models.
Status: Complete
Accelerate the process of updating geological models with GPUs and use data assimilation techniques with key information automatically extracted from 4D seismic data through deep learning, focusing on improving the predictability of production curves.
Segmentation and parameter inference of rock physics models in computed tomography using deep learning.
Status: Complete
The scope of this project was to investigate computational intelligence techniques for predicting petroelastic properties of rocks and lithofacies from digitized images at 2D, 3D, and whole-core scales, and to implement a system applicable to real data using the new research algorithms. The main product of this project was the first version of GeoSlicer.