Premium
Evaluating the Combined Effect of Temperature, Shear Rate and Water Content on Wild‐Flower Honey Viscosity Using Adaptive Neural Fuzzy Inference System and Artificial Neural Networks
Author(s) -
AlMahasneh M.A.,
Rababah T.M.,
Ma'Abreh A.S.
Publication year - 2013
Publication title -
journal of food process engineering
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.507
H-Index - 45
eISSN - 1745-4530
pISSN - 0145-8876
DOI - 10.1111/jfpe.12014
Subject(s) - adaptive neuro fuzzy inference system , artificial neural network , shear rate , rheology , viscosity , inference system , biological system , mean squared error , artificial intelligence , mathematics , materials science , computer science , fuzzy logic , fuzzy control system , statistics , composite material , biology
Handling, processing and shelf life determination of honey requires the knowledge of rheological properties. In this study, rheological properties of eight different untreated wild‐flower honey samples collected in J ordan were modeled using three important parameters including temperature (28–58 C ), water content (16.1–17.3%) and shear rate (2.2–47/s). Artificial neural network ( ANN ) (feed forward [ FF ] and radial basis function [ RBF ] ANNs ) and adaptive neural fuzzy inference system ( ANFIS ) were used because conventional analytical models were insufficient. The results showed that both ANNs and ANFIS were able to model honey viscosity very well. RBF ‐ ANNs and FF ‐ ANNs were able to model viscosity with R 2 from 0.961 to 0.986 and mean square error ( MSE ) of 1.43–4.71. An important analysis of input variables showed that shear rate was the most influential factor with 60% weight, followed by temperature (25% weight) and finally water content (15% weight). ANFIS was also found to adequately model honey viscosity using three, four and five triangular and bell‐shaped membership functions. The results showed that R 2 and MSE varied between 0.953 and 0.984 and between 1.41 and 4.24, respectively. ANFIS provided an extra advantage over ANNs because it presented a set of fuzzy if‐then rules that can be used for predictions of new viscosity data. It was concluded that both ANNs and ANFIS were able to provide an excellent alternative to conventional analytical rheological models. Practical Applications Determination of honey rheological properties is of great interest to honey keepers, processors and handlers. It is difficult to predict the combined effect of temperature, water content and shear rate on honey rheological properties using conventional rheological models. More recent intelligent techniques such as artificial neural networks and neural fuzzy logic systems have been alternatively used to model such complex systems. This paper deals with the implementation of intelligent modeling to predict untreated honey viscosity as a function temperature, shear rate, water content and geographic location, which cannot otherwise be modeled using conventional analytical models.