Optical fiber spectroscopy for measuring quality indicators of lubricant oils

Optical fiber spectroscopy for measuring quality indicators of lubricant oils
of 4
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
    Optical fiber spectroscopy for measuring quality indicators of lubricant oils A.G. Mignani a ∗ , L. Ciaccheri a , N. Díaz-Herrera  b , A.A. Mencaglia a , H. Ottevaere c , H. Thienpont c , S. Francalanci d , A. Paccagnini d , F. Pavone e   a CNR-IFAC – Sesto Fiorentino (FI), Italy  b Dpto. Óptica, Universidad Complutense de Madrid –Madrid, Spain c  Vrije Universiteit Brussel, Department of Applied Physics and Photonics – Brussels, Belgium d  MECOIL Diagnosi Meccaniche srl, – Firenze, Italy e  Dipartimento di Fisica, Università di Firenze – Sesto Fiorentino (FI), Italy ABSTRACT A collection of lubricant oils from different types of turbines, which were characterized by different degrees of degradation, were analyzed by means of UV-VIS-NIR absorption spectroscopy, fluorescence spectroscopy and scattering measurements. All these measurements were performed by means of optical fiber-based instrumentation that made use of LEDs or compact lamps for illumination and miniaturized spectrometers for detection. Multivariate data analysis was used to successfully correlate the wide optical spectral signature of lubricant oils to some of the most important  parameters for indicating the degree of degradation of the oil, such as TAN, JOAP-index, water content, and phosphorus.   Keywords:  Absorption spectroscopy, fluorescence spectroscopy, lubricant oil, scattering, optical fibers 1.   LUBRICANT OILS AND QUALITY INDICATORS Lubricant oil is a key element in the proper functioning of industrial machinery such as turbines, compressors, and  presses. The condition of the oil becomes irreversibly degraded due to machine operation, and needs to be monitored frequently so as to avoid structural damage to the machinery. Furthermore, lubricant oil status is an indication of the correct machinery operation, since any wrong functioning impairs the chemical-physical properties of the oil 1 . Currently, according to a regularly-programmed operational schedule, oil samples are manually drawn and sent to specialized laboratories, where they are analyzed by conventional analytical instrumentation. In order to minimize manual sampling and to reduce the number of costly laboratory analyses, a rapid online alert by means of low-cost sensors would be advisable. Spectroscopy techniques, especially fluorescence 2  and infrared absorption spectroscopy 3 , have been experimented. Molecular spectroscopy is frequently used to investigate the quality control parameters of  petroleum-based products, while fluorescence spectroscopy can assess oil viscosity and the integrated condition of oil. Moreover, multivariate data processing has been used to complement spectroscopic techniques, so as to achieve classification maps of oil degradation 4 . Recently, we have experimented scattered colorimetry followed by multivariate data processing in order to obtain an indication of the general chemical-physical status of lubricant oils, without seeking information about other quality indicator parameters 5 . This paper presents the measurement of some important quality parameters of lubricant oils, which was obtained by means of a combination of optical fiber spectroscopic techniques, such as UV-VIS-NIR absorption spectroscopy, fluorescence spectroscopy, and scattering measurements. All oils were previously measured in the MECOIL laboratories with the use of conventional techniques, and analytical data were available. The spectral data were processed by means of multivariate analysis so as to predict the quality parameters from the optical measurements. All instruments used for spectroscopy and scattering measurements used multimode optical fibers and compact sources and detectors, to ∗ ; phone 0039 055 522 6361 ; website 19th International Conference on Optical Fibre Sensors, edited by David Sampson, Stephen Collins, Kyunghwan Oh,Ryozo Yamauchi, Proc. of SPIE Vol. 7004, 70045R, (2008) 0277-786X/08/$18 doi: 10.1117/12.786663Proc. of SPIE Vol. 7004 70045R-1    demonstrate the feasibility of the online monitoring of lubricant oils, not only for assessing the general chemical-physical status, but also for the quantitative determination of key quality parameters. Among those reported in the MECOIL analytical forms, four quality indicators showed a good fit with the data coming from the multiple optical measurements. These were: -   Total Acid Number (TAN). It measures the concentration of acid groups in oil, and is determined via a  potentiometric or colorimetric titration. The acidity of an oil increases as it is degraded by continuous use. The excessive acidity induces the formation of varnish or sludge 6 . -   Water  . The presence of water in oil is usually due to defective gaskets, which allow water to infiltrate the oil circuit. Water produces oxidative phenomena (rust), increasing wear and causing, in the long run, turbine damage. -    Phosphorus . This element is introduced in oils as an anti-wear additive. -    JOAP ‘anti-wear’ index  (JOAP-index). This is a quality parameter that was introduced by the Joint Oil Analysis Program. It is derived from FTIR spectral measurements at 1022 and 960 cm -1 , where the phosphoric group contained in many anti-wear additives absorbs. It is strongly correlated to the phosphorus content, especially in new oils, when the anti-wear additive has not yet been consumed. The interaction between the anti-wear additive and the metal surface often involves physical or chemical adsorption of the anti-wear molecule 7 . 