摘要
提出了一种基于可见-近红外光谱分析技术快速鉴别汽车制动液品牌的新方法。采用美国ASD公司的便携式光谱仪对五种不同品牌的制动液进行光谱分析,各获取60个样本数据。采用平均平滑法和标准归一化方法对样本数据进行预处理,再对光谱数据进行主成分分析,建立第一主成分和第二主成分的二维散点图,表明不同品牌制动液具有较好的聚类特性。将前6个主成分作为输入量,制动液品牌作为输出量,建立了基于逐步判别分析法的鉴别模型。随机抽取225个样本用于建模,余下的75个样本用于模型验证。试验结果表明验证准确率达到94.67%,说明所提出的方法具有很好的分类和鉴别作用,为制动液品牌的快速鉴别提供了一种新方法。
A new method was developed to fast discriminate brands of brake fluid by means of visual-near infrared spectroscopy. Five different brands of brake fluid were analyzed using a handheld near infrared spectrograph, manufactured by ASD Company, and 60 samples were gotten from each brand of brake fluid. The samples data were pretreated using average smoothing and standard normal variable method, and then analyzed using principal component analysis （PCA）. A 2-dimensional plot was drawn based on the first and the second principal components, and the plot indicated that the clustering characteristic of different brake fluid is distinct. The foregoing 6 principal components were taken as input variable, and the band of brake fluid as output variable to build the discriminate model by stepwise discriminant analysis method. Two hundred twenty five samples selected randomly were used to create the model, and the rest 75 samples to verify the model. The result showed that the distinguishing rate was 94. 67%, indicating that the method proposed in this paper has good performance in classification and discrimination. It provides a new way to fast discriminate different brands of brake fluid.
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2008年第6期1296-1299,共4页
Spectroscopy and Spectral Analysis
基金
“十一五”国家科技支撑项目(2006BAD10A0403)
高等学校优秀青年教师教学科研奖励计划项目(02411)资助
关键词
制动液
光谱分析
主成分分析
逐步判别分析
品牌鉴别
Brake fluid
Spectral analysis
Principal component analysis
Stepwise discriminant analysis
Brand discrimination
