Automated identification and quantification of rock types from drill cuttings
Youssef Tamaazousti, M C Francois, Josselin Kherroubi · 2020
Identifying and quantifying rock types from drill cuttings is a key step in characterizing the reservoir. Today, because it is manually done, this step is subjective and time-consuming. Automating this task is crucial to gain efficiency and objectivity. Some work has been proposed in the literature, but all consider a classification approach that does not allow for quantification. Here, we propose formalizing the problem as a segmentation task and solve it with a convolutional neural network in a transfer-learning scenario. Extensive experiments have been conducted and very promising results were obtained on single lithology samples, mixed ones and even wet cuttings. Presentation Date: Wednesday, October 14, 2020 Session Start Time: 1:50 PM Presentation Time: 4:45 PM Location: 351F Presentation Type: Oral