2.   OPTICAL FIBER SPECTROSCOPY As shown in Figure 1, absorption spectroscopy in the entire UV-VIS-NIR spectral range, fluorescence spectroscopy and scattering measurements have all been used for the characterization of lubricant oils. The instrumentation for absorption spectroscopy made use of an optical fiber-compatible lamp as source (Micropack Inc., DH-2000-BAL), an optical fiber spectrometer as detector (Instrument Systems, model Spectro 320), and a holder for standard quartz cuvette containing the lubricant oil sample. Two standard multimode optical fiber strands, with a 200 µm core diameter, were used for connecting both source and detector to the cuvette holder, which was equipped with collimating and focussing optics so as to perform transmission measurements efficiently. The instrumentation for fluorescence spectroscopy made use of an LED source emitting at 404 nm and an Ocean Optics spectrometer (model USB4000) for detection. Front-face fluorescence spectroscopy was achieved by means of a  bifurcated optical fiber bundle, the common end of which was dipped in the lubricant oil being tested, while the individual strands were coupled to LED and spectrometer, respectively. The lubricant oil was contained inside a small glass vial, the cap of which was modified for compatibility with the end face of the optical fiber bundle. The fluorescence peak showed an almost linear temporal decay. The shape of the time-decay curve was characteristic of each sample. Consequently, the lubricant oils were characterized by means of both fluorescent spectra and fluorescence time-decay curves. The instrumentation for the scattering measurements made use of an LED emitting at 650 nm and two detectors: one for transmission measurements and one for scattering measurements at 60°. This particular wavelength-angle combination was chosen from previous scattering measurements performed at many angles and wavelengths, as it was found to be the most convenient one for an optimized signal-to-noise ratio. Standard multimode optical fiber strands, with a 200 µm core diameter, were used for connecting the source and the detectors to the glass vial containing the lubricant oil sample. The optical fibers ended with GRIN µ -lenses for beam collimation. 3.   EXPERIMENTAL RESULTS A collection of 27 lubricant oils was considered. These samples came from three different types of turbines: 6 samples were from an aeronautic-type gas turbine (TGAD), 6 samples were from a steam turbine (TV), and 15 samples were from a gas turbine (TG). They were both mineral-paraffin and polyester-synthetic oils belonging to various brands, such as Mobil, AGIP, Chevron, Shell, and Castrol. All the oils had been previously analyzed in the Mecoil laboratory using conventional techniques. The results of the absorption and intensity fluorescence measurements, which were achieved by means of optical fiber- based instrumentation, are shown in Figure 2. The time-decay curves were almost linear in the 0 ÷ 500 sec range, with slopes in the -33 ÷ +12 counts/sec range, mean value of -9 counts/sec, and standard deviation of 12 counts/sec. The scattering values varied in the 3.5 ⋅ 10 -5  ÷ 4.0 ⋅ 10 -4  range, with mean value of 1.3 ⋅ 10 -4  and standard deviation of 9.6 ⋅ 10 -5 .  These data were processed as follows:  Proc. of SPIE Vol. 7004 70045R-2    1.   The spectra were smoothed by using the boxcar integration. The boxcar width was 11 points for fluorescence spectra and 3 points for absorption spectra. The resulting spectral resolutions were 2.2 nm for fluorescence spectroscopy, and 4.1 nm for absorption spectroscopy, respectively. 2.   The data were compressed by means of Principal Component Analysis (PCA) 8 . Four PCA models were  produced: one for the normalized fluorescence spectra, one for the fluorescence decay curves, one for the UV-VIS absorption (290-800 nm), and one for the NIR absorption (1100-1690 nm). The absorption spectra were split because the UV absorption bands had a much higher amplitude and variance than the NIR bands. 3.   The score matrices of the four PCA models were concatenated along the 2 nd  dimension to form a joint data-matrix. One further variable was added that came from the scattering data. 4.   