Hybrid Transformer-Graph Neural Network for Nuclear Tracer Analysis of Mineral Salts in Crude Oil

This study introduces a novel approach integrating Graph Neural Networks (GNNs) with Transformer architectures (GNN-Transformer) for the identification of mineral salt depositions within oil pipelines. Utilizing nuclear tracer techniques, where a radioactive source (e.g., Cobalt-۶۰) is employed to irradiate the pipe and detectors are placed externally, we analyze the transmitted radiation patterns. The core challenge lies in interpreting these complex radiation signatures to infer the type and extent of deposited salts, which significantly impact pipeline integrity and flow. The methodology involves simulating the nuclear transport process using MCNPX, a widely adopted code for simulating radiation transport. The simulation models the pipeline geometry, the external radiation source, and crucially, the presence of various precipitated mineral salts within the fluid. Key salts considered include NaCl, CaCl۲, MgCl۲, CaSO۴, SrSO۴, BaSO۴, and CaCO۳, alongside water (H۲O), each with distinct elemental compositions and densities. These simulations generate intricate datasets representing the radiation signals detected outside the pipe under different salt deposition scenarios. We then employ a GNN-Transformer model to learn effective representations from these simulated tracer data. The GNN component is adept at capturing the relational dependencies within the spatial configuration of the pipeline and the deposition, while the Transformer's attention mechanism excels at processing sequential or graph-based information to identify subtle patterns indicative of specific salt types. The model is trained to map the complex detector signals to the corresponding mineral salt types. Preliminary results demonstrate the model's potential in accurately classifying the dominant salt constituents based on the tracer data analyzed, offering a promising non-destructive method for real-time monitoring and management of mineral scaling in the oil and gas industry.

کلیدواژه ها:

Nuclear Tracing ، Mineral Salt Identification ، GNN-Transformer ، MCNPX Simulation ، Pipeline Integrity