基于激光诱导水拉曼抑制法的油膜厚度测量方法研究
陈宇男1,2,3, 杨瑞芳1,3,*, 赵南京1,3,*, 祝玮1,2,3, 陈晓伟1,2,3, 张瑞琦1,2,3
1.环境光学与技术重点实验室, 中国科学院安徽光学精密机械研究所, 安徽 合肥 230031
2.中国科学技术大学, 安徽 合肥 230026
3.安徽省环境光学监测技术重点实验室, 安徽 合肥 230031
*通讯作者
摘要

为了实现水体表面油膜厚度的快速测量分析, 以266 nm的激光作为探测系统的激发光源, 基于激光诱导水拉曼散射光谱检测技术, 通过获取不同种类不同厚度油膜存在下水拉曼光谱信息, 建立油膜厚度反演模型。采用高斯函数拟合法校正了荧光光谱对拉曼光谱的干扰。 然后根据水拉曼抑制法结合非线性最小二乘优化算法, 建立油膜厚度反演模型。结果表明: 对92#汽油、 0#柴油、 美孚机油20w-40、 壳牌润滑油10w-40、 采埃孚变速箱油AG6和原油油膜能探测到的油膜厚度范围为0.19~379.22 μm。 采用水拉曼光谱-油膜厚度反演模型预测油膜厚度的平均相对误差在8.14%~15.81%之间。 该方法能实现实验室条件下对微米级油膜的测量。

关键词: 油膜厚度; 拉曼光谱; 快速检测
中图分类号:O657.3 文献标识码:A
Research on Measuring Oil Film Thickness Based on Laser-Induced Water Raman Suppression Method
CHEN Yu-nan1,2,3, YANG Rui-fang1,3,*, ZHAO Nan-jing1,3,*, ZHU Wei1,2,3, CHEN Xiao-wei1,2,3, ZHANG Rui-qi1,2,3
1. Key Laboratory of Environmental Optics and Technology, Anhui Institute of Optics and Fine Mechanics, Chinese Academy of Sciences, Hefei 230031, China
2. University of Science and Technology of China, Hefei 230026, China
3. Key Laboratory of Optical Monitoring Technology for Environment, Hefei 230031, China
*Corresponding authors e-mail: rfyang@aiofm.ac.cn; njzhao@aiofm.ac.cn
Abstract

In this work, using a 266 nm laser as the excitation light source of the detection system, the water Raman spectra under the oil film of different thicknesses were obtained based on the laser-induced water Raman spectroscopy technology. The Gaussian function fitting method was used to correct the interference of the fluorescence spectrum on the Raman spectrum. And then, the oil film thickness inversion model could be established according to the water Raman suppression method combined with the nonlinear least square optimization algorithm. The results show that the six oils’ detectable oil film thickness range (92# gasoline, 0# diesel, Mobil oil 20w-40, Shell Helix10w-40, Lifeguard Fluid AG6 and crude oil) is from 0.19 to 379.22 μm. The average relative error of oil film thickness prediction ranges from 8.14% to 15.81% according to the water Raman spectrum-oil film thickness inversion model. This method can realize the measurement of micron-level oil film under laboratory conditions.

Key words: Oil film thickness; Raman spectroscopy; Rapid detection
Introduction

The frequent oil spills have caused marine environment pollution and threatened ecological safety. When oil spill accidents occur, the rapid acquisition of oil spill information is an important part ofthe emergency treatment process, which also has great significance in protecting the marine environment. The oil spill volume calculation includes the oil spill area measurement and the oil film thickness.Currently, the existing remote sensing technology can measure the oil spill area, but the measurement of the oil film thickness is still in the preliminary research stage.

In recent years, the main methods for measuring oil film thickness include the optical method[1, 2], acoustic method[3], electrical method[4], and microwave radar[5]. Optical methods such as hyperspectral and laser triangulation methods. Foundation[6] used the airborne visible/near-infrared imaging spectrometer (AVIRIS) to study the oil spill on the coast of Santa Barbara in the United States at 380~2 500 nm. They found that the increase in the thickness of the oil film leads to an increase in light absorption and a decrease in reflectivity, which provided the possibility of detecting the thickness of the oil film. Subsequently, the scientists analyzed the reflectance spectra of the changes in the thickness of the oil film on the sea surface. The oil film was divided into thick and thin oil film, and the reflection enhancement effect of thin oil film was studied. It was found that the spectral reflectance of the oil film had a negative power function relationship with the thickness. The green-light and red-light bands were used as the selected bands for oil spill hyperspectral remote sensing detection[7, 8]. These studies provided technical support for oil film identification and quantitative inversion. However, the hyperspectral data were collected in multiple spectral bands, resulting in redundancy. Therefore, it was necessary to perform dimensionality reduction and denoising processing on the data, which would increase the complexity of data processing. Laser triangulation usesthe principle of imaging ranging for measurement. The oil film was approximated as a transparent layer, and the laser was focused and projected onto the object’ s surface through a converging lens. The reflected or scattered light was imaged on the detector PSD or CCD through the imaging lens, and the imaging displacement was subtracted to obtain the oil film thickness[9, 10, 11, 12]. Wu Di studied the oil film thickness sensor using the vertical incidence differential laser triangulation method[13]. The measuring range was from 0.1 to 10 mm, but the measuring device needed to be placed in a buoy, which made it difficult to carry out a wide range of measurements. Acoustic methods, such as laser ultrasound, use high-energy pulsed lasers to irradiate the oil film. The Doppler vibrometer was used to measure the propagation time of sound waves in the oil film to obtain the oil film thickness. Li Yibo et al. realized oil film thickness measurement in the range of 0.64~15.738 mm under laboratory conditions[14]. Environment Canada has developed a laser ultrasonic oil film thickness telemetry sensor system. However, the system was expensive, and the price of a high-energy carbon dioxide pulsed laser was about one million yuan [15]. The electrical method, such as the antenna resistance method, measured the oil film thickness with the inverse relationship between the antenna resistance and the oil film thickness. The detection range was 1 to 18 mm, and the detection limit can reach 0.25 mm[16]. Microwave radar, such as Synthetic Aperture Radar (SAR), isan active phased sensor. SAR data contains rich polarization feature information and texture information.It had the characteristics of long measuring distance and high spatial resolution. However, the classification information of oil film thickness was not accurate enough, which would make it difficult to estimate the oil film thickness[17] accurately.

This work selects six kinds of oil as the research objects. The oil film thickness is calculated using the oil film’ s water Raman suppression effect. A quantitative inversion model of oil film thickness is established based on the nonlinear least square algorithm. Relative error (REP) and average relative error are selected to evaluate the model’ s accuracy.

1 Materials and methods
1.1 Sample preparation

92# gasoline, 0# diesel, Mobil oil 20w-40, Shell Helix 10w-40, Lifeguard Fluid AG6 and crude oil were selected to prepare the oil film. Sodium chloride was added to the deionized water to prepare the simulated seawater with a salt content of 34 g· L-1. A pipette was used to drop different volumes of oil onto the surface of simulated seawater in a glass beaker. The added oil film thickness was calculated according to the formula V=Sd, where V represented the volume of the oil product, S represented the area of the generated oil film, and d represented the thickness of the oil film.

1.2 Experimental set-up

In order to realize the measurement of the oil film spectrum, a laser-induced water Raman detection system was designed and built. Figure 1 was a schematic diagram of the detection system. It was mainly composed of a laser transmitter, an optical coupling collection system and a spectrum detection device. The control and spectrum processing processes were carried out through the computer. The laser was a quadruple frequency solid laser with a wavelength of 266 nm, an energy of 100 mJ, a pulse time of 6 ns, and a laser beam diameter of 9 mm. The fiber aperture was 500 μ m. The spectrometer was a Maya-2000pro spectrometer, whose measuring range was 200~1 100 nm.

Fig.1 Schematic diagram of the detection system

The system’ s working process was as follows: the control system sends a trigger signal to trigger the Nd:YGA laser to emit a 266 nm laser, which acts on the oil film in a 45° direction via a mirror. Based on this, oil emitted a fluorescence spectrum signal, and water was stimulated to generate a Raman spectrum signal. The optical fiber coupling system effectively collected the generated signals and transmitted them to the fluorescence detection system. Finally, the spectrometer transmitted the signals to the computer.

