Search

The construction of 3D geological models is an essential task in oil/gas exploration, development and production. However, it is a cumbersome, time-consuming and error-prone task mainly because of the model's geometric and topological complexity. The models construction is usually separated into interpretation and 3D modeling, performed by different highly specialized individuals, which leads to inconsistencies and intensifies the challenges. In addition, the creation of models following geological rules is paramount for properly depicting static and dynamic properties of oil/gas reservoirs. In this work, we propose a sketch-based approach to expedite the creation of valid 3D geological models by mimicking how domain experts interpret geological structures, allowing creating models directly from interpretation sketches. Our sketch-based modeler (Geo-Sketcher) is based on sketches of standard 2D topographic and geological maps, comprised of lines, symbols and annotations. We developed a graph-based representation to enable (1) the automatic computation of the relative ages of rock series and layers, and (2) the embedding of specific geological rules directly in the sketching. We introduce the use of Hermite-Birkhoff Radial Basis Functions to interpolate the geological map constraints, and demonstrate the capabilities of our approach with a variety of results with different levels of complexity., Comment: 21 pages, 30 Figures

We present a novel multimodal language model approach for predicting molecular properties by combining chemical language representation with physicochemical features. Our approach, MULTIMODAL-MOLFORMER, utilizes a causal multistage feature selection method that identifies physicochemical features based on their direct causal effect on a specific target property. These causal features are then integrated with the vector space generated by molecular embeddings from MOLFORMER. In particular, we employ Mordred descriptors as physicochemical features and identify the Markov blanket of the target property, which theoretically contains the most relevant features for accurate prediction. Our results demonstrate a superior performance of our proposed approach compared to existing state-of-the-art algorithms, including the chemical language-based MOLFORMER and graph neural networks, in predicting complex tasks such as biodegradability and PFAS toxicity estimation. Moreover, we demonstrate the effectiveness of our feature selection method in reducing the dimensionality of the Mordred feature space while maintaining or improving the model's performance. Our approach opens up promising avenues for future research in molecular property prediction by harnessing the synergistic potential of both chemical language and physicochemical features, leading to enhanced performance and advancements in the field., Comment: 14 pages, 6 Figures, 5 tables. Submited to NEURIPS 2023, Under review