Inteligência Artificial aplicada em sistemas de embalagens inteligentes com indicadores visuais de variação de pH
Data
2022-11-29
Autores
Brazolin, Isadora Fernandes [UNIFESP]
Orientadores
Yoshida, Cristina Maria Pedroso [UNIFESP]
Tipo
Dissertação de mestrado
Título da Revista
ISSN da Revista
Título de Volume
Resumo
Embalagens inteligentes são sistemas de embalagens que informam sobre a
qualidade de um produto de maneira fácil e acessível, sem a necessidade da abertura
da embalagem. O controle de qualidade e a segurança do alimento são fatores
importantes para o mercado, garantindo a saúde do consumidor. O pH dos alimentos
pode estar relacionado com aspectos de qualidade e segurança do alimento embalado,
indicando deterioração, crescimento microbiano e adulteração. Neste estudo, a criação
e treinamento de uma rede neural artificial (RNA) foram desenvolvidos visando facilitar
o controle de qualidade de produtos alimentícios com alterações/adulterações a partir
de reações que apresentam variação de pH, baseado na resposta de um indicador
colorimétrico natural funcional de um sistema de embalagem inteligente sustentável e
de fácil fabricação (filmes de quitosana). Filmes inteligentes de quitosana-antocianina
(CH-ATH) foram formulados com diferentes concentrações de quitosana (Cch, 0,5, 1,0
e 2,0%, m/m) e de indicador colorimétrico natural, a antocianina (Cath, 0,5, 1,0 e 2,0%,
m/m). Os filmes CH-ATH foram caracterizados quanto a solubilidade em água,
propriedades mecânicas e análises térmicas (DSC e TGA). Para avaliar a indicação
colorimétrica dos filmes CH-ATH, os mesmos foram imersos em uma ampla faixa de pH
(1,20 a 12,58), e foram medidas as variações dos parâmetros de cor dos filmes (L*, a*,
b*). Com os resultados experimentais, foi desenvolvido um modelo multivariável
empírico baseado em inteligência artificial (redes neurais artificiais) para determinação
de pH pela variação da cor do indicador e as concentrações de quitosana e antocianina.
A rede neural facilita a avaliação da segurança e qualidade do produto alimentício
embalado, sem a necessidade de um dispositivo ou análises químicas para medir a
condição do pH. Os filmes com maiores Cath apresentaram menor resistência,
coloração violeta mais forte e espessos e baixa solubilidade. Os filmes com maiores Cch
apresentaram maior resistência, cor violeta mais visível e flexibilidade, porém mais
solúveis em água. Todos os filmes CH-ATH apresentaram estabilidade térmica
compatível com a indústria alimentícia. A RNA apresentou assertividade de 79% na
classificação; expressando a necessidade de um maior número de dados para o
treinamento.
Intelligent packaging systems provide information about the quality of a food product in an easy and accessible way, without the need to open the package. Food quality control and safety are important factors for the market, ensuring consumer health. The pH of foods is strongly related to quality and safety of the packed food products, and may indicate deterioration, microbial growth, and adulteration. In this study, the creation and training of an artificial neural network (ANN) were developed to facilitate the food quality control with alterations/adulterations from reactions that present pH variation, based on the response of a natural functional colorimetric indicator of a sustainable and easy-to-manufacture smart packaging system (chitosan films). Chitosan-anthocyanin (CH-ATH) intelligent films were formulated with different concentrations of chitosan (Cch, 0.5, 1.0, and 2.0%, w/w) and natural colorimetric indicator, anthocyanin (Cath, 0.5, 1.0, and 2.0% w/w). CH-ATH films were characterized by water solubility, mechanical properties, and thermal analysis (DSC and ATG). To efficiency of colorimetric indication of the devices was evaluated by immersing the CH-ATH films in a wide pH range (1.20 to 12.58), and the variations of the color parameters (L*, a*, b*) were measured. Based on the experimental results, an empirical multivariable model based on artificial intelligence (artificial neural networks) was developed to determine pH by varying the color of the CH-ATH film indicator and the concentrations of chitosan and anthocyanin. The neural network facilitates the assessment of the safety and quality of the packaged food product, without the need for a device or chemical analysis to measure the pH condition. Films with higher Cath showed lower strength, are darker violet color and thicker, with low water solubility. CH-ATH films with higher Cch showed good strength, more visible violet color, and good flexibility, but more water soluble. All CH-ATH film indicator showed thermal stability compatible with the food industry. RNA presented a classification accuracy of 79%, therefore expressing the need for a greater amount of data for training.
Intelligent packaging systems provide information about the quality of a food product in an easy and accessible way, without the need to open the package. Food quality control and safety are important factors for the market, ensuring consumer health. The pH of foods is strongly related to quality and safety of the packed food products, and may indicate deterioration, microbial growth, and adulteration. In this study, the creation and training of an artificial neural network (ANN) were developed to facilitate the food quality control with alterations/adulterations from reactions that present pH variation, based on the response of a natural functional colorimetric indicator of a sustainable and easy-to-manufacture smart packaging system (chitosan films). Chitosan-anthocyanin (CH-ATH) intelligent films were formulated with different concentrations of chitosan (Cch, 0.5, 1.0, and 2.0%, w/w) and natural colorimetric indicator, anthocyanin (Cath, 0.5, 1.0, and 2.0% w/w). CH-ATH films were characterized by water solubility, mechanical properties, and thermal analysis (DSC and ATG). To efficiency of colorimetric indication of the devices was evaluated by immersing the CH-ATH films in a wide pH range (1.20 to 12.58), and the variations of the color parameters (L*, a*, b*) were measured. Based on the experimental results, an empirical multivariable model based on artificial intelligence (artificial neural networks) was developed to determine pH by varying the color of the CH-ATH film indicator and the concentrations of chitosan and anthocyanin. The neural network facilitates the assessment of the safety and quality of the packaged food product, without the need for a device or chemical analysis to measure the pH condition. Films with higher Cath showed lower strength, are darker violet color and thicker, with low water solubility. CH-ATH films with higher Cch showed good strength, more visible violet color, and good flexibility, but more water soluble. All CH-ATH film indicator showed thermal stability compatible with the food industry. RNA presented a classification accuracy of 79%, therefore expressing the need for a greater amount of data for training.
Descrição
Citação
BRAZOLIN, Isadora Fernandes. Inteligência Artificial aplicada em sistemas de embalagens inteligentes com indicadores visuais de variação de pH. 2022. 97 f. Dissertação (Mestrado em Engenharia Química) - Instituto de Ciências Ambientais, Químicas e Farmacêuticas, Universidade Federal de São Paulo, Diadema, 2022.