遗传算法-反向传播神经网络优化气相色谱质谱联用法测定香蕉挥发性组分

Determination of volatile components in banana by gas chromatography -mass spectrometry optimized by genetic algorithm combined with back propagation neural network

中文关键词:  香蕉  挥发性组分  固相微萃取  气相色谱质谱联用法  反向传播神经网络  遗传算法  自动质谱退卷积定性系统

英文关键词:banana  volatile components  solid phase microextraction  gas chromatography-mass spectrometry  back propagation neural network  genetic algorithm  automated mass spectral deconvolution and identification system

中文摘要:

      目的 应用反向传播(back propagation, BP)神经网络结合遗传算法(genetic algorithm, GA)优化固相微萃取(solid phase microextraction, SPME)条件, 建立更优的测定香蕉果肉挥发性组分的气相色谱质谱联用法(gas chromatography-mass spectrometry, GC-MS)和组分定性方法。方法 通过正交实验优化SPME提取参数: 超声时间、样品量和萃取温度, 在正交实验基础上, 运用BP神经网络结合GA寻找SPME最佳的萃取参数, 应用GC-MS对香蕉果肉挥发性组分进行定性和定量分析。结果 根据GA-BP神经网络寻优得到萃取参数: 超声时间25.0 min、样品量2.9 g、萃取温度49.0 ℃。在此最佳萃取参数条件下测定香蕉果肉的挥发性化合物, 共鉴定出香蕉果肉63个挥发性组分, 主要以酯类为主(占相对总含量的75.75%), 相对含量最大的组分是乙酸异戊酯(1281.26 μg/kg)。结论 本研究通过GA-BP神经网络优化SPME条件, 再通过解卷积软件处理原始质谱图, 提高了香蕉挥发性组分鉴定的可靠性和鉴定组分数量, 为测定果蔬等农产品的挥发性组分与评价其品质提供参考。

英文摘要:

      Objective To establish a better gas chromatography-mass spectrometry (GC-MS) and component qualitative method of volatile components in banana pulp, and optimize solid phase microextraction (SPME) conditions by using back propagation (BP) neural network combined with genetic algorithm (GA). Methods The extraction parameters of SPME were optimized by orthogonal test: Ultrasonic time, sample weight and extraction temperature, on the basis of the orthogonal test, the BP neural network combined with GA was used to find the optimal extraction parameters of SPME, and GC-MS was used to qualitatively and quantitatively analyze the volatile components in banana pulp. Results The extraction parameters optimized by GA-BP neural network were as follows: Ultrasonic time 25.0 min, sample weight 2.9 g, and extraction temperature 49.0 ℃. Under the optimal extraction conditions, the volatile compounds in banana pulp were determined and 63 volatile components in banana pulp were identified, esters were the main components (accounting for 75.75% of the relative total content), and the component with the largest relative content was isoamyl acetate (1281.26 μg/kg). Conclusion In this study, the SPME conditions are optimized by GA-BP neural network, and then the original mass spectrogram is processed by deconvolution software, which improved the reliability of identification of banana volatile components and the number of identified components, and provided reference for determining volatile components of fruits and vegetables and other agricultural products and evaluating their quality.