对羟基苯甲酸甲酯钠是一种常见的食品添加剂,如果长时间食用或者超量食用会对人体造成一定的危害。采用FS920荧光光谱仪对对羟基苯甲酸甲酯钠橙汁溶液和水溶液进行检测,实验结果表明两者的特征峰发生了明显的变化。经分析得出,对羟基苯甲酸甲酯钠橙汁溶液的荧光光谱受到橙汁荧光特性干扰,一定浓度范围的溶液光谱图存在较大差距,对羟基苯甲酸甲酯钠浓度与荧光强度之间的关系复杂。为了精确地检测橙汁中对羟基苯甲酸甲酯钠的浓度,结合荧光光谱法与最小二乘支持向量机,建立了橙汁溶液中对羟基苯甲酸甲酯钠的检测模型,使用改进的粒子群优化算法得到影响模型性能的正则化参数和核函数。实验得到了较为理想的结果,与普通反向传播(BP)神经网络、基本粒子群寻参的最小二乘支持向量机等方法相比,该方法性能最优,得到的平均回收率为97.05%,平均相对误差为2.71%,均方根误差为3.04%,模型输出与真实值之间的相关系数是0.9999。该方案可以做为橙汁中对羟基苯甲酸甲酯钠浓度的精确检测方法。 Sodium methylparaben is a common food additive that can cause some harm to the human body if it is consumed for a long time or over consumed. FS920 fluorescence spectrometer was used to detect sodium methylparaben orange juice and aqueous solution. The experimental results show that the characteristic peaks of the two have obvious changes. The results showed that the fluorescence spectra of methylparaben sodium orange juice solution were disturbed by the fluorescence characteristics of orange juice. There was a big gap in the solution spectra of certain concentration range. The relationship between methylparaben sodium concentration and fluorescence intensity complex. In order to accurately detect the concentration of sodium methylparaben in orange juice, the detection model of sodium methylparaben in orange juice was established by combining fluorescence spectrometry and least square support vector machine. The improved Particle Swarm Optimization The algorithm obtains regularization parameters and kernel functions that affect the performance of the model. Compared with ordinary backpropagation (BP) neural network and least square support vector machine based on PSO, this method has the best performance and the average recovery is 97.05% , The average relative error is 2.71% and the root mean square error is 3.04%. The correlation coefficient between the model output and the real value is 0.9999. The scheme can be used as an accurate method for the determination of the sodium methylparaben concentration in orange juice.