基于荧光光谱的五维数据处理方法

石油是一种成分复杂的混合物,通过常规的检测方法很难对其进行定性识别。本文用汽、煤、柴油的混合物来模拟环境中的油类污染物。汽、煤、柴油在特定波长范围的激发下可以发出含有物质自身信息的荧光,根据朗伯-比尔定律可知荧光强度与物质浓度成正比,利用该性质对特定物质进行识别。通过FS920稳态荧光光谱仪对样本进行测量,将实验所得的三维数据拓展为五维数据,提出了一种将展开偏最小二乘耦合到残差四线性的五维数据处理方法,同时采用五维平行因子法和该算法分解数据,实现了对汽、煤油的定量分析,并恢复出了其激发和发射光谱。结果表明,展开偏最小二乘法的分析效果更好。

Processing Method of Five-dimensional Data Based on Fluorescence Spectra

As a complex mixture of components, petroleum is difficult to be qualitatively identified by conventional detection methods. In this paper, the mixture of gasoline, kerosene and diesel was used to simulate the oil pollutants in the environment. The gasoline, kerosene and diesel could emit fluorescence with the material self information under the excitation of the specific wavelength range, and the fluorescence intensity was directly proportional to the concentration of the substance from the Lambert-Beer law, which was used to identify the kind of the oil. These samples were measured by FS920 steady state fluorescence spectrometer. The data were added to five-dimensional array data by Savitzky-Golay method, then the fourth-order date that contained complex information is obtained to analyze applications. A fourth-order correction method, which coupled unfolded par-tial least-squares to residual quadrilinearization, was proposed to deal with the five-way data. In or-der to test its predictive ability, the parallel factor method was used as a reference. Both of them can retrieve the excitation and emission profiles from the test samples. However, the REP value shows that the new method has higher precision than parallel factor analysis.