Determinação de parâmetros físico-químicos em laranjas Valência por métodos não-destrutivos
Carregando...
Arquivos
Data
2016-07-20
Autores
Borba, Karla Rodrigues [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Objetivo: Alguns atributos em sua composição química e física, como teor de sólidos solúveis (SST), pH (potencial hidrogeniônico), ácidos orgânicos, açúcares e cor são utilizados como parâmetros na aferição da qualidade de frutas. Os métodos utilizados, nos dias atuais, para a análise da qualidade destes produtos são invasivos, que demandam tempo, muitas amostras, reagentes e geram resíduos. Técnicas não destrutivas, como a espectroscopia no infravermelho e de ressonância magnética nuclear, têm sido propostas para avaliar qualidade em frutas e hortaliças. Este projeto teve como objetivo: desenvolver um modelo de calibração, a partir da espectroscopia de infravermelho de médio e próximo, e ressonância magnética com o auxílio das ferramentas quimiométricas PCA (análises de componentes principais) e PLS (Regressão em mínimos quadrados parciais) com a utilização do programa computacional Pirouette versão 4.5, para estimar parâmetros físico-químicos da qualidade de laranjas da variedade ‘Valência’. Métodos: Foram coletadas um total de 450 laranjas de cultivo comercial no interior de São Paulo, transportadas para a Embrapa Instrumentação, São Carlos, SP e em laboratório foram selecionadas, higienizadas e sanitizadas. Em seguida foram realizadas análises espectroscópicas de infravermelho médio (MIR) e próximo (NIR) e de ressonância magnética nuclear (RMN). Logo após realizadas as análises não destrutivas as laranjas foram analisadas com métodos de referência para os parâmetros de qualidade. Os sinais espectroscópicos foram processados e os modelos de predição foram construídos. Com os sinais de NIR e MIR foram desenvolvidos modelos PLS para determinação de: SST, pH, massa fresca, vitamina C, ácido cítrico e açucares. Para os sinais de decaimento de RMN foram construídos modelos para SST, pH e massa fresca. Resultados: Para a predição de SST o modelo PLS – NIR mostrou melhor resultado, com SEP 0,74 °BRIX e coeficiente de correlação de 0,75. O parâmetro que apresentou melhor aplicação com a técnica de RMN-dt foi para determinação de massa fresca, com SEP 21,41 gramas e r = 0,82. A determinação do pH foi mais precisa com o modelo PLS-MIR, onde observou-se SEP de 0,20 e r = 0,86. Para a predição de vitamina C e ácido cítrico tanto o NIR quanto o MIR resultaram em modelos com aplicabilidade. O modelo mais preciso foi o desenvolvido com sinas de MIR, com coeficiente de correlação de Pearson (r) de 0,81 e erro quadrático médio de predição (RMESP) de 7,51 para vitamina C e 0,82 e 3,36 para ácido cítrico. Em relação a predição de açúcares o NIR não resultou em modelos precisos. Todavia, a utilização de MIR para a predição de sacarose, glicose e açúcares totais apresentaram baixos erros de predição 11,69, 7,79 e 24,28 (g L-1) e coeficientes de correlação de forte a moderado. Conclusão: A aplicação de espectroscopia na determinação de parâmetros de qualidade mostrou-se uma possível alternativa na análise de qualidade de laranjas sem que seja necessária a destruição das amostras.
