摘要
由于气体数据有限以及气体分类的准确率不高,解决气体分类问题有较大的难度。传统的模式识别算法虽然能够运用到气体分类中,但在不同浓度的气体数据下训练和测试性能会下降,准确率也比较低。本研究对8阵列传感器和24阵列传感器采集到的混合气体进行分类,提出了结合主成分分析(Principal component analysis, PCA)应用支持向量机(Support vector machine, SVM),构建PCA-SVM模型对不同浓度的一氧化碳(CO)、甲烷(CH_(4))、硫化氢(H_(2)S)以及乙醇(C_(2)H_(6)O)的混合气体进行分类的方法。与SVM、参数优化的BP神经网络(Back propagation neural network, BPNN)和PCA-BP神经网络模型对比的结果表明,在随机选择的数据中,采用PCA-SVM模型能够提高分类的性能,在含有13个特征的气体数据集中,利用PCA-SVM模型准确度达到98.974%。在含有27个特征的数据集中,利用PCA-SVM模型准确度达到100%,能够满足对混合气体分类的实际需求。该方法在应用于重复和复杂的气体数据时,具有足够的鲁棒性,可以提供准确的结果。
Due to the limitation of existing gas-data and the low accuracy of gas classification,it is still difficult to solve the gas classification problem in practice.Although traditional pattern recognition algorithms can be applied to the gas classification problem,the achieved training and testing performance will decline on the gas-data with different concentrations,the accuracy is also relatively low.In this study,in response to classify the mixed gas collected by the 8-array sensor and the 24-array sensor,the authors propose to combine with the Principal component analysis(PCA)and apply the Support vector machine(SVM),to build a PCA-SVM model,which can be used for the classification of carbon monoxide(CO),methane(CH_(4)),hydrogen sulfide(H_(2)S)and ethanol(C_(2)H_(6)O) with different concentrations.After comparing with the results of the SVM model,the parameter-optimized Back-propagation-neural-network(BPNN)model and the PCA-BP-Neural-Network model,the PCA-SVM model can greatly improve the classification effect on the randomly selected data.In the gas-dataset with thirteen features,the accuracy achieved by the PCA-SVM model can reach 98.974%.In the gas-dataset with twenty-seven features,the accuracy achieved by the PCA-SVM model can reach 100%,which can meet the actual requirements for the classification of mixed gases.This method is robust enough to provide accurate results when applied to repetitive and complex gas-data.
作者
杨朝
张平平
汪国强
杜宝祥
YANG Zhao;ZHANG Pingping;WANG Guoqiang;DU Baoxiang(College of Electronic Engineering,Heilongjiang University,Harbin 150080,China;Department of Research and Development,Suzhou Huiwen Nano Technology Limited Company,Suzhou 215000,China)
出处
《黑龙江大学自然科学学报》
CAS
2022年第3期345-354,共10页
Journal of Natural Science of Heilongjiang University
基金
国家自然科学基金资助项目(51607059)
黑龙江省自然科学基金资助项目(QC2017059)。
关键词
气体分类
主成分分析
支持向量机
模式识别
gas classification
principal component analysis
support vector machine
pattern recognition
