Document Type : Research Article
Authors
1 Department of Biosystems Engineering, Faculty of Agriculture, University of Kurdistan, Kurdistan, Iran
2 Department of Agricultural Machinery Engineering, Faculty of Agriculture, University of Tabriz, Tabriz, Iran
Abstract
Introduction
The noticeable proportion of producing wheat losses occur during production and consumption steps and the loss due to harvesting with combine harvester is regarded as one of the main factors. A grain combines harvester consists of different sets of equipment and one of the most important parts is the header which comprises more than 50% of the entire harvesting losses.
Some researchers have presented regression equation to estimate grain loss of combine harvester. The results of their study indicated that grain moisture content, reel index, cutter bar speed, service life of cutter bar, tine spacing, tine clearance over cutter bar, stem length were the major parameters affecting the losses.
On the other hand, there are several researchswhich have used the variety of artificial intelligence methods in the different aspects of combine harvester.
In neuro-fuzzy control systems, membership functions and if-then rules were defined through neural networks. Sugeno- type fuzzy inference model was applied to generate fuzzy rules from a given input-output data set due to its less time-consuming and mathematically tractable defuzzification operation for sample data-based fuzzy modeling. In this study, neuro-fuzzy model was applied to develop forecasting models which can predict the combine header loss for each set of the header parameter adjustments related to site-specific information and therefore can minimize the header loss.
Materials and Methods
The field experiment was conducted during the harvesting season of 2011 at the research station of the Faulty of Agriculture, Shiraz University, Shiraz, Iran. The wheat field (CV. Shiraz) was harvested with a Claas Lexion-510 combine harvester. The factors which were selected as main factors influenced the header performance were three levels of reel index (RI) (forward speed of combine harvester divided by peripheral speed of reel) (1, 1.2, 1.5), three levels of cutting height (CH)(25, 30, 35 cm), three levels of the horizontal distance of reel tine bar from cutter bar (Hd)( 0, 5, 10 cm) and three levels of vertical distance of reel tine bar from cutter bar (Vd)( 5, 10, 15 cm) which are taken as the input variables for neuro-fuzzy model and only combine header loss is output of the model.
Some frames with the dimensions of 50 × 50 cm2 were randomly used to determine the amount of header loss. In order to determine the header loss, the frame was placed on the ground in the vacant place behind the cutter bar, where output material from the back of the combine was not allowed to pour on the ground. Grains and ears found inside the frame were gathered, weighed and then the amount of pre-harvest loss was subtracted from it. A fractional factorial design based on a completely randomized design was used to determine the header loss. Each test was repeated three times and for each repetition.
The structure of neuro- fuzzy model for this study has four inputs and each input variable as mentioned and three Gaussian membership functions (mf), result in 81 rules.
Results and Discussion
A neuro- fuzzy model was developed for predicting the combine header loss based on reel index, cutting height, the horizontal distance of reel tine bar from the cutter bar and vertical distance of reel tine bar from cutter bar as input variables. The Model has three membership functions for each input. Gaussian membership functions and rules were defined for knowledge representation of header loss.
Predicting header loss is an important issue for minimizing the amount of harvest grain losses. Neuro-fuzzy model presented a satisfactory application to describe header loss of a combine harvester. It showed R² equal to 0.95 which is superior to multiple regression method with 0.71. In fact, the amount of coefficient of determination is a good indicator to check the prediction performance of the model. Based on developed neuro-fuzzy system model, levels of reel index, cutting height, the horizontal distance of reel tine bar from cutter bar and vertical distance of reel tine bar from cutter bar could be recommended according to minimize header loss.
Conclusions
In the final step, the designed controller was simulated in SIMULINK. The Controller can change setting of header components in order to their impaction gathering loss and in each step, compare gathering loss with optimal value and If it was more than optimum then change the settings again. The simulation results were evaluated satisfactory.
Keywords
Send comment about this article