Construction of electronic nose system for wine SO2 content detection and optimization of sensor array

Traditional techniques to detect wine SO2 content are time-consuming, require expensive equipment, and professional personnel. In this study, an electronic nose system consisting of 16 commercial gas sensors is used to establish a rapid detection method for SO2 content in wine. To improve the detection performance of the electronic nose, we propose a sensor array optimization algorithm based on Recursive Sensor Elimination (RSE). The Maximum Information Coefficient (MIC) is used as the standard to measure the relationship between variables, and the Feature Importance (FI) as well as the Sensor Importance (SI) are defined. RSE removes the sensor with the lowest SI each time until the Coefficient of Determination (R-2) of the retained feature subset reaches the maximum value. Four regression models including PLSR, MLP, SVR and BRR are used to compare the detection performance of the data before and after optimization for the samples. The results show that the number of sensors in the optimized array is reduced from 16 to 5, and the quantity of features is decreased by 68.75%. The PLSR model based on the optimized array has the best performance, R-2 and Root Mean Square Error (RMSE) are 0.9686 and 12.11 ppm, respectively, which are better than the original array, and the running time is 0.6851 s.