Artificial Neural Network Model as a statical analysis tool in pipe-framed greenhouse design

Author: Unal Kizil ^1* and Melis Sacan ²

Received 12 January 2010, accepted 8 April 2010.

Abstract

The aim of this study was to investigate the performance of a feed-forward back-propagation artificial neural network (ANN) model on design of single span, pipe-framed, plastic-covered greenhouse structures. Statical analysis of different size greenhouses (n = 75) was conducted using Kleinlogel formulas to develop a database in Microsoft Excel. Frame dimensions for database development were randomly selected using MS Excel. Sixty percent of the database was used to train ANN model. The remainder of the data was used to validate the model. The model parameters were selected by trial and error method. It was shown that a single hidden layer with 25 nodes and a standard sigmoid function yielded the highest classification results. The model was able to assign a pipe diameter for roof and columns of the frame. Number of errors was “0” in classification. Probability values for classes were between 0.50 and 0.99. Based on the results of this study, ANN seems to be a useful method in the statical design of pipe-framed, single span greenhouses.

Key words: Greenhouse, statical analysis, artificial neural networks, frame systems.