1.3 Principle of the method

When the laser irradiates the oil film, the oil is excited to emit a fluorescence spectrum signal, and the water also produces a background fluorescence spectrum signal and a water Raman scattering signal. The spectrum signals detected when the water surface is covered with oil film and without oil film are shown in Figure 2. Among them, the dotted line represents the spectral signal received when there is an oil film on the water surface, and the solid line represents the spectral signal received when there is no oil film. It can be seen that under the excitation light of 266 nm, water produces Raman scattering in the range of 285~300 nm, and the center of the Raman peak is located at 291 nm.

Fig.2 Seawater background fluorescence spectrum, seawater Raman spectrum and oil film fluorescence spectrum

The spectral signal received by the detector can be expressed as:

Ki=ηiP0{1-exp[-(ke+ki)d]}+δΨiP0exp[-(ke+ki)d]+ζiP0exp[-(ke+ki)d](1)

The η i represents the oil film’ s fluorescence conversion efficiency at the i-th wavelength. P0 represents the intensity of excitation light; ke and ki respectively represent the extinction coefficient of the oil film at the emission wavelength and the i-th wavelength; drepresents the thickness of the oil film; ζ i represents the fluorescence conversion efficiency of seawater; Ψ i represents the water Raman scattering conversion coefficient; δ represents the δ function.

It can be seen from equation (2) that the received optical signals from seawater are the fluorescence spectrum signal of oil ϕ i, o, the background fluorescence spectrum signal of seawater ϕ i, b and the optical signal of seawater Raman scattering spectrum Ki, r. Therefore, the expression of the optical signal received by the detector at the i-th wavelength is as follows:

Ki=ϕi, o+ϕi, b+Ki, r(2)

When there is no oil film on the water surface, the received signal Ji by the detector can be expressed as:

Ji=δirΨiP0+ζiPo(3)

By transforming equations (2) and (3), and combining the Lambert-Beer attenuation law, the following equations can be obtained:

Ki-(ϕi, o+ϕi, b)Ji-ζiPo=exp[-(ke+ki)d](4)

When on substance can emit fluorescence in the water, the above formula can also be expressed according to the Lambert-Beer attenuation law as:

Ki-ϕi, oJi=exp[-(ke+ki)d](5)

Putting the wavelength signal at the Raman peak into equation (5), the above formula can also be expressed as:

d=-1ke+krlnKr-ϕr, 0Jr(6)

Through simplification, the expression of the oil film thickness d is as follows:

R'=Bexp[-Ad](7)

Where R'=Kr-ϕ r, o represents the Raman signal received when there is an oil film on the water surface, that is, Kr minus the fluorescence signal of the oil film at the wavelength λ r; B=Jr represents the water Raman signal received when there is no oil film on the water surface. A=(ke+kr), a quantitative inversion model of oil film thickness is established by a nonlinear least squares algorithm, and relative error (REP) and average relative error are selected to evaluate the model’ s accuracy.

2 Results and discussion
2.1 Analysis of oil spectral characteristics

The spectra of 92# gasoline, 0# diesel, Motor oil 20w-40, Shell Helix 10w-40, Lifeguard Fluid AG6 and crude oilare shown in Figure 3. The water Raman signal in the presence of different types of oil films is extracted.

Fig.3 Oil film spectra of different thicknesses obtained by tie detection system

It can be seen from Figure 3 that the measurement range of the six oil film thicknesses is 0.19~379.22 μ m, the fluorescence spectrum of the oil is distributed in the range of 270~530 nm, and the fluorescence intensity increases with the increase of the oil film thickness. The Raman spectrum of water is in the range of 280~300 nm. When the water surface is covered with 0# diesel, Motor oil 20w-40, Shell Helix 10w-40, Lifeguard Fluid AG6 and crude oil, the intensity of the water Raman spectrum decreases as the thickness of the oil film increases. When the water surface is covered with 92# gasoline, the water Raman intensity increases continuously with the oil film thickness. Because the fluorescence spectrum of gasoline has a fluorescence peak in the range of 270~310 nm, it will interfere with the water Raman spectrum signal.