A fit of various quality parameters of lubricant oil was attempted using Orthogonal Signal Correction (OSC) followed by Partial Least Square (PLS) regression 9,   10 . The OSC filter was first applied in order to remove the interference due to the systematic noise, i.e. the data variation not linked to the target variable. One PLS factor was then extracted for prediction purposes. The columns of the joint data matrix were auto scaled (mean = 0, variance = 1) before starting the OSC + PLS regression, thus removing the difference of size among these non-homogeneous variables. Table I summarizes the spectral bands used for PCA, the number of PC extracted, and the correspondent explained variance. Table II summarizes the results of PLS regression for the selected quality indicators. The goodness of fit is expressed by the determination coefficient (R  2 ) and by the standard error of cross validation (SECV). The correlation  between the conventionally-measured data and the predicted data was found to be extremely satisfactory for the JOAP-index and for the phosphorous content, and was very good also for the water content and TAN. Figure 1. Experimental setup for lubricant oil characterization by means of optical fiber instrumentation: scattering measurements (left), fluorescence spectroscopy (top), and wide range absorption spectroscopy (bottom).   4.   PERSPECTIVES Absorption spectroscopy in the UV-VIS-NIR spectral range, together with fluorescence spectroscopy and scattering measurements, complemented by a multivariate data processing, were used to predict important quality parameters of lubricant oils, such as TAN, water, phosphorus content and the JOAP-index. These results open up the possibility of implementing an online device for the continuous assessment of oil condition, given that all measurements were  performed by means of compact optical fiber-based instruments.   ACKNOWLEDGMENTS 'This work was partially supported by a grant within the program ‘Profesores de la UCM en el extranjero’ and a Community of Madrid  project, FUTURSEN (S-0505/AMB-0374).The European Network of Excellence ‘NEMO’ is also acknowledged for partial financial support. The authors would like to thank Mr. Franco Cosi and Mr. Daniele Tirelli for technical assistance. Proc. of SPIE Vol. 7004 70045R-3    Figure 2. Spectral characterization of the lubricant oil collection: absorption spectra in the UV-VIS-NIR range (left), and intensity fluorescence measurements (right). Data type Spectral band (nm) # PC Explained variance Fluorescence spectrum 420 – 700 4 99% Fluorescence temporal decay -- -- 2 100% UV-VIS absorption spectrum 290 – 800 4 97%  NIR absorption spectrum 1100 – 1690 3 99.5% Table I. Spectral characteristics used for multivariate data analysis, and PCA specifications. Quality parameter R  2  SECV Water content 0.843 56.8 JOAP-index 0.973 10.1 TAN 0.739 0.1 Phosphorus 0.969 527 Table II. PLS regression characteristics of the lubricant oil’s quality indicators . REFERENCES 1    Basic Handbook of Lubrication , Society of Tribologists and Lubrication Engineers – Alberta Section, 2003. 2  T.D. Downare, O.C. Mullins, Visible and near-infrared fluorescence of crude oils,  Appl. Spec. , vol. 49, 1995, pp. 754-764. 3  M. Wiseman, A. Ah-Sue, Monitoring oil degradation using FTIR analysis,  Lubr. Eng. , vol. 48, 1992, pp. 236-241. 4  C.M. Stellman, K.J. Ewing, F. Bucholtz, I.D. Aggarwal, Monitoring the degradation of a synthetic lubricant oil using infrared absorption, fluorescence emission and multivariate analysis: a feasibility study,  Lub. Eng. , vol. 55, 1999, pp. 42-52. 5  A.G. Mignani, L. Ciaccheri, S. Francalanci, A. Paccagnini, F.S. Pavone, M. Galimberti, Monitoring of lubricant oil degradation using fiber-optic scattered colorimetry, OSA Technical Digest ISBN 1-55752-B16-0, 18 th  Int. Conf. on Optical Fiber Sensors , A.Mendez Ed., 2006, file WB3.pdf. 6  C.V. Ossia, H. Kong, L.V. Markova, N.K. Myshkin, On the use of intrinsic fluorescence emission ratio in the characterization of hydraulic oil degradation, Trib. Int. , vol. 41, 2008, pp. 103-110. 7  W. Huang, Y. Tan, B. Chen, J. Domg, X. Wang, The binding of antiwear additives to iron surface: quantum chemical calculations and tribological tests, Trib. Int. , vol. 36, 2003, pp. 163-168. 8  M.J. Adams, Chemometric in Analytical Spectroscopy , Royal Society of Chemistry, 1995, Cambridge UK. 9  E. Bertran, H. Iturriaga, S. Maspoch, I. Montoliu, Effect of orthogonal signal correction on the determination of compounds with very similar near infrared spectra,  Anal. Chim. Acta , vol. 431, 2001, pp. 303-311. 10  S. Wold, M. Sjostrom and L. Erikkson, PLS-regression: a basic tool for chemometrics, Chem. Intel. Lab. Sys. , vol. 58, 2001, pp. 109-130.   4505005506006500.511.522.5x 10 -5 Wavelength ( nm )    N  o  r  m  a   l   i  z  e   d   U  n   i   t  s TGTVTGAD4006008001000120014001600012345Wavelength ( nm )    A   b  s  o  r   b  a  n  c  e TGTVTGAD Proc. of SPIE Vol. 7004 70045R-4
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks