Volume 13, Issue 52 (2016)                   FSCT 2016, 13(52): 161-172 | Back to browse issues page

XML Persian Abstract Print


Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:

Khayati S, Amiri Chayjan R. Prediction of some thermal, physical and mechanical properties of terebinth fruit after semi-industrial continuous drying using artificial neural networks. FSCT 2016; 13 (52) :161-172
URL: http://fsct.modares.ac.ir/article-7-2027-en.html
Abstract:   (4318 Views)
  The purpose of this study was prediction of thermal (effective moisture diffusivity and specific energy consumption), physical (shrinkage and color) and mechanical properties (rupture force) of terebinth fruit in a semi industrial continuous dryer using artificial neural networks (ANNs). Three effective factors on thermal, physical and mechanical properties, were air temperature, air velocity and belt linear speed as independent variables. Experiments were conducted with a semi industrial continuous dryer in temperature levels of 45, 60, 75 °C, air velocity levels of 1, 1.5 and 2 m/s and belt linear speed levels of 2.5, 6.5, 10.5 mm/s. Necessary data were collected using a the semi-industrial continuous dryer. Feed and cascade forward back propagation networks with learning algorithms of Levenberg-Marquardt and the Bayesian regulation were used to train the patterns. To predict the effective moisture diffusivity, feed forward networks with the Bayesian regulation, topology of 3-10-13-1 and 108 training cycles with R2=0.9999 was optimal arrangement. The optimal topology to predict the specific energy consumption was 3-10-1 with feed forward network, Levenberg-Marquardt algorithm, 117 training cycles and R2=0.9961. The best network for shrinkage prediction was feed forward network with the Bayesian regulation algorithm, topology of 3-6-4-1, 101 training cycles and R2=0.9926. To predict the total color change, feed forward networks with the Levenberg-Marquardt algorithm, topology of 3-6-7-1, 24 training cycles and R2=0.9139 was the optimal arrangement. The best network to predict the rupture force was feed forward network trained with the Bayesian regulation, topology of 3-8-6-1, 69 training cycles and R2=0.9990.
Full-Text [PDF 220 kb]   (2319 Downloads)    

Received: 2015/04/21 | Accepted: 2015/09/23 | Published: 2016/05/21

Add your comments about this article : Your username or Email:
CAPTCHA

Rights and permissions
Creative Commons License This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.