QUANTITATIVE ANALYSIS OF SULPHATE IN CANE JUICE USING NEAR-INFRARED SPECTROSCOPY

By: Aamarpali Puri

ABSTRACT:

Presence of sulphate in cane juice has detrimental effect it leads to formation of hard scales on the surface of evaporators. Near Infrared Spectrophotometer of Elico (India) in the spectral range of 600-2500nm has been used for the quantitative estimation sulphate in cane juice. A calibration model was set up for this in transmission mode using Partial Least Square Regression Analysis with thirty samples of cane juice containing varying concentrations of sulphate. Statistical parameters like standard error of calibration and correlation coefficient have been evaluated. The model was used to predict the concentration of sulphate in twenty unknown samples of cane juice not present in the calibration file. The standard error of prediction was found to be 0.522 for sulphate. Multi Linear Regression Analysis was also done and the wavelengths have been identified at which the absorbance correlated well with the concentrations of sulphate.

Keywords: sulphate, calibration, transmittance, partial least square regression, stepwise multi linear regression

INTRODUCTION:

The NIR region is the one portion of IR region towards the visible wavelength region and ranges from 0.8 mm (wave number: 12500 cm-1) and goes up to 2.5 mm (wave number: 400 cm-1).

NIR spectroscopy is effective for determination of moisture, fat and protein content in the fish and other meats (Solid and Solberg, 1992; Osborne et al., 1993; Shimamoto et al., 2003) NIR analysis is used for the determination of cotton in polyester yarns (Blanco et al., 1994), and seed oil content and fatty acid composition in sunflower through the analysis of intact seeds, husked seeds, meal and oil (Vich et al, 1998).

In sugar industry the testing of sugarcane for pol, brix, sucrose content, invert and other common constituents have traditionally been done by a series of ICUMSA (International Commission for Uniform Methods of Sugar Analysis) and AOAC (Association of Official Agricultural Chemists) test methods. As many of these methods are time consuming, operator dependent and involve the use of hazardous reagents so Near-Infrared analysis has gained rapid acceptance as an alternative method. NIR spectroscopy can be used for determination of moisture, alcohol, oil, protein, fat, starch, amino acids, hydroxyl ion, film thickness, latex, total carbohydrates, nicotine, attributes like stability and internal damage etc. It can be successfully applied in the areas of baked foods, beverages, fruits, grains, dairy products, meats, flows, pharmaceuticals, paper, textiles, plastics, sugar, vegetable and petrochemicals etc. The various applications (Edye and Clarke, 1996) of NIR in sugar industry are analysis of raw sugar, refinery liquors, run-off syrups, remelts streams, molasses and low purity streams. Near Infrared analysis of shredded (Schaffler and Meyer, 1996) cane is being used as potential replacement for direct analysis of cane. NIR spectroscopy is used for determination of chemical composition (Brix, sugar content, purity) of molasses (Salgo, Nagy and Miko, 1998). NIR spectrophotometeric analysis is an alternative polarization method for raw sugar that uses NIR wavelengths (Player et al., 2000). As near infrared wavelengths specific to individual component can be identified, so it can be used for quantitative analysis of constituents of sugarcane. It is beneficial to opt this method because with it the results are produced in a matter of seconds, with little or no sample preparation required. It calibrates against approved primary methods. Moreover, the light in the near Infrared region is able to penetrate sugarcane juice, molasses and dark color solutions of massecuites etc. Thus, it can be used to predict the values of impurity content in cane juice, purity, pol and brix directly of a given sample without clarification using lead sub acetate. It is online (Fiedler et al, 2001) situation and permits rapid multicomponent analysis by spectrally scanning the sample. It is accurate with minimum of chemical expertise (Davies and Grant, 1986). It is simple (Bruijn, 1997) fast and versatile, with only dilution of liquid samples required, their easy disposal and uses filters.

