Prediction and Optimization of Oxygen Production under Microgravity Conditions Using Machine Learning

Efficient oxygen generation under microgravity and magnetohydrodynamic (MHD) conditions is a cornerstone of NASA's life-support and in-situ resource utilization (ISRU) systems for lunar and Martian missions. This study presents a comparative analysis of three machine learning models-Long Short-Term Memory (LSTM), Support Vector Machine (SVM), and Decision Tree Regressor (DTR)-for predicting the rate of oxygen evolution in MHD-assisted electrolysis systems. A comprehensive dataset containing operational parameters such as current density, magnetic field intensity, and temperature was employed to train and validate the models. The proposed LSTM architecture, consisting of dual recurrent layers with dropout regularization, demonstrated superior predictive performance with an accuracy of ۹۱%, an R_x۰۰۰۰_B۲ of ۰.۹۳, and a mean absolute error (MAE) of ۰.۰۴۵ mL min, outperforming the SVM and DTR baselines by ۵-۸%. Additionally, the Prediction Recall (PR) metric reached ۹۴.۷%, confirming the reliability of the proposed approach in maintaining predictions within a _x۰۰۰۰_B۱۵% tolerance of actual experimental data. The results indicate that the LSTM model effectively captures the temporal dependencies and nonlinear dynamics governing oxygen evolution, offering a robust framework for real-time monitoring and control of electrochemical oxygen generation in space environments.

کلیدواژه ها:

MHD electrolysis ، microgravity ، oxygen generation ، machine learning ، LSTM ، space life-support systems