Parametric regression of sound velocity for predicting carbon dioxide content

The detection of air gas components is of great significance to prevent the occurrence of harmful gases, monitor gas leaks, and evaluate the state of air pollution. The sound velocity method has the characteristics of a wide range, high stability, and applicability to many types of gases, so it has been widely used in the concentration detection of binary known mixed gases. However, the variation of sound velocity is non-selective, and it cannot determine the composition type; environmental factors such as temperature, humidity, and atmospheric pressure also affect the fluctuation of sound velocity, which restricts the accurate measurement and analysis of the actual air composition using the sound velocity method. Therefore, based on the actual air sound velocity equation, this paper adopts the support vector regression (SVR) algorithm and takes the carbon dioxide content in air as an example for fitting prediction, combining physical quantities such as sound velocity, temperature, humidity, and atmospheric pressure. A machine learning algorithm is used to measure specific gas concentrations in real air. The measured results show that the SVR algorithm can well predict the change trend of carbon dioxide gas under the test conditions, with the predicted accuracy index coefficient of determination (R²) not less than 0.70, and mean absolute error (MAE) not exceeding 0.35.