The average sulphate (Mathur, 1986) content is 300 to 500 ppm of juice. Sometimes, it is much higher and is as high as 2000 ppm SO3 per litre of juice, depending upon the cultural practices and soil conditions. The sulphate content in mixed juice has a great influence on the operational results. If the sulphate content is over 800 ppm of juice, the scale formation constituting sulphate scale in the last vessel of the multiple effect evaporator is well marked. It is a very hard scale, insoluble in acids and difficult to remove even by mechanical scrappers. A good amount of sulphate in juice passed on from the evaporator gets precipitated in the crystallization process and fouls the heating tubes of the pans.

Treating the juice with superphosphate can reduce the intensity of the sulphate scaling. Online estimation of phosphate (Kaur and Aamarpali, 2004) and silica (Puri and Kaur, 2006) has already been done using NIR spectroscopy. So in continuation with my studies, NIR spectroscopic method for analysis of sulphate in cane juice is given in the present investigation, which is essential in factories before starting the process of clarification of cane juice.

 INSTRUMENTATION

  • Near Infrared Spectrophotometer of Elico (India): Spectral range 600-2500nm, bandwidth 10nm, accuracy +5nm, repeatability + 0.2nm and with advanced state of the art MS Windows® based software for data acquisition was used. With this instrument processing, storage, retrieval and interpretation of complex spectra can be done. This spectrophotometer helps in quantitative estimation of impurities in cane juice using regression analysis. The instrumental set up is shown in the Figure.1. It is PC based user friendly and menu driven. It is having high performance concave grating monochromator and two colour detector.

MATERIALS AND METHODS:

The initial concentration of sulphate was found to be 410 ppm in cane juice. As the accuracy of NIR analysis is wholly dependent on the quality of calibration set so utmost care was taken in gathering, selecting and preparing samples to be used for calibration. While collecting samples it was taken into consideration that the samples should cover wide range of constituent’s concentration. The cane juice samples undergo chemical and biological (microbial) degradation with time. So the samples were analyzed on the same day without any delay. The instrument was away from direct sunlight and electromagnetic radiations. There was no radio frequency interference. The lab was well ventilated with ambient temperature maintained between 288-308 K and relative humidity 45 – 80 %. The samples were scanned in transmittance mode. While scanning there was no dust and corrosive gases. Scanning was done in the surroundings, which was free of vibrations and shocks. In this particular application NIR light is transmitted through the cane juice samples. The samples used for calibration were containing added sulphur in such a manner so as to cover it’s complete range in different local cane-juice varieties. In this spectroscopic analysis the samples are illuminated with light, which gives different wavelengths. The sample reacts with light producing a unique spectrum. This resultant transmittance spectrum gives the measure of the composition of sugar cane juice. In calibration process the sample specific transmittance data is compared with known wet chemical analysis of a selected set of sample so that if the sample transmission data matches with calibration sample set, it gives accurate results.

Calibration model was prepared for sulphate in cane juice using PLSR. For setting calibration model, thirty samples of this sulphate were scanned in transmittance mode. Calibration file was having a concentration range sulphate 300-500 ppm. While preparing calibration file for the sulphate the range was adjusted by dilution with water and by more addition of the sulphate. The dilution effect was nullified by having constant volume of solution in each sample.

Prediction file was prepared for sulphate in cane juice. For the preparation of prediction file twenty samples of cane juice having varying concentrations of sulphate were scanned and their prediction was done using respective calibration models. The prediction value is obtained from software, which gives required value by comparing with the calibration file chosen. Very carefully calibration file was chosen which was generated from same mode through the same range.

RESULTS AND DISCUSSION:

Two multivariate calibration procedures ie Partial Least Square Regression analysis (PLSR) and Stepwise Multivariate Linear Regression analysis (SMLR) were applied. Calibration model was set up for sulphate. Calibration models were set up using Partial Least Square Regression analysis. Standard error of calibration (SEC) and correlation coefficient were determined. The Standard Error of Prediction (SEP) and it’s correlation coefficient were also determined. SEC/SEP were determined using (Eq. 1)

SEC/SEP   =             Σ (Y-X)2   (N-1)                                    (.1)

Where Y is the result predicted from chemical analysis or the actual value, X is predicted from NIR measurements and N is the number of samples in the calibration/ prediction set. The actual values and NIR predicted values for sulphate in ppm are given in the table .1.

