8290614314

7. Opublikowane badania własne

7. Opublikowane badania własne

B4

J. Onet et ol / Totonfo tOI (2012) 78-84

Table 3

Robustness of proposed method calculated for three conceniration levels tested after 0 h . 48 h. and 96 h. Mcan valurs were calculatrd for three laboratory rrplicates. uncertalnty calculated for 95X confldence lntrrval. PLS models were constmcted with eight facton

Coocrntratioa (mg* l' ^ł|	Time |h|	PLS SY124			SR 19
		Mean [mg L '|	RSD|X|	± l«niL *|	Mean |agL-'|	RSD(X|	X l“fL *1
4.000	cond.0 48	3559 3233	3.65	0.360	3541	1.92	0.164 0.3S0
	96	3.410	2.33	0242	3.917	1.65	0.197
5.000	cond. 0	4J94	3.25	0.435	5.129	1.24	0.194
	48	4727	3.27	0.471	5.076	3.29	0.663
	%	4.518	4.37	0505	5.017	3.69	0564
0.000	cond.0	62)27	3.31	0.607	6282	4.42	0.846
	48	5.647	2.26	0.388	6.100	4.21	0.967
	96	5.787	4.82	0549	5.996	352	0.644

are 0.9492 and 1.053, respectively. The nuli hypothesis is rejected when the p-values are smaller chan a definite significant level value. a. of 0.01 or 0.05. For the studied case. the nuli hypothesis of no significant differences in the two variances calculated for the PLS predicted values can be accepted for both types of samples at a level of significance of 0.05. The second nuli hypothesis is that there is no difference in the variances of the predicted values obtained from PLS for the samples analyzed at a conditional time of 0 h and after 96 h. The values of the estimated F ratios were 1.154 (p-value of 0.9284) for the samples with the dye and 0.899 (p-value of 1.053) for the samples mixed with the marker. Again the nuli hypothesis can be accepted at a significant level or 0.05.

Assuming a significant level of 0.05 it can be conduded that the proposed analytical approach is stable over time (there are no significant differences in models* performance) and constmcted calibration models allow for the determination of analytes in samples after 48 and 96 h with comparable and acceptable error levels.

4. Condusions

In this paper. a novel approach Gombining an excitation-emission matrix fluorescence spectroscopy as an analytical technique and partial least squares regression as a mulriple modeling tooł was developed to quantitattveły and qualitarively determine Solvent Red 19 and Solvent Yellow 124 in diesel oiL It is a noodestmctive procedurę which does not retjuire expensive reagents and laborious sam ple preparation prior to analysis. The results obtained from the validation procedurę give evidence that the approach is stable over time. These attractive features make the proposed methodology a potential screening technique that can be used directly at the place wheie samples are collected and thus to support the actions of the cusroms Office and relared agend es. Two calibration methods were evaluated and compared. namely PLS and N-PLS. Calibration models constmcted to predict concentration of SY124 in samples have comparable fit and predktion properties. Compared to PLS, the N-PLS model describing content of SR 19 was characterized with higher RMSE and RM SEP values. In our application. the PLS model is preferred duc to its conceptual simpfidty. In terms of va!idation parameters (RSD and LOD), the HPLC method for SY124 determina-tion performs better than the proposed approach. On the other hand. simplicity and Iow cost of fluorescence spectroscopy encourage its use.

Acknowtedgment

M.D. wishes to express his gratitode to the Minister of Sdence and Higher Education of the Polish Republic for funding the scholarship.

References

111 European Council Dirrciive 95/60/EC.

(2| European Commission Decbion 2003/900/EC.

(31 T. Lisinger. C. Koomen. H. Emteborg. C. Roebben. C. Kramer. A. Lamberty.

Energy Fueli 18 (2004) I8SI-18S4.

|4| DIN 51430.

|S) H.D. SrotL B.W WrighL Talanu 86 (2011) 148- 156.

|6) G.-a Chen. C.-Q. Ma. Y.-M. Wu. H.-J Lu. S.-M. Cao. T. Zhu. Determination and identJHcation of Sudan IV uting fluorortncf spectrometry and artlficial neural networks. in: The 4th International Conference on Information and Computing. Phuket Island: 2S-27 April 2011.

17) R. Bro. Anal. Chim. Acta S00 (2003) 185-194.

181 A.C. Olivieri. AnaL Chem. 80 (2008) 5713-5720.

|9| M. Meira. C.M. Quintella. A.D.S. Tanajura. H.R.C. Da Silva, J.D5. Fernando. P.R. Da Costa Neto. I.M. Pepe. MA Santos, LL Nascimento. Talanta 85 (2011) 430-434.

110| A.C. Olivieri. C.M. Escandar. A. Munoz de la Pena. Trendj AnaL Chem. 30(2011)607-617.

1111 M.J Culroni. H.C. Coicoechen. A.P. Pagani. MA Cabezon. A.C. Ollvteri. Arulyst 131 (2006) 718-723.

(12| O. Divya. A K. Muhra. AppL Spectrojc. 62 (2008) 753-758. jl3| D.L Massart. B.C.M. Vandeginste. LM.C. Buydeiu. S. de Jong. P.J. Lewi. J. SmeyervVert>eke. Handbook of Chemometrics and Qualimetrtcs Part A. Elsevter. London. 1992.

1141 J. Orzeł. M. Da&zykowski. B Walczak. Chemom. Inrell. Lab. Sysi. 110 (2012) 89-96.

115| LC. Thygesen. A Rinnan. S. Barsberg. JXi. Molier. Chemom. Intell. Lab. SysL 71 (2004) 97-106.

116| C. Tomasi. R. Bro. Chemom. Intell. Lab. Sy«. 75 (2005) 163-180.

1171 R.G. Zepp. W.M. Sheldon. MA Moran. Mar. Chem. 89 (2004) 15-36. j 1 B| M. Bahram. R. Bro. C. Stedmoo. A. Afkhami. J. Chemometr. 20 (2006) 99-105. 119] R. Bro. J. Chemometr. 10 (1996)47-62.

(20| R. Bro. Multiway Analysis in the Food Industry. Models. Ałgorithms. and Applications. Royal Veterlnary and Agrlcultural University Denmark. 1998 Available from:.

|211 T. Naes. T. Isaksson. T. Feam. T. Davles. A User-friendly Guide to Multivariate Calibration and Classifiration. NIR Pubłirations. Chichester. 2002.

|22| Q.-S. Xu. Y.-Z. Uang. Chemom. Intell Ub. Syst. 56 (2001) I-II.

(23j M. Daszykowski. S. Semeels. K. Kaczmarek. P. Van Espen. C. Croux, B. Walczak. Chemom. IntelL Ub. Syst. 85 (2007) 269-277.

|24| < http://www.models.kvl.dk/nwaytoolbox>.

|25J C.B. Barber. D.P. Dobkin. H.T. Huhdanpaa. ACM Trans. Math. Softw. 22 {1996) 469-483.

|26| A Lorber. Anal. Chem. 58 (1986) 1167-1172.

|27| AC. Olivierl N.M. Faber. J. Ferre. R. Boque. J.H. Kalivas. H. Mark. Pure Appl. Chem. 78(2006)633-661.

Strona 69