Application of Artificial Neural Network and Genetic Algorithm for Predicting three Important Parameters in Bakery Industries




Farinograph is the most frequently used equipment for empirical rheological measurements of dough. It’s
useful to illustrate quality of flour, behavior of dough during mechanical handling and textural
characteristics of finished products. The percentage of water absorption and the development time of dough
are the most important parameters of farinography for bakery industries during production. However,
farinograph quality number is also a profitable factor for rapid evaluation of flour. Our purpose in present
research is to apply artificial neural networks (ANNs) for predicting three important parameters of
farinograph from simple measurable physicochemical properties of flour. Genetic algorithm (GA) was also
applied in the training phase for optimizing different parameters of ANN’s structure and inputs. Sensitivity
analyses were also conducted to explore the ability of inputs in predicting the networks outputs. Two neural
networks were developed; the first for modeling water absorption and dough development time and the
second for modeling farinograph quality number. Both developed ANNs using GA have excellent potential
in predicting the farinograph properties of dough. In developed models, gluten index and Zeleny, suitable
parameters for qualitative measurements of samples, played the most important role for predicting dough
farinograph characterisations.