Model-Guided Extremum Seeking Control for Mechanical Ventilators
Mechanical ventilators are an important lifesaving medical device and because of the COVID-19 pandemic, they were in a massive demand for them to help the patients most severely affected by the disease. A large number of open-source hardware ventilator projects were established during the pandemic, most of which did not provide sufficient information on the design of their control systems. In this thesis, a model-guided extremum seeking control scheme is proposed for use on a open-source hardware ventilator design. The performance of the control scheme is intended to not be significantly affected by changes to the patient, nor to variations in the ventilators geometry. To test the robustness of the model-guided extremum seeking control scheme to patient variation, a prototype ventilator was built, using the hardware design from one of the open-source hardware ventilator projects, and experiments were conducted using a test lung. In this process it was determined that the control scheme was able to achieve its goal of converging unto a targeted volumetric flow rate during inspiration for all volume-controlled ventilation tests performed. To test the robustness of the model-guided extremum seeking control scheme to variations in the ventilator design, a numerical model was created that replicates the experimental device. The numerical model was able to reproduce the experimental results with a 16% difference for the tidal volumes and a 14% difference for the positive end expiratory pressures. Various ventilator parameters were investigated, and it was found that the only geometric parameters that had a significant effect on the ventilators performance were the diameter of the main ventilator piping and the length of the expiratory piping. Through all the tests performed via the numerical model, the model-guided extremum seeking control scheme was always able to converge upon the target volumetric flow rate during inspiration.