This thesis rigorously investigates the practicality of data-driven models in predicting the time derivative of bulk velocity within oscillatory flows through porous media. By employing a comprehensive approach, it delves into the utilization of macroscopic pressure gradients and time-lagged values of bulk velocity derivatives as important input features. Furthermore, the study advances innovative methodologies, including data balancing techniques to mitigate inherent dataset biases, and introduces a specialized inception-like model specifically designed for numeric data, aiming to bolster prediction accuracy. Additionally, the research uses two distinct lagging techniques: one employing linear interpolation and the other leveraging Laguerre polynomials, with a keen focus on understanding their individual impacts on prediction efficiency. Through priori and posteriori validation procedures, the thesis ensures the robustness and reliability of the proposed methodologies in accurately predicting flow behavior in porous media under oscillatory conditions, thereby shedding light on new avenues for future research and development in this domain.     «

This thesis rigorously investigates the practicality of data-driven models in predicting the time derivative of bulk velocity within oscillatory flows through porous media. By employing a comprehensive approach, it delves into the utilization of macroscopic pressure gradients and time-lagged values of bulk velocity derivatives as important input features. Furthermore, the study advances innovative methodologies, including data balancing techniques to mitigate inherent dataset biases, and in...     »