| Multi-Attribute Identification of Salt Structures in Seismic Data Using a Multilayer Perceptron Neural Network |
| Paper ID : 1042-UGESR |
| Authors |
|
Ali Madah Jalali, Masoume Lotfi * Faculty of Advanced Petroleum and Energy Technologies, University of Science and Technology of Mazandaran (MAZUST), Behshahr, Mazandaran, Iran |
| Abstract |
| Accurately delineating seismic geological features, such as salt domes, is a crucial step in understanding subsurface geology and identifying potential hydrocarbon traps. Traditional methods relying on single-attribute analyses are often labor-intensive and inefficient. Since each seismic attribute reveals unique details through its own distinct computational algorithm, an integrated approach is required to effectively analyze all these results. In this study, a semi-automatic method for merging seismic attributes was proposed to identify salt domes. After preprocessing the seismic data, a combination of attributes helpful in identifying salt bodies and sets of interpreted target and background points were fed into a supervised neural network. In this process, the neural network distinguished areas associated with salt masses from the background by utilizing information derived from various features. Ultimately, this approach yielded clearer, higher-resolution images of the salt structures. A comparison of the binary output model with the reference model yielded an accuracy of 94.3% and an F-score of 91.7%. |
| Keywords |
| Salt structures, 3D seismic data, Seismic attributes, Artificial neural network, Back-propagation algorithm |
| Status: Accepted (Poster Presentation) |