模式识别用于压电晶体传感器阵列定量检测有机蒸汽混合物

本文应用9个压电晶体组成传感器阵列, 每片晶体上分别涂有不同种类的冠醚衍生物, 用它来定量检测二元及三元有机蒸汽混合物, 在数据处理中比较了两种模式识别方法---偏最小二乘法(PLS)和人工神经网络法(ANN), 实验证明, ANN法在预测准确度上明显优于PLS法, 本文还讨论了解决神经网络训练过拟合现象的方法。

Organic vapor mixture detection with piezoelectric crystal sensor array using pattern-recognition techniques

An array of nine piezoelectric quartz crystals, each coated with a different derivative compound of crown-ether, was constructed for mullticomponent analysis of organic vapors. The usefulness of this array was evaluated by quantitating known mixture samples in both two and three component cases using two chemometric techniques, partial least squares (PLS) and artificial neural networks (ANN). The results show that the prediction was better for ANN in both two and three component cases. The method to avoid overfitting in ANN training was also discussed.