反射光谱法快速检测润滑油污染浓度的回归模型

发布时间：2018-05-06 22:33

本文选题：反射光谱 + 润滑油　；参考：《光谱学与光谱分析》2017年07期

【摘要】：基于反射光谱技术原理研究润滑油的污染浓度预测模型。选用纯净润滑油和原态污染油为原料,采用等体积法逐一配置不同浓度的样本;调整好光纤探头与样品之间的距离及方位角并保持不变,采用脉冲氙灯为激发光源,利用波段在200~850nm范围的光谱仪对不同浓度的样本进行反射光谱测量实验。实验曲线观察分析表明,润滑油的污染浓度越高,反射光谱的强度越低。在采用变步长算法对原始实验数据进行稀疏采样的基础上,分别采用相关系数法、主成分分析法、主成分分析结合相关系数法(主成分-相关系数法)选取工作润滑油在220~780nm宽波段范围内的优势特征波长,并在所选取的优势波长处建立了润滑油污染浓度与光谱反射率的指数关系模型。测试结果表明,主成分-相关系数法选取的优势波长378.93nm,所建立的污染浓度与光谱反射率的指数回归模型能够较好地实现润滑油污染浓度大于0.06时的预测和估计。基于主成分-相关系数法选取优势特征波长所建立的指数关系回归模型,因混合介质及浓度的特异性条件,不服从低浓度均匀透明溶液中的朗伯-比尔定律,适用于较高污染浓度的润滑油质量状态估计,为进一步实现工作润滑油污染浓度的反射光谱法在线快速准确测定提供了可行性实验依据。
[Abstract]:Based on the principle of reflectance spectroscopy, the prediction model of lubricating oil pollution concentration is studied. The pure lubricating oil and the original polluted oil were selected as raw materials, the samples of different concentrations were allocated one by one by the equal volume method, the distance and azimuth between the optical fiber probe and the sample were adjusted and remained unchanged, and the pulsed xenon lamp was used as the excitation light source. The reflectance spectra of samples with different concentrations were measured by using a spectrometer in the 200~850nm range. The observation and analysis of the experimental curves show that the higher the concentration of lubricating oil, the lower the intensity of reflectance spectrum. Based on the sparse sampling of the original experimental data with variable step size algorithm, correlation coefficient method and principal component analysis method are used respectively. Principal component analysis (PCA) combined with correlation coefficient method (PCA) selects the dominant characteristic wavelength of working lubricating oil in 220~780nm wide band range. An exponential model of the relationship between lubricating oil pollution concentration and spectral reflectivity was established at the selected dominant wavelength. The test results show that the dominant wavelength of 378.93 nm selected by principal component correlation coefficient method and the exponential regression model of pollution concentration and spectral reflectivity can be used to predict and estimate the lubricating oil pollution concentration over 0.06. The exponential regression model based on selecting dominant characteristic wavelength by principal component correlation coefficient method was established. Because of the specific conditions of mixing medium and concentration, Lambert-Beer 's law from low concentration uniform transparent solution was not accepted. It can be used to estimate the quality state of lubricating oil with high concentration of pollution. It provides a feasible experimental basis for the on-line rapid and accurate determination of the contamination concentration of working lubricating oil by reflection spectrum method.
【作者单位】：燕山大学电气工程学院;
【基金】：国家自然科学基金项目(20577038) 河北省自然科学基金项目(D2004000195)资助
【分类号】：O657.3;TE626.3

【相似文献】