Objective: Some attributes in their chemical and physical composition, as soluble solids content (SSC), pH (hydrogen potential), organic acids, sugars and color are used as parameters in measuring the quality of fruit. The methods used nowadays, to analyze the quality of these products are invasive, requiring time, many samples, reagents and generate waste. Nondestructive techniques such as infrared spectroscopy and nuclear magnetic resonance, have been proposed to evaluate quality in fruits and vegetables because they allow rapid implementation of measurements without sample preparation. This project aimed to: develop a calibration model by near ande mid-infrared spectroscopy and nuclear magnetic resonance with the aid of chemometric tools PCA (principal component analysis) and PLS (regression in partial least squares) with use of the computer program Pirouette version 4.5, to estimate physical-chemical parameters of quality oranges from ‘Valencia’ variety. Methodology: A total of 450 commercial cultivation oranges were collected in São Paulo, transported to Embrapa Instrumentation, São Carlos/SP, and in laboratory were selected, cleaned and sanitized. Then spectroscopic analysis near (NIR) and mid-infrared (MIR), and nuclear magnetic resonance (NMR) were made. After performed the non-destructive analysis, the oranges were analyzed with reference methods for quality parameters. Spectroscopic data were processed and prediction models were developed to determinith MIR and NIR signal PLS models were developed to determine SSC, pH, fresh weight, vitamin C, citric acid and sugars. For the NMR signal decay models were constructed to SSC, pH and fresh weight. Results: For SSC prediction PLS model - NIR showed better results, with RMSEP (mean squared prediction error) 0.74 °BRIX and 0.75 correlation coefficient. The parameter that showed better application with DT- NMR technique was to determine fresh weight with RMSEP 21.41 grams and r = 0.82. The determination of pH was more accurate with the PLS-MIR model where there was RMSEP 0.20 and r = 0.86. For the prediction of vitamin C and citric acid as both the MIR NIR models resulted in applicability. The most accurate model was developed with MIR, with Pearson's correlation coefficient (r) of 0.81 and RMESP of 7.51 for vitamin C and 0.82 and 3.36 for Citric acid. Regarding the prediction of sugars NIR did not result in accurate models. However, the use of MIR for the prediction of sucrose, glucose and total sugars were low prediction errors 11.69, 7.79 and 24.28 (g L-1 ) and strong correlation coefficients to moderate.Conclusion: The application of spectroscopy to determining quality parameters proved to be a viable alternative in quality analysis oranges without destruction of samples.
Objective: Some attributes in their chemical and physical composition, as soluble solids content (SSC), pH (hydrogen potential), organic acids, sugars and color are used as parameters in measuring the quality of fruit. The methods used nowadays, to analyze the quality of these products are invasive, requiring time, many samples, reagents and generate waste. Nondestructive techniques such as infrared spectroscopy and nuclear magnetic resonance, have been proposed to evaluate quality in fruits and vegetables because they allow rapid implementation of measurements without sample preparation. This project aimed to: develop a calibration model by near ande mid-infrared spectroscopy and nuclear magnetic resonance with the aid of chemometric tools PCA (principal component analysis) and PLS (regression in partial least squares) with use of the computer program Pirouette version 4.5, to estimate physical-chemical parameters of quality oranges from ‘Valencia’ variety. Methodology: A total of 450 commercial cultivation oranges were collected in São Paulo, transported to Embrapa Instrumentation, São Carlos/SP, and in laboratory were selected, cleaned and sanitized. Then spectroscopic analysis near (NIR) and mid-infrared (MIR), and nuclear magnetic resonance (NMR) were made. After performed the non-destructive analysis, the oranges were analyzed with reference methods for quality parameters. Spectroscopic data were processed and prediction models were developed to determinith MIR and NIR signal PLS models were developed to determine SSC, pH, fresh weight, vitamin C, citric acid and sugars. For the NMR signal decay models were constructed to SSC, pH and fresh weight. Results: For SSC prediction PLS model - NIR showed better results, with RMSEP (mean squared prediction error) 0.74 °BRIX and 0.75 correlation coefficient. The parameter that showed better application with DT- NMR technique was to determine fresh weight with RMSEP 21.41 grams and r = 0.82. The determination of pH was more accurate with the PLS-MIR model where there was RMSEP 0.20 and r = 0.86. For the prediction of vitamin C and citric acid as both the MIR NIR models resulted in applicability. The most accurate model was developed with MIR, with Pearson's correlation coefficient (r) of 0.81 and RMESP of 7.51 for vitamin C and 0.82 and 3.36 for Citric acid. Regarding the prediction of sugars NIR did not result in accurate models. However, the use of MIR for the prediction of sucrose, glucose and total sugars were low prediction errors 11.69, 7.79 and 24.28 (g L-1 ) and strong correlation coefficients to moderate.Conclusion: The application of spectroscopy to determining quality parameters proved to be a viable alternative in quality analysis oranges without destruction of samples.
Descrição
Palavras-chave
Espectroscopia, Infravermelho, Ressonância magnética nuclear, Qualidade de laranja intacta, Quimiometria, PLS, Spectroscopy, Infrared, Nuclear magnetic resonance, Orange quality intact, Chemometrics