基于介电特性的水中钾盐浓度检测

Detection of sylvine concentration in water based on permittivity

为了寻求快速检测水中污染物的方法,该文分析了钾盐水溶液中不同离子浓度与其介电特性的关系,研究基于介电特性的水中钾盐浓度快速检测方法。采用新型液体介电特性检测传感器,通过试验测量不同浓度的钾盐水溶液在宽频率范围内的频率响应特性,采用偏最小二乘(partialleastsquare,PLS)回归,建立钾盐水溶液中各种离子浓度的预测模型。通过主成分分析识别钾离子的特征频率,用得到的特征频率建立预测钾离子浓度的新模型。结果表明,每种离子都有其独特的频率响应模式,不同种类的钾盐溶液其频率响应特性在低频区存在明显差异。对低浓度混合溶液,在只使用增益数据时,钾离子浓度偏最小二乘(PLS)回归训练模型的R2高达0.98865,相应的均方根误差为0.37598mg/L;而验证模型的R2达0.98861,相应的均方根误差为0.41031mg/L。主成分分析和特征频率的识别使数据量大幅度减少,同时也有较高的预测精度。

In order to find out a fast detection method for pollutants in water,through analysis on the relationship between concentration of different ion or cation in sylvine solutions and dielectric properties,the rapid and accurate detection method for sylvine concentration in water based on permittivity was studied.Using the new permittivity detection sensor designed for liquid dielectric materials,frequency-response data of sylvine solutions with different concentrations over a wide frequency range were measured by experiments.With PLS regression,the concentration prediction model for various ions or cation in sylvine solutions was established.By principal component analysis,the characteristic frequencies of potassium cation have been identified.With the characteristic frequencies,the new prediction model for concentration of potassium cation was established.The results showed that each ion or cation solution has its unique frequency-response pattern;frequency-response properties of different kinds of sylvine solution are significant difference at the low frequency region.For mixing solutions of low concentration,the coefficient of determination（R2）was high up to 0.98865 while RMSE was 0.37598 mg/L for training PLS regression model of potassium cation concentration when only using gain data.For validation of PLS regression model,R 2 was high up to 0.98861 while RMSE was 0.41031 mg/L.Principal component analysis and identification of the characteristic frequencies can reduce the amount of data significantly,and also have a high accuracy.