Automated porosity estimation using CT-scans of extracted core data

Abstract Estimation of porosity at a millimeter scale would be an order magnitude finer resolution than traditional logging techniques. This enables proper description reservoirs with thin layers and fine heterogeneities. To achieve this, we propose end-to-end convolutional neural network (CNN) regression model that automatically predicts continuous using two-dimensional whole core CT scan images. More specifically, CNN is trained to learn from routine analysis (RCA) measurements. characterize the performance such approach, compare this two linear models relationship between average attenuation standard deviation same images RCA porosity. Our investigations reveal are outperformed by CNN, indicating capability in extracting textures important for estimations. We predicted results against total logs calculated density log. The obtained show values proposed method well correlated plug measurements importantly, approach can provide accurate estimations, while log averaged over interval thus do not variations. Thus, employed calibrate logs, thereby reducing uncertainties associated indirect calculations logs.