The water Raman spectrum and the oil fluorescence spectrum have a certain range of spectral overlaps. When the method of water Raman suppression is used to measure the oil film thickness, it is necessary to extract the interpolation data of the oil film fluorescence spectrum at the water Raman signal and to make the intensity difference with the water Raman signal value.

2.2 Extraction of water Raman signals

In order to eliminate the interference caused by the oil fluorescence spectrum, an algorithm for subtracting oil film fluorescence background was written based on Matlab software. The process of subtracting the oil film fluorescence background and extracting the water Raman signal is shown in Figure 4. Fig.4(a) shows the fluorescence spectrum of 0# diesel and the water Raman spectrum generated by excitation at 266 nm. Fig.4(b) shows the fluorescence spectrum of 0# diesel after subtracting the spectral signal in the range of 280~300 nm. Fig.4(c) shows the fluorescence spectrum fitted by the Gaussians. Fig.4(d) shows the water Raman spectrum after subtracting the fluorescence background. It clearly shows that the fluorescence background is effectively removed.

Fig.4 Flow chart of water Raman signal extraction in the presence of 0# diesel oil film

The Gaussian function fitting method is used to fit the oil film spectrum. The fitting results of the six oils, 92# gasoline, 0# diesel, Mobil oil 20w-40, Shell Helix 10w-40, Lifeguard Fluid AG6 and crude oilare shown in Figure 5. It can be seen from Figure 5 that the fitting correlation coefficients of the fluorescence spectra of the six oils using the Gaussian function fitting method are all greater than 0.99, which can accurately fit the shape characteristics of the fluorescence spectra of the oil film.

Fig.5 The fitting results of the Gaussian function fitting method to the oil film spectrum
(a): 92# Gasoline; (b): Lifeguard Fluid AG6; (c): 0# Diesel; (d): Motor oil 20w-40; (e): Shell Helix 10w-40; (f): Crude oil

2.3 Inversion analysis of oil film thickness by water Raman suppression method

The water Raman strength-oil film thickness fitting curve was established based on the nonlinear least squares algorithm. The fitting curves of the six oils, 92# gasoline, 0# diesel, Motor oil 20w-40, Shell Helix 10w-40, Lifeguard Fluid AG6 and crude oil are shown in Figure 6. The oil film thickness fitting curve parameters of the six oils are shown in Table 1. It can be seen that the water Raman signal intensity gradually decreases with the increase of the oil film thickness. The correlation coefficients of the water Raman intensity-oil film thickness fitting curve are all greater than 0.95.

Fig.6 Fitting curve of water Raman Intensity-oil film thickness
(a): Shell Helix 10w-40; (b): 0# Diesel; (c): Motor oil 20w-40; (d): Crude oil; (e): Lifeguard Fluid AG6; (f): 92# Gasoline

Table 1 Fitting curve parameters of six oils

In order to further analyze the quantitative inversion capability of the water Raman suppression method for oil film thickness, six oils of different thicknesses were generated on the water surface, and the accuracy of the oil film thickness inversion model was evaluated according to REP and average relative error. The inversion results of the oil film thickness of the six oils are shown in Table 2 to Table 7.

Table 2 Quantitative inversion results of the oil film thickness of Motor oil 20w-40
Table 3 Quantitative inversion results of the oil film thickness of 0# diesel
Table 4 Quantitative inversion results of the oil film thickness of Shell Helix 10w-40
Table 5 Quantitative inversion results of the oil film thickness of 92# gasoline
Table 6 Quantitative inversion results of the oil film thickness of Lifeguard Fluid AG6
Table 7 Quantitative inversion results of the oil film thickness of crude

It can be seen from table 2 that the relative prediction error of oil film thickness of Motor Oil 20w-40 obtained by establishing oil film thickness inversion model based on the water Raman suppression method and nonlinear least square algorithm is between 1.68%~19.36%, and the average relative error is 8.14%.

It can be seen from Table 3 that the relative error of the oil film thickness prediction of 0# Diesel obtained by establishing the oil film thickness inversion model based on the water Raman suppression method and nonlinear least square algorithm is between 1.06%~20.24%. The average relative error is 10.67%.

