Neural network-based prediction of K-۴۰ dispersion trajectory in a hypothetical crude oil desalter accident and its health impact assessment

This study presents an integrated framework for predicting airborne K-۴۰ dispersion trajectory and assessing associated health impacts following a hypothetical accident at the Ahvaz Central Desalter Unit in Iran (approximately ۳۱.۲۱°N, ۴۸.۹۴°E). The methodology combines HYSPLIT۴ atmospheric transport modeling, MLP neural network optimization, and MCNPX Monte Carlo dosimetry. HYSPLIT_۴ simulates the initial plume trajectory and dispersion pattern of K-۴۰ released during the accident, utilizing its hybrid Lagrangian-Eulerian approach for accurate atmospheric transport tracking. To enhance prediction accuracy and computational efficiency, a Multi-Layer Perceptron (MLP) neural network is developed and trained on key meteorological parameters, including wind speed, wind direction, ambient temperature, and atmospheric stability class, to capture the complex, non-linear relationships governing plume behavior. The optimized MLP generates highly precise trajectory predictions, providing rapid and reliable estimates of contaminant transport to downwind receptor locations, thereby supporting timely safety decisions. Subsequently, the MCNPX code is employed to calculate the radiological dose delivered to various human organs from the ۱.۴۶ MeV gamma radiation emitted by K-۴۰. Using computational anthropomorphic phantoms, the simulation determines organ-absorbed doses and effective dose conversion factors for occupationally exposed workers at the facility and for members of the public residing in surrounding residential zones. The Monte Carlo approach yields statistically reliable effective dose coefficients, enabling a comprehensive evaluation of potential health risks, including stochastic effects. The seamless integration of HYSPLIT trajectory modeling, MLP-optimized dispersion predictions, and MCNPX organ dosimetry offers a robust, multi-disciplinary HSE risk assessment tool. This framework effectively supports emergency response planning, safety zone delineation, and radiation protection decision-making for desalter unit accidents in the oil and gas industry, with specific application to the Ahvaz facility

کلیدواژه ها:

Neural Network ، HYSPLIT ، MCNPX ، K-۴۰ ، Oil Desalter