The modelling of rupture force of white kidney beans (Phaseolus vulgaris L.) using the multiple linear regression (MLP) and artificial neural networks (ANN)

Authors : Sefa ALTIKAT

Pages : 129-136

Doi:10.20289/zfdergi.554929

View : 29 | Download : 12

Publication Date : 2020-03-31

Article Type : Research Paper

Abstract :Objective: The objective of this study modelling the rupture force of white kidney beans with the multiple linear regression (MLR) and artificial neural networks (ANN). Material and Methods: It was used four different white kidney bean varieties  (Akman, Topçu, Göynük and Karacaşehir) at the five different moisture contents (14.28%, 24.32%, 33.45%, 42.54% and 53.48%). In the MLR and ANN models the moisture contents,  length, width, thickness, arithmetic mean diameters, geometric mean diameters, surface area and  sphericity of the beans were used as input parameters while the rupture force as output parameter. In addition, 24 different ANN architectures were used in the ANN. Results: The highest R 2 values for the Akman (0.979) and Karacaşehir (0.986) varieties were obtained in the ANN11 architecture used by the Levenberg-Marquard learning function and the logarithmic sigmoid - linear transfer function pairs with 12 neurons. However, the best prediction values for Topçu (0.963) and Göynük (0.944) were obtained in ANN 7 and ANN 2 architectures, respectively. In addition, the best pair of learning functions for Topçu and Göynük were observed in Logarithmic sigmoid - Symmetric sigmoid and Logarithmic sigmoid- linear transfer functions, respectively. Conclusion: The results of the study clearly showed that the ANN successfully modeled rupture force in all the white kidney bean varieties.
Keywords : Physical property, mechanical property, simulink, moisture

ORIGINAL ARTICLE URL