Model identification of ventilation air pump utilizing Ridge-momentum regression and Grid-based structure...

Area of Science:

• Engineering
• Control Systems
• Biomedical Engineering

Background:

• Mechanical ventilators are critical for respiratory support during pandemics.
• Accurate modeling of ventilator air generation units presents challenges due to system uncertainties and time-varying parameters.
• Existing modeling techniques struggle with complex dynamics and limited computational resources.

Purpose of the Study:

• To develop a novel and robust algorithm for identifying Autoregressive Moving Average (ARMA) models in ventilation pumps.
• To enhance model identification stability and adaptability for time-varying systems.
• To optimize the selection of ARMA model orders and time-delay parameters.

Main Methods:

• Implementation of Ridge regression modified with momentum (Ridge-M).
• Integration of a grid search strategy for joint optimization of model orders and time delays.
• Utilizing the Akaike Information Criterion (AIC) for model structure selection.
• Experimental validation using ramp and multistep input signals.

Main Results:

• Ridge-M demonstrated superior performance in capturing dynamic system behaviors compared to Recursive Least Squares (RLS) and standard Ridge regression.
• Ridge-M achieved an average reduction in Root Mean Squared Error (RMSE) of 2.7% over RLS and 6.8% over standard Ridge regression for multistep inputs.
• Standard Ridge regression showed a slight advantage for ramp inputs, while Ridge-M yielded the lowest overall average RMSE (31.6236).

Conclusions:

• The proposed Ridge-M algorithm offers a stable and adaptable approach for mechanical ventilator model identification.
• This method is suitable for embedded applications with noisy data, time-varying systems, and limited computational power.
• The findings contribute to improving the reliability and performance of essential respiratory equipment.