Publications

Geologists usually use wellbore cores to identify subsurface facies. Since not all wells have wellbore cores, wireline logs are also commonly used to identify subsurface facies as data that tends to be available on all wells. By utilizing computing technology, depositional facies can be interpreted more quickly and with high accuracy.

In this study, a Convolutional Neural Network (CNN) deep learning algorithm was used for generating dummy gamma ray (GR) log data and classifying depositional facies for each GR log trend. The log patterns used in this research include bell, funnel, and cylindrical patterns, which identify different depositional environments. To get the desired results, the dummy data generated is input into the convolutional layer by proposing models, including dropout, max pooling, flattening, and dense layer. The training and validation accuracy results attained 96%, demonstrating the effectiveness of the CNN algorithm in identifying facies in GR log data.

However, it is essential to acknowledge that classification results obtained using computational methods like this inevitably have limitations, as successful accuracy may not reach 100%. Therefore, assistance with manual interpretation is still necessary to refine the data obtained after obtaining the results. In future work, this method could be applied to actual data, leveraging computational technology to shorten the required time and reduce multiple interpretations compared to conventional interpretation work.