8290614306

7. Opublikowane badania własne

Chłmomnra and Inwlligtm Ubcrjtcry Systems 110 ( 2012) 69 96

CM(«*o«fnacs ^: Iaw ifmucncT PTiiiAWMTwn Braniu

Contents lists available at SciVerse ScienceDirect

Chemometrics and Intelligent Laboratory Systems

n SFYIFR

lournal homepagc: www.ciscvicr.com/locate/cherrolab

Controlling sugar ąuality on the basis of fluorescence fingerprints using robust calibration

J. Orzeł, M. Daszykowski *. B. Walczak

Depanir.ent of Anofyurel Cf-rmarry. Chtmomemc Mruorrh Cntp. tnirinte of Oirmuiry Ttrr Uah/tnUy of SilnM. 9 SzJkotaid Sirwi. 40 006 KarmtKr. Poiant

ABSTRACT

A R T I C L E 1NF0

AmrV (liitanr Rernwd II July 2011

Rrceiwrd In rrwlwd form 28 Septrmber 20| 1 Acrrptfd 3 On o ber 2011 Awailiblr onbre B Ociofarf 2011

Kfywotó*

Proctti jnjlytKal trchnology Par tul leatl tquares Panial robun M-rrgTtuinn N FIS

Tbe Jim of our study was lo highligtit che benefits of robust calibration in the contexl of process control Twe properties were monitored — the color and ash content of sugar samples. It was shown for the data being studied thai robust modHs. constructed using the partia! robust M-regression tcchnique. have a better fil ta the majonty of the data and prediciion properties than the classic partial ieast spuares and N-way partia! least squares models. In pamcular. the constructed calibratioo models were characterized by a root mean square errors improved by 1.60* and 1.82* and a roor mean squarr errcrs of prediction (for independent test samples) impraved by 2.39* and l.llt rnmpared ta classic panial least squares modeis constructed for color and ash conient. respeciively.

© 20U Elsevier B V. Ali rights teserved.

1. Inrrodurfion

Nowadays. the quality of a given food product is a significant cri-terion that is considered by consumers. A high product quality implies its safety and satisfactory sensory properties. The safety of food products is controlled by authorized laboratories on the basis of regulations for describing the attributes of food products. These regulations are established by local and intemational organizations eg. the International Standardization for Organization (ISO), the Quality Contro! Council of the Llnired States, the Pohsh Standardiza-rion Committee, etc. On the other hand. sensory evaluation is often morę definirive than the laboratory one because some product fea-tures such as color, texture and smell ran be judged by consumers themselves. Therefore. it is very important to control the finał product and its production process in order to obtain prcduds that are virtu-ally identical from batch ta batch. This philosophy is the core of process analytical technology (PAT). PAT efficiently combines knowledge of analytical chemisny and the use of chetnometric tools to design and monitor the quality of manufactured products. The PAT methnd-ology is greatly appreciated in many branches of industry, including the food. pharmaceutical and Chemical industries 11-5|. The sugar industry is one of many eicamples where the PAT concepts are exercised extensively |6.7). The main properties that are important for sugar quality assessment are its color and purity (defined as the total amount of organie and inorganic impurities). To monitor changes in these rwo parameters. multivariate calibration models such as

• Ceifcipondinfl julher

E maiJ addnen: mtcluUUsrytawikifui edt.pl (M Ojuylcowskl).

0HW-7O9/5 - front nurter O 2011 Elwier B.V Ali rijthti reterwd doi 10 1016 j chrmoljh 2011 10 001

Principal component regression (PCR) |8J and partial least squares regression (PIS) (8J can be usedL

Recently, rhere has been a steadily growing interest in the use of Chemical fingerprints for describing different processes and their end products for the purpose of process control. Spectroscopic tech-niques, such as near-infrared specrroscopy (NIR) and fluorescence spectroscopy have gained a great deal of attention in the field of PAT. They are relatively inexpensive, allow for the rapid acquisition of spectroscopic fingerprints (also on-line) and are non-destructive techniques.

In this study, the use of fluorescence fingerprints in PAT is ex-plored. Excitarion and emission fluorescence speara of samples (the so-called fluorescence landscapes) that are collected simultaneously contain richer Chemical information than a single excitarion or emission fluorescence spectrum. Therefore. they seem to be attractive for the purposes of process control To datę. fluorescence fingerprints have been used to construct multivanate calibration models that are able to predict the riboflavin content in yogurt |9| and to monitor the deterioration of extra virgin olive oil during heating 110). In the context of sugar quality control. fluorescence fingerprints have been used to construct calibration models to describe the color of raw cane sugar |11) and the physicochemical quality parameters of thick juice and beet sugar (12).

The presence of smali amounrs of impurities. such as phenols, amino acids. products of their reactions and inorganic rcsiduals. in sugar is responsible for its quality. Therefore, their Ievels should be monitored. As illustrated in 113]. owing to its high sensitivity, fluorescence spectroscopy can be used for this purpose.

In this study. the possibilicies offered by the robust calibration technique. called panial robust M-regression (PRM). which enables

Strona 52