Correlação entre reologia e textura de emulsões para predição de desempenho sensorial pelo uso de redes neurais
Data
2021-07-23
Autores
Franzol, Angelica 
Orientadores
Rezende, Mirabel
Tipo
Tese de doutorado
Título da Revista
ISSN da Revista
Título de Volume
Resumo
Emulsões são sistemas complexos e termodinamicamente instáveis e estão presentes em larga escala em produtos farmacêuticos, cosméticos e de higiene pessoal. Nessa área, o desafio para estabelecer critérios técnicos para a escolha dos componentes das emulsões, de modo que as mesmas atinjam estabilidade cinética, é constante. Tem-se ainda que as características reológicas e de textura das emulsões estão intimamente relacionadas com o sucesso de sua aceitação pelos consumidores. Nesse contexto, o presente trabalho tem como objetivo estabelecer uma metodologia de predição de desempenho sensorial de emulsões cosméticas, pelo uso de um modelo estatístico por rede neural artificial (RNA), alimentado com dados reológicos e de textura, em apoio à tomada de decisão na formulação de emulsões. Para isto, 39 amostras de emulsões, disponíveis comercialmente na Ásia, América do norte, América latina e Europa, foram caracterizadas. A caracterização de todas as amostras envolveu ensaios reológicos nos regimes oscilatório e estacionário (módulos de armazenamento e perda, tixotropia e ponto de fluidez), análises de textura (espalhamento, consistência e pegajosidade) e a determinação do tamanho de gotículas. Ensaios sensoriais com consumidores também foram realizados. Os resultados obtidos mostraram a complexidade presente em cada amostra analisada e que não existe uma tendência de comportamento entre as amostras de uma mesma região ou entre as diferentes regiões. As simulações por RNA utilizando os dados reológicos permitiram predizer o comportamento do consumidor com relação a gosto geral, consistência, espalhabilidade e sensação pesada ou leve sobre a pele, com precisão superior a 60%. As simulações com os dados de textura permitiram predizer, além dos atributos citados, a sensação oleosa ou seca sobre a pele, com precisão entre 37 e 91%. O uso do tamanho de gotículas não se mostrou efetivo nas simulações realizadas e foi desconsiderado nos estudos de predição. Os resultados mostraram que o uso de dados reológicos e de textura na predição de comportamento sensorial de emulsões, pelo uso de um modelo por RNA, foi confiável e gerou predições robustas, em apoio ao desenvolvimento de novos desenvolvimentos nessa área.
Emulsions are complex and thermodynamically unstable systems and are widely present in pharmaceuticals, cosmetics, and personal hygiene products. In this area, the challenge to establish technical criteria for choosing the components of the emulsions, so that they achieve kinetic stability, is constant. There is also that the rheological and texture characteristics of emulsions are closely related to the success of their acceptance by consumers. In this context, the present work aims to establish a methodology for predicting the sensory performance of cosmetic emulsions by using a statistical model by artificial neural network (ANN), fed with rheological and texture data, in support of decision making in the formulation of emulsions. For this, 39 samples of emulsions, commercially available in Asia, North America, Latin America, and Europe, were characterized. The characterization of all samples involved rheological tests in oscillatory and stationary regimes (storage and loss moduli, thixotropy, and yield point), texture analysis (spreading, consistency, and stickiness) and the determination of droplet size. Sensory tests with consumers were also performed. The results obtained showed the complexity present in each sample analyzed and the there is no trend in behavior between samples from the same region or between different regions. ANN simulations using rheological data allowed the prediction of consumer behavior in relation to general taste, consistency, spreadability, and the sensation of heavy/light on the skin, with accuracy greater than 60%. The simulations with the texture data allowed predicting, in addition to the attributes mentioned, the sensation oily/dry on the skin, with precision between 37-91%. The use of droplet size was not effective in the simulations performed and was excluded in the prediction studies. The results showed that the use of rheological and texture data to predict the sensory behavior of emulsions, using an ANN model, was reliable and generated robust predictions, in support of new developments in this area.
Emulsions are complex and thermodynamically unstable systems and are widely present in pharmaceuticals, cosmetics, and personal hygiene products. In this area, the challenge to establish technical criteria for choosing the components of the emulsions, so that they achieve kinetic stability, is constant. There is also that the rheological and texture characteristics of emulsions are closely related to the success of their acceptance by consumers. In this context, the present work aims to establish a methodology for predicting the sensory performance of cosmetic emulsions by using a statistical model by artificial neural network (ANN), fed with rheological and texture data, in support of decision making in the formulation of emulsions. For this, 39 samples of emulsions, commercially available in Asia, North America, Latin America, and Europe, were characterized. The characterization of all samples involved rheological tests in oscillatory and stationary regimes (storage and loss moduli, thixotropy, and yield point), texture analysis (spreading, consistency, and stickiness) and the determination of droplet size. Sensory tests with consumers were also performed. The results obtained showed the complexity present in each sample analyzed and the there is no trend in behavior between samples from the same region or between different regions. ANN simulations using rheological data allowed the prediction of consumer behavior in relation to general taste, consistency, spreadability, and the sensation of heavy/light on the skin, with accuracy greater than 60%. The simulations with the texture data allowed predicting, in addition to the attributes mentioned, the sensation oily/dry on the skin, with precision between 37-91%. The use of droplet size was not effective in the simulations performed and was excluded in the prediction studies. The results showed that the use of rheological and texture data to predict the sensory behavior of emulsions, using an ANN model, was reliable and generated robust predictions, in support of new developments in this area.
Descrição
Citação
Tese de defesa de Doutorado