Fluctuation-enhanced sensing of organic solvent vapors mixture by machine learning

The data set consists of exemplary results of the product of voltage noise power spectral density S(f) multiplied by frequency f and normalized to squared DC voltage U^2 recorded in the graphene back-gated Field Effect Transistor under UV light assistance (275 nm) in the selected ambient atmospheres (Figure 3) and the results of gas detection by SVM algorithm: 1) chloroform (Figure 5), 2) acetonitrile (Figure 6), and predicted gas concentrations using various number of frequency bins (Figure 7, Figure 8). The demonstrated data reveals we can determine two gas components in the considered gas mixture (chloroform and acetonitrile) by utilizing flicker noise and the SVM detection algorithm. When we considered noise power spectra in the frequency range 0.5 Hz—2 kHz, the gas detection limit reached 2.9 ppm for chloroform and 49.5 ppm for acetonitrile.

本数据集包含两类核心结果：其一为石墨烯背栅场效应晶体管（Field Effect Transistor）在275 nm紫外光（UV）辅助、选定环境氛围中采集的电压噪声功率谱密度S(f)与频率f的乘积，并归一化至直流电压平方U²的示例性结果（对应图3）；其二为基于支持向量机（Support Vector Machine，SVM）算法的气体检测结果与不同频率仓数量下的气体浓度预测结果，其中气体检测结果分别对应三氯甲烷（图5）与乙腈（图6），预测浓度结果对应图7、图8。
本数据集展示的数据表明，通过结合闪烁噪声与支持向量机检测算法，可在所研究的三氯甲烷与乙腈混合气体中识别出两种气体组分。当将噪声功率谱的分析范围设定为0.5 Hz至2 kHz时，三氯甲烷的气体检测限可达2.9 ppm，乙腈的气体检测限可达49.5 ppm。