不同分辨率近红外光谱对汽车差速器油品牌鉴别的研究

发布时间：2018-02-06 02:29

本文关键词： 差速器油分辨率特征波数连续投影算法　出处：《光谱学与光谱分析》2015年07期 　论文类型：期刊论文

【摘要】：研究了不同分辨近红外光谱对汽车差速器油品牌鉴别的效果与影响。分别采集了五种不同品牌汽车差速器油在4,8,16和32cm-1分辨率下的透射光谱。去除采集波段范围两端噪声明显部分后采用10522.28~4 443.425cm-1范围光谱进行分析。不同分辨率光谱的主成分分析(principal component analysis,PCA)结果表明不同品牌差速器油可以被鉴别。同时基于不同分辨率光谱全谱所建立的偏最小二乘判别分析(partial least squares-discriminant analysis,PLS-DA)以及支持向量机(support vector machine,SVM)模型所获得的判别正确率相近且均高于90%。其中以8cm-1分辨率下PLS-DA与SVM模型的判别正确率最高。上述研究结果表明分辨率对判别分析结果的影响较小。进一步对不同分辨率下的近红外光谱采用连续投影算法(successive projections algorithm,SPA)选择特征波数,发现不同分辨率下得到的特征波数均不相同,表明分辨率影响特征波数的选择。以不同分辨率特征波数建立的PLS-DA以及SVM模型取得了较好的判别分析效果,与以不同分辨率下全谱建立的判别分析模型效果相当。研究结果表明,分辨率对差速器油品牌鉴别影响较小,而对特征波数的选择影响较大,在实际应用中应考虑到分辨率对特征波数选择的影响。总体而言,不同分辨率的近红外光谱全谱与对应的特征波数都能有效的实现对汽车差速器油品牌的鉴别。
[Abstract]:The effects and effects of different resolution near infrared spectra on the brand identification of automobile differential oil were studied. Transmission spectra at resolution 16 and 32 cm-1. After removing the obvious noise at both ends of the range of the acquisition band, 10522.284 is used. The spectrum of 443.425 cm ~ (-1) was analyzed. Principal component Analysis (PCA) of different resolution spectra (. Principal component analysis. The results of PCA show that different brands of differential oils can be identified. At the same time, the partial least squares discriminant analysis based on the spectral full spectrum of different resolutions (. Partial least squares-discriminant analysis. PLS-DAA and support vector machine support vector machine. SVM). The discriminant accuracy obtained by the model is similar and higher than that of 90. The PLS-DA and SVM models with 8cm ~ (-1) resolution have the highest discriminant accuracy. The results show that the discriminant analysis is based on the resolution analysis. The influence of the results is small. Furthermore, the continuous projection algorithm is used for the near infrared spectra with different resolution (. Successive projections algorithm. Spa) selects the characteristic wavenumber and finds that the characteristic wavenumber obtained at different resolution is different. The results show that the resolution affects the choice of the characteristic wavenumber. The PLS-DA and SVM models based on the different resolution characteristic wavenumber have good discriminant analysis effect. The results show that the resolution has little effect on differential oil brand identification, but has a greater impact on the choice of characteristic wave number. In practical application, the influence of resolution on the choice of characteristic wavenumber should be taken into account. In general, the full spectrum of near-infrared spectrum with different resolution and the corresponding characteristic wave number can effectively identify the brand of automobile differential oil.
【作者单位】：浙江经济职业技术学院;浙江大学生物系统工程与食品科学学院;
【基金】：国家支撑计划项目(2014BAD10B02)资助
【分类号】：U473.74;TN219

【参考文献】