Priscila Magalhaes Ribeiro - Petrobras
Rodrigo Surmas - Petrobras
Pablo Machado Barros - Petrobras
Guillaume Durance - Total Energies
Sylvia dos Anjos - COPPE/UFRJ
Diretora da ABGP
Diretora da ABGP
Pedro Mário Cruz e Silva - NVIDIA
Felix T. T. Gonçalves - COPPE/UFRJ
Reconstructing fluid mechanics problems from partial data is essential in science and engineering, especially regarding optimal design, biomedical and geophysical flows, parameter estimation, and more. These inverse problems are often ill-posed; thus, it's challenging (sometimes even impossible) to solve them using traditional methods. Moreover, generating simulated data for ill-posed inverse problems can become very costly since simulations must be performed several times to either discover missing physics or calibrate the free parameters in the model. We'll show a physics-informed neural network framework for reconstructing two examples of complex fluid mechanics problems: gravity currents and bubble dynamics. It reconstructs the other missing fields (velocity, pressure, etc.), given a set of sparse partial data, such as heavy fluid concentration in gravity currents or snapshots of the bubble position. We also discuss optimal sensor placement and dimensionality reduction for gravity current.
Alvaro Coutinho Professor, COPPE/UFRJ
RISC2 (https://www.risc2-project.eu) é um projeto de colaboracao em HPC entre a América Latina e a União Europeia, que objetiva criar uma rede de apoio para coordenar as ações em HPC.
https://www.youtube.com/watch?v=sJXrSJIjhQI&t=32s
https://www.youtube.com/watch?v=sJXrSJIjhQI&t=32s
Petrophysical characterization of carbonate reservoirs remains a challenge. The widespread use of X-ray computed microtomography (micro-CT) images in the classical reservoir characterization workflow allows the use of recent artificial intelligence algorithms to improve the process. This work presents an end-to-end workflow for permeability prediction using deep learning models and micro-CT images. A dataset of 37,600 slices from 376 plug samples from Brazilian presalt carbonate rock, along with the laboratory determined absolute permeability of each sample, were used for model training. Three models were tested: two convolutional neural network models (CNN and CNNSPP) and an ImageNet pretrained model (Densenet161). The models were trained using MSE or the Huber loss and with/without data augmentation.All experiments were performed using 10-fold cross-validation, and the models performance were evaluated by the average prediction of all slices for each sample. In this study, the Densenet161 model achieved the best results. The comparison with other models shows that pretrained models have less influence of data augmentation and almost no difference with respect to the loss function. This shows the effect of transfer learning, even if micro-CT images are very different from ImageNet. The results show that the proposed workflow can automate and speed up the characterization of Brazilian presalt carbonate samples by processing micro-CT slices thereby allowing accurate estimations of absolute permeability within a few seconds.