Highly Sensitive Porphyrin Sensor Modified by Organic Nano-Skeleton Material Combined with Convolutional Neural Network Model for Discriminating Large-Leaf Yellow Tea Roasting Degree

Abstract

Roasting is an important part of forming the unique roasting flavor of large-leaf yellow tea (LYT). However, rapid and scientific methods for monitoring the roasting degree have not yet been developed. In this study, novel colorimetric sensors based on nano-modified and PSN/MOF porous materials modified TPP dyes were proposed for monitoring the roasting degree of LYT. First, four TPPs were screened according to their response to LYT aroma. Scanning electron microscope (SEM), Transmission electron microscope (TEM) and energy dispersive spectrometer (EDS) were used to characterize nanoporphyrin (N-TPP) and PSN/MOF. Then, the RGB and hyperspectral response features of different sensing arrays were extracted and the performance of extreme learning machine (ELM), least square support vector machine (LSSVM) and convolutional neural network (CNN) algorithms for processing sensing array data were compared. Among the established roasting degree models, CNN shows the highest discriminant rate. Both the PSN/MOF@N-TPP-based CNN models achieve 100% discriminant rate, which is better than the TPP-based CNN model (90%). The results show that the proposed method can effectively improve the monitoring accuracy of LYT roasting degree.

Keywords: Large-leaf yellow tea, Roasting degree, Colorimetric sensor array, Nanoporphyrin, Porous silica nanosphere, Metal organic framework