The summary statistics for the calibration of sulphate in cane juice is reported in the table .2. The table shows the standard error of calibration along with a multiple regression of the analyte. The SEC is found to be negligible for sulphate.

The summary statistics for prediction of sulphate in cane juice is reported in the table .3. The correlation is found to be almost 1 and SEP is also very less. Good agreement was found between the actual and the NIR predicted values. The error may vary to some extent depending on geographical location of samples used for prediction, difference in lab personnel or error from lab methods. Standard Error of Calibration (SEC) is found to be much less then the Standard Error of Prediction (SEP).

Stepwise Multivariate Linear Regression analysis has also been carried out to identify the wavelengths, which respond well to impurity under consideration. The order of correlation was determined. From the spectrum of each impurity, eight wavelengths were selected, out of these best-correlated wavelengths were selected. For selection of best-correlated wavelengths the regression was applied to the wavelength that correlated maximum and the standard deviation was determined. To this next wavelength was added in turn and retained only if it helped lower the standard deviation. The absorbances at specific wavelengths are correlated with concentrations of absorbing species. The quantity of the analyte (C) was determined. This can be expressed (Eq. 2) as follows.

         C    =    k0    +   k1 [A]x    +     ———–kn [A]x                             (2)

Where k0, k1——n are constants and [A]x are the absorbance values at different wavelengths. The data at these wavelengths has been reported in table .4. It also shows standard deviations of the analyte. The standard deviation for sulphate is found to be . The table .5 shows wavelengths identified for sulphate in cane juice. The plot of sulphate content (ppm) in cane juice vs. absorbance values of the selected wavelengths can help knowing concentration of sulphate in given cane juice sample very easily. In this study the results are confined to zero order derivative because at higher order derivative the value of SEP was found to be very high. So best prediction was obtained at zero order derivatives.

CONCLUSIONS:

In the calibration models set up by the Partial Least Square Regression analysis, SEC was found to be negligible which confirms the reliability of calibration model set up for sulphate in cane juice. The standard error of prediction was found to be very less with best correlation so NIR spectroscopic method can be successfully used for accurate prediction of the concentration of unknown samples of cane juice in few seconds. The best-correlated wavelengths were identified with Stepwise Multivariate Linear Regression analysis for sulphate. These wavelengths can give useful information about particular impurity in cane juice. From the results it was confirmed that the Partial Least Square analysis give much better results than Stepwise Multi Linear Regression analysis.

ACKNOWLEDGEMENT:

Corresponding Author Dr. Aamarpali Ratna Puri is very thankful to her mother (Ms.Sneh Lata) for her invaluable support and Guru Nanak Dev University, Amritsar. for providing necessary instruments and infrastructure for the present research.

 REFERENCES:

Blanco, M. Coello, J. Iturriaga, H. Maspoch, S. and Bertan, E., 1994. Analysis of cotton Polyester yarn by Near Infrared Spectroscopy. Analyst. 119: 1779-1782.

Bruijn, J, M., 1997. Development and application of automatic NIRS in factory laboratories. Technical Session, Brit Sug. 3B (iii), 11.

Davies, A. M. C., Grant, A. 1986. Review: Near Infrared analysis of food. Int. J. Food. Sci. Tech., 22: 191-207.

Edye, L. A., Clarke, M. A., 1996. Near Infrared Spectroscopy in sugar refining: Five years down the road. Proc. Annual. Meeting. Sug. Ind. Tech. 55, 1-8.

Fiedler, F. M., Edye, L. A., Watson, L. J., 2001. The application of discriminant analysis to online near infrared spectroscopy of prepared sugar cane. Proc. Aust. Sug. Cane Tech. 23, 317-321.