It can be seen from Table 4 that the relative error of the oil film thickness prediction of Shell Helix 10w-40 obtained by establishing the oil film thickness inversion model based on the water Raman suppression method and nonlinear least square algorithm is between 3.15%~17.54%, and the average relative error is 9.94%.

It can be seen from Table 5 that the relative error of the oil film thickness prediction of 92# Gasoline obtained by establishing the oil film thickness inversion model based on the water Raman suppression method and nonlinear least square algorithm is between 4.36%~30.26%, and the average relative error is 15.81%.

It can be seen from Table 6 that the relative error of the oil film thickness prediction of Lifeguard Fluid AG6 obtained by establishing the oil film thickness inversion model based on the water Raman suppression method and nonlinear least square algorithm is between 3.93%~21.36%, and the average relative error is 11.57%.

It can be seen from Table 7 that the relative error of the oil film thickness prediction of Crude oil obtained by establishing the oil film thickness inversion model based on the water Raman suppression method and nonlinear least square algorithm is between 2.51%~22.36%, and the average relative error is 10.43%.

3 Conclusion

This work studied the oil film thickness measurement based on the laser-induced water Raman spectrum method. The fluorescence background of the oil was subtracted by Gaussian fitting, and the water Raman signal was extracted. Then the non-linear least squares algorithm was used to establish a quantitative inversion model of water Raman intensity- oil film thickness, and the oil film thickness of six oils could be predicted. The results showed that the measurement range of the oil film of the six oils by this method was between 0.19~379.22 μ m. The average relative errors of the six oil, 92# gasoline, 0# diesel, Motor oil 20w-40, Shell Helix 10w-40, Lifeguard Fluid AG6 and crude oil were 15.81%, 10.67%, 8.14%, 9.94%, 11.57%, and 10.43%, respectively.This method can realize the quantitative inversion of the oil film on a micron scale.

参考文献
[1] Sun Changsen, Yu Longcheng, Sun Yuxing, et al. Applied Optics, 2005, 44(25): 5202. [本文引用:1]
[2] Peterson J P, Peterson R B. Applied Optics, 2006, 45(20): 4916. [本文引用:1]
[3] Brown C E, Fingas M F. Marine Pollution Bulletin, 2003, 47: 485. [本文引用:1]
[4] Sabah A Abdul. Electroanalysis, 2006, 18(21): 2148. [本文引用:1]
[5] Singhal, S, Dadhich, et al. Indian Journal of Radio & Space Physics, 2013, 42(1): 52. [本文引用:1]
[6] Foudan M F S. Fairfax: George Mason University, 2003. [本文引用:1]
[7] LU Ying-cheng, TIAN Qing-jiu, QI Xiao-ping. Spectroscopy and Spectral Analysis, 2009, 29(3): 986. [本文引用:1]
[8] Wettle M, Daniel P J, Logan G A, et al. Remote Sensing of Environment, 2009, 113(9): 2000. [本文引用:1]
[9] Kukhtarev N, Kukhtarev T, Gallegos S C. Applied Optics, 2011, 50(7): B53. [本文引用:1]
[10] Qieni L, Baozhen G, Wenda Y, et al. Optics & Lasers in Engineering, 2011, 49(1): 13. [本文引用:1]
[11] Qieni, Lu L, Ge B, et al. Journal of Modern Optics, 2012, 59(11): 947. [本文引用:1]
[12] Baozhen G, Jingbin S, Pengcheng L, et al. Review of Scientific Instruments, 2013, 84(1): 2148. [本文引用:1]
[13] Wu Di, Lv Qieni, Chen Xi. Journal of Tianjin University, 2013, (11): 52. [本文引用:1]
[14] Li Yibo, Qi Xiang, Wang Huifang. Nanotechnology and Precision Engineering, 2017, 15(3): 159. [本文引用:1]
[15] Christian Néron, Padioleau C, Daniel Lévesque, et al. Optical Instrumentation for Energy & Environmental Applications. 2013. [本文引用:1]
[16] Li Xinghua, Xi Meng, Zhao Yu. Nanotechnology and Precision Engineering, 2016, 14(2): 106. [本文引用:1]
[17] Solberg A H S. Proceedings of the IEEE, 2012, 100(10): 2931. [本文引用:1]