Aplicação de métodos analíticos não destrutivos na melhoria do processo de controle de qualidade de noz macadâmia
Carregando...
Data
2019-02-13
Autores
Carvalho, Livia Cirino
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
Objetivo: Desenvolver métodos analíticos com uso da espectroscopia do infravermelho próximo (NIRS) e ressonância magnética nuclear (NMR) para a melhoria do processo de controle de qualidade de nozes macadâmia (Macadamia integrifolia Maiden & Betche). Métodos: Foram coletadas nozes de macadâmia intactas de várias cultivares (HAES 246, ‘IAC 4-20’, ‘IAC 2-23’, ‘IAC 5-10’,‘IAC 8-17’,HAES 344, HAES 660, IAC 4-12 B e IAC Campinas B) e amêndoas sem defeitos e que apresentavam diferentes defeitos, ou seja, amêndoas imaturas; amêndoas danos causados por picada de insetos, amêndoas mofadas e amêndoas descoloridas. Além disso, foi analisado o efeito de diferentes temperaturas de secagem quanto os paramentos de oxidação, índice de peróxido (IP) e de acidez (IA). Os espectros NIR foram coletados na superfície das nozes intactas e também das amêndoas (NIRS e NMR-TD) e estes foram utilizados para a construção dos modelos de classificação e predição utilizando análise discriminante linear por componente principal (PCA-LDA), algoritmo genético com análise discriminante linear (GA-LDA) e regressão por mínimos quadrados parciais (PLSR). Resultados: Em relação a classificação das cultivares, os modelos desenvolvidos com GA-LDA resultaram em acurácia de 71,57% com os espectros pré-processados com suavização de Savitzky-Golay. Não foi observado efeito do processo de secagem no IP e IA. Os melhores resultados de predição do IP (R²c=0,57 e SEP=0,55 meq·kg−1) e IA (SEP=0.14%, R2C=0.29) foram obtidos empregando PLSR. Em relação aos defeitos, foi possível obter uma acurácia e especificidade de 97,8% e 100% utilizando GA-LDA, respectivamente para os espectros NIR. O melhor modelo RMN de classificação foi obtido utilizando PCA-LDA, 84,7% de acurácia para o grupo de treinamento e 83.3% para o grupo de validação. Conclusão: A NIRS e RMN–TD podem ser utilizada no processo de controle de qualidade de nozes e amêndoas de macadâmia com potencial benefício para as indústrias processadoras.
Objective: To develop analytical methods with the use of near infrared spectroscopy (NIRS) and nuclear magnetic resonance (NMR) to improve the quality control process of macadamia nuts (Macadamia integrifolia Maiden & Betche). Methods: Intact macadamia nuts were collected from several cultivars (HAES 246, ‘IAC 4-20’, ‘IAC 2-23’, ‘IAC 5-10’,‘IAC 8-17’,HAES 344, HAES 660, IAC 4-12 B, and IAC Campinas B) and kernels that had different defects, as such: nuts without defects; immature kernels; damages caused by insect, mold kernels, and discolored kernels. In addition, the effect of different types of drying on the oxidation parameters, peroxide value (PV) and acidity index (AI) were analyzed. The NIR spectra were collected on the surface of intact nuts and kernels (NIR and NMR-TD), and the spetra were used to develop the classification and prediction models using linear discriminant analysis by principal component analysis (PCA-LDA), genetic algorithm with linear discriminant analysis (GA-LDA), and partial least squares regression (PLSR). Results: In relation to cultivar classification, the models developed with GA-LDA resulted in an accuracy of 71,57% with Savitzky-Golay smoothing pre-processed spectra. No effect of the drying process was observed on IP and AI. The best PV prediction results (R²c = 0.57 and SEP = 0.55 meq.kg-1) and AI (SEP = 0.14 %, R2C = 0.29) were obtained employing PLSR. Regarding kernel defects, it was possible to obtain an accuracy and specificity of 97.8% and 100% using GA-LDA, respectively. The best NMR classification models using PCA-LDA, achieving 85.7 % accuracy for the training set and 83.3 % for the validation set. Conclusion: NIRS and NMR-TD can be used in the quality control process of macadamia nuts and kernels with potential benefits for the processing industries.
Objective: To develop analytical methods with the use of near infrared spectroscopy (NIRS) and nuclear magnetic resonance (NMR) to improve the quality control process of macadamia nuts (Macadamia integrifolia Maiden & Betche). Methods: Intact macadamia nuts were collected from several cultivars (HAES 246, ‘IAC 4-20’, ‘IAC 2-23’, ‘IAC 5-10’,‘IAC 8-17’,HAES 344, HAES 660, IAC 4-12 B, and IAC Campinas B) and kernels that had different defects, as such: nuts without defects; immature kernels; damages caused by insect, mold kernels, and discolored kernels. In addition, the effect of different types of drying on the oxidation parameters, peroxide value (PV) and acidity index (AI) were analyzed. The NIR spectra were collected on the surface of intact nuts and kernels (NIR and NMR-TD), and the spetra were used to develop the classification and prediction models using linear discriminant analysis by principal component analysis (PCA-LDA), genetic algorithm with linear discriminant analysis (GA-LDA), and partial least squares regression (PLSR). Results: In relation to cultivar classification, the models developed with GA-LDA resulted in an accuracy of 71,57% with Savitzky-Golay smoothing pre-processed spectra. No effect of the drying process was observed on IP and AI. The best PV prediction results (R²c = 0.57 and SEP = 0.55 meq.kg-1) and AI (SEP = 0.14 %, R2C = 0.29) were obtained employing PLSR. Regarding kernel defects, it was possible to obtain an accuracy and specificity of 97.8% and 100% using GA-LDA, respectively. The best NMR classification models using PCA-LDA, achieving 85.7 % accuracy for the training set and 83.3 % for the validation set. Conclusion: NIRS and NMR-TD can be used in the quality control process of macadamia nuts and kernels with potential benefits for the processing industries.
Descrição
Palavras-chave
Macadamia integrifolia, NIRS, PLS-DA, GA-LDA, PLSR, NMR