Kaur, S., Aamarpali., 2004. Quantitative estimation of phosphate in cane juice using near infrared Spectrophotometer. Indian Sugar, Vol (LIV), 513-517. Oct.

Mathur, R.L. (1986). “Handbook of cane sugar technology”. IInd Edition, Oxford and IBH Publishing Co., India, 96.

Osborne, B. G., Fearn, T., Hindle, P. H., 1993. Practical spectroscopy with application in food and beverage analysis. UK, Longman Scientific & Technical.

Puri, A. R., Kaur, S., 2006. Estimation of silica in cane juice using near infrared spectroscopy, Cooperative sugar, Vol (37): No 6. 29-33. Feb.

Player, M. R., Rowe, G. S., Urquhart, R. M., McCunnie, K. A., McCarthy, D., 2000. Polarization of raw sugar without basic lead acetate: Int. Collaborative test. Proc. 22nd Conf. Aust. Soc. Sug. Tech., 385-392.

Salgo, A., Nagy, J., Miko, E., 1998. Application of near infrared spectroscopy in the sugar industry. J. NIR. Spectrosc. 6: 101-106.

Schaffler, K. J and Meyer, J. H., 1996. Near Infrared analysis of shredded cane: A potential replacement for direct analysis of cane, Proceeding of South African. Sugar. Technologists .Association. 70, 5, 131-139.

Shimamoto, J., Hiratsuka, S., Hasegawa, K., Sato, M., Kawano, S., 2003. Rapid non-destructive determination of fat content in frozen skipjack using a portable near infrared spectrophotometer, Fisheries Sci. 69. 856-60.

Solid, H, and Solberg, C., 1992. Salmon fat content estimation by Near-Infrared transmission spectroscopy. J Food Sci. 57: 792-93.

Vich, B.P., Velaso, L, Martinez, J. M. F., 1998. Determination of seed oil content and fatty acid composition in sunflower through the analysis of intact seed, husked seed meal and oil by Near Infrared Reflectance Spectroscopy. J. Am. Oil Chem. Soc. 75 (5) 547-555.

TABLES:

Table .1 Showing actual values and values predicted from the NIR spectroscopy of sulphate (ppm) in cane juice. 

S.No Actual values

(ppm)

Predicted values

(ppm)

.1 296.0 296.5
.2 350.0 350.3
.3 448.5 448.0
.4 498.0 499.0
.5 425.0 425.5
.6 380.5 381.0
.7 320.0 320.5
.8 360.5 360.0
.9 308.5 308.0
.10 395.5 395.0
.11 444.0 444.5
.12 458.0 458.5
.13 470.5 471.0
.14 452.0 452.5
.15 311.0 311.5
.16 334.0 334.5
.17 447.5 447.0
.18 386.0 385.5
.19 488.5 489.0
.20 408.0 407.5

 

Table .2 Summary statistics for the PLS calibration of sulphate (individually) from 30 samples of cane juice.

S.No Analyte

in cane juice

Range of values

(ppm)

    Calibration

SEC           MR

1. Sulphate  300-500 0.128        0.999

Where: SEC = Standard Error of Calibration, MR =Multiple Regression coefficient

Table .3 Summary statistics for the prediction of sulphate (individually) from 20 samples of cane juice.

S.No Analyte

in cane juice

SEP   R
1. Sulphate 0.522 0.999

Where: SEP = Standard Error of Prediction, R = correlation coefficient

Table .4 Regression coefficients and Standard Deviation (s) data for sulphate using Stepwise Multivariate Linear Regression (SMLR) analysis.

S.No Analyte k0 k1 k2 k3 s
1. sulphate 583.627 -400.767 -78.358 -18.124 6.47

Table .5 Wavelengths identified from Stepwise Multivariate Linear Regression analysis that respond well to impurity under consideration.

Analyte   Wavelengths (nm)    
sulphate   2025 900 1450

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s