基于传感器阵列结合GA-BP算法对VOC的识别研究

VOC recognition model based on sensor array and GA-BP algorithm

为提高空气中挥发性有机物(VOC)检测可靠性,提出了一种基于气敏传感器阵列结合遗传算法(GA)优化反向传播(BP)神经网络算法的VOC检测模型。选用多个气体传感器组建阵列对VOC混合气体样本进行响应测试,使用主成分分析(PCA)对响应数据进行数据降维及初步分类探索,使用构建的GA-BP神经网络算法模型在PCA探索性分析的基础上进行定性及定量识别,并与BP神经网络识别结果进行对比。结果表明:遗传算法优化后的BP神经网络多元分类和回归模型性能优良且稳定,气体分类识别准确率达96%,浓度回归预测均方根误差为1.8×10-2,平均相对误差为5.2%,平均训练耗时分别降至1.5 s和1.12 s,效果显著优于BP神经网络算法模型。这些研究结果进一步拓展了GA-BP算法结合气敏传感器在挥发性有机物检测识别中的应用前景。

In order to improve the detection reliability of volatile organic compounds(VOC)in the air,a VOC detection model based on gas sensor array combined with genetic algorithm(GA)optimized back propagation(BP)neural network algorithm was proposed.Multiple gas sensors were used to form an array to test the response of VOC mixed gas samples.Principal component analysis(PCA)was used to reduce the dimension of the response data and explore the preliminary classification.The GA-BP neural network algorithm model was constructed,and the qualitative and quantitative identification were carried out on the basis of PCA exploratory analysis,and it was compared with BP neural network recognition results.The results show that the performance of BP neural network multivariate classification and regression model optimized by genetic algorithm is excellent and stable.The accuracy of the gas classification and recognition is 96%,the root mean square error of the concentration regression prediction is 1.8×10-2,the average relative error is 5.2%,and the average training time is reduced to 1.5 s and 1.12 s,respectively.And this is significantly better than that of BP neural network algorithm model.The results further expand the application prospect of GA-BP algorithm combined with gas sensor in the detection and identification of volatile organic compounds.