Modelling thermal conductivity of foods

Abstract

In this study, mixtures were made in the laboratory having different proportions of food components and are referred to as Model Foods. They were employed in this study to minimize variation in samples due to factors other than composition and density. Food items which were purchased from a local retailer are referred to as 'real' foods. Data on Model Foods was used to determine effect of composition and density on thermal conductivity of foods and to develop thermal conductivity prediction models. Three sets of 'real' food data were used for testing the performance and validity of the proposed models. Variation in the composition and densities were examined over the following ranges: water content, 0% to 80%; protein content, 0% to 30%; carbohydrate content, 0% to 60%; fat content, 0% to 100%; and density, 0.849 g/cm^2 to 1.189 g/cm^2. Effects of water content, fat content and density on thermal conductivity of Model Foods were found to be highly significant (PR > F = 0.0001) while effects of protein and carbohydrate contents were non-significant (PR > F = 0.7543 and PR > F = 0.7445, respectively). Linear models of thermal conductivity as a function of composition and density inadequately explain thermal conductivity of Model Foods. Two relatively simple quadratic models, Model 5 and Model 6, were developed. Model 5 expresses thermal conductivity as a function of water content, water content squared, and density. Model 6 contains all terms in Model 5 and includes the first and second order terms of fat content. Both models did equally well in predicting thermal conductivity of 'real' foods. However, Model 6 did better for high fat content foods. The performance of both models in prediction of thermal conductivity of 'real' foods is comparable to their performance with the Model Foods. Both proposed models were compared with two literature models which express thermal conductivity as a function of composition. The proposed model did equally well or better in predicting thermal conductivity of 'real' foods. The presence of density terms in Model 5 and Model 6 enables the models to predict thermal conductivity of unusually high or low density foods with smaller errors than that of literature models.