基于递归图和深度学习的电子鼻气体识别算法数据集

电子鼻系统作为一种仿生传感装置，可通过集成多通道传感器阵列与人工智能算法，实现对复杂气体成分的识别及浓度定量分析功能。本工作以8个商用气体传感器进行传感器阵列构建，完成电子鼻系统的设计。提出了一种基于传感器阵列融合递归图的数据处理方法用于传感器信号重构，结合深度神经网络构建气体识别模型。通过简单的二维卷积神经网络模型实现了6种气体98.37 %的识别准确率。对四种浓度识别算法进行比较，其中深度神经网络算法预测效果最好，实现了1.0695的均方根误差（RMSE）。进一步结合VGG-19模型构优化气体识别模型，实现了98.91 %的气体识别准确率，将浓度预测的RMSE降至0.5121。As a bionic sensing device, the electronic nose system can enable the identification of complex gas components and quantitative concentration analysis by integrating multi-channel sensor arrays and artificial intelligence algorithms. In this study, a sensor array was constructed with 8 commercial gas sensors, and the design of the electronic nose system was finalized. A data processing method based on recurrence plots fused with sensor array data was proposed for sensor signal reconstruction, and a gas recognition model was built by combining deep neural networks. A simple two-dimensional convolutional neural network (2D-CNN) model achieved a recognition accuracy of 98.37% for 6 types of gases. Four concentration prediction algorithms were compared, among which the deep neural network algorithm performed optimally, achieving a root mean square error (RMSE) of 1.0695. Furthermore, the gas recognition model was optimized by incorporating the VGG-19 model, reaching a gas recognition accuracy of 98.91% and reducing the RMSE of concentration prediction to 0.5121.

搜集汇总

背景与挑战

背景概述

该数据集旨在支持基于递归图和深度学习的电子鼻气体识别算法研究，通过构建传感器阵列并采用深度神经网络模型，实现了对多种气体的高精度识别与浓度预测。数据集包含实验数据文件，适用于人工智能领域的气体传感分析应用。