Neural Network-Based Self-Calibrating Control for Advanced Oxygen Supply Systems

• Published in: Chinese Control and Decision Conference pp. 3575 - 3580
• Main Authors: Yang, Jinshan, Sun, Qinglin, Tan, Panlong, Sun, Hao
• Format: Conference Proceeding
• Language: English
• Published: IEEE 16.05.2025
• Subjects: Control systems; Mathematical models; Neural Networks; Optimization; Oxygen; Oxygen Mask; Pressure Regulation; Real-Time Optimization; Real-time systems; Regulation; Regulators; Robustness; Uncertainty; Ventilators
• ISSN: 1948-9447
• DOI: 10.1109/CCDC65474.2025.11090443

The oxygen supply system (OSS) is characterized by its compact size, high airtightness, and rapid pressure adjustment capabilities, necessitating advanced control performance in oxygen regulators. Unlike traditional controllers with fixed parameters, this study presents a neural network-based self-calibrating controller that dynamically optimizes control parameters in response to system states. The structural design and operating principles of the OSS are systematically detailed, followed by the development of a nonlinear mathematical model derived through mechanistic analysis. Leveraging this model, the proposed controller employs real-time feedback of differential pressure between the interior and exterior of the oxygen mask, with online weight updates using a backpropagation algorithm. This approach effectively addresses challenges such as environmental uncertainties and variations in system characteristics. To evaluate performance, respiratory patterns under various scenarios-modulating respiratory rates and tidal volumes-are mathematically simulated. Beyond its primary application, the system demonstrates potential for broader adoption in civilian domains, including medical ventilators and high-altitude exploration, ensuring reliable oxygen delivery across diverse environments.