LITHOLOGY IDENTIFICATION BY ADAPTIVE FEATURE AGGREGATIONON GRAPH NEURAL NETWORKS

Abstract

The current complex problems in the field of geological science and hydrocarbon exploration require the introduction of innovative technologies to improve the prediction of rock lithology. In this context, automatized lithology prediction becomes extremely relevant. The scientific research represented in this article offers an integrated approach that combines modern machine learning methods with the analytical power of graph neural networks. This stack allows for the effective processing of logging data, improving the quality and accuracy of lithological classification processes traditionally performed by experts. Traditional methods of logging interpretation, which rely on the human factor, often result in significant time and resource costs. In addition, differences in the knowledge and experience of different experts can lead to inconsistencies in results. With the increasing volume and complexity of data to be interpreted, it is becoming increasingly important to use new, automated approaches. Besides, wavelet analysis combined with graph structures provides new opportunities for faster and more accurate processing of geological data.

The purpose of this study is to improve conventional methods of lithology identification by introducing adaptive feature aggregation into graph neural network models. The well-defined purpose of the study allows us to focus on the possibilities of introducing the latest technologies into the geological modeling process. The structure of the work is logically organized, ensuring consistency and coherence of the presentation. The author have demonstrated a thorough understanding of the subject area, which enhances the value of the obtained scientific conclusions. The article is particularly valuable in that the methods presented can significantly improve accuracy and reduce the risks associated with the exploration process, especially in conditions of limited data. This demonstrates the author's high level of expertise in the subject and his ability to adapt to current challenges in the field. The methodological part of the study is characterized by an ideal combination of theory and practice, which makes the proposed approach understandable and useful for the scientific community. A significant advantage is the development of algorithmic principles that take into account the specific needs and limitations when working with logging data. This is a significant contribution to solving the problem of computational complexity that often arises in the context of geological exploration. The scientific novelty of the research consist in the implementation of adaptive methods that open up new perspectives in the automation of learning processes based on the collection of identification signals.

Consequently, this scientific article is intended not only to improve the accuracy and reliability of lithological identification, but also to contribute to the improvement of existing approaches to geological research. Provided results can be effectively integrated into the practice of modern geological exploration, demonstrating broad potential for further development in this area.