Automatic identification and separation of reflection patterns with the help of clustering of seismic attributes in a Rain optimization meta-heuristic algorithm

Abstract

Seismic exploration, a key component of geophysical methods, is crucial for analyzing subsurface structures and evaluating their potential for hydrocarbon resources. However, the interpretation of geological structures based on seismic data frequently entails ambiguity and uncertainty, making it a labor-intensive endeavor that is heavily reliant on the interpreter's expertise. Seismic attributes are essential instruments for the quantitative assessment of seismic information, facilitating the identification and delineation of structural and stratigraphic elements by revealing concealed details. This paper aims to conduct a multi-attribute analysis for the automatic and unsupervised stratigraphic interpretation of two-dimensional seismic data. The research employs optimization-based clustering utilizing the Rain meta-heuristic algorithm to enhance the detection of reflection patterns within the seismic data. To optimize computational efficiency and mitigate data redundancy, a subset of extracted seismic attributes was selected through the Laplacian scoring feature selection method. The results were validated against geological evidence to ensure both reliability and accuracy. The findings underscore the effectiveness of unsupervised clustering methodologies, particularly meta-heuristic optimization strategies, in enhancing the efficiency and precision of seismic interpretation. Notably, these methods automatically ascertain the optimal number of clusters, thus providing a degree of flexibility that traditional techniques, such as k-means, do not afford. The study further elucidates those meta-heuristic methods, especially the ROA method, yield superior clustering outcomes in comparison to genetic algorithms (GA) and particle swarm optimization (PSO).