| Titles |
|
English :
|
Using Neural Networks and Linear Regression For Predicting Energy Requirements in Plowing
|
|
Arabic :
|
إستخدام الشبكات العصبية والإرتداد الخطى لتوقعات الطاقة اللازمة للحرث.
|
|
| Abstract |
In this research work, two methods were utilized to predict both fuel consumption and energy requirements during tillage process under different operation conditions. The first method was multiple linear regression (MLR) and the second one was artificial neural network (ANN). Data of three plows were obtained from (Mamkagh, 2002), the first plow is chisel , the second is moalboard and the third one is disk. These plows performed tillage on clay loam soil surface changing in moisture (8, 11 and 14 % , d.b). Three forward speed were utilized namely, 5.5, 7.0 and 9.0 km/h. During executing the MLR to obtain the regression coefficients, factors were used as 1 for chisel, 2.5 for disk and 2. for moalboard. These factors are the ratios of the average of fuel consumption of selected plow to fuel consumption of chisel plow. Multilayer feedforward network with the error backpropagation learning algorithm was used to build the ANN model. The ANN was trained and tested with: plow type (chisel, moldboard and disk), soil moisture and forward speed as input parameters and fuel consumption and energy requirements as output parameters. The training data were 18 observations. The architecture of the ANN was consisted of three layers, the first layer for inputs, the second was hidden layer with 16 processing elements, and the third layer for output. The hidden layer and the output layer had been (Tanh) transfer function. The learning rate was 0.098 and the best results were achieved at 60000 training runs, which gave minimum root mean squared error (RMSE) equals to 0.000246 during training process. The results during testing the ANN showed that the variation of observed and predicted fuel consumption and energy requirements was small and coefficient of determination (R2) and root mean squared error (RMSE) were 0.98 and 2.699 MJ/fed for energy requirements prediction, while these values of (R2) and RMSE were 0.96 and 3.729 MJ/fed using MLR respectively. The study suggests that the ANN approach is very useful for predicting fuel consumption (lit/fed) and energy requirements (MJ/fed) although the relationship between them and the affecting variables is unseen.
|
| Publication year |
2002
|
| Pages |
126 - 115
|
| Availability location |
معهد بحوث الهندسة الزراعية- ب 256 الدقى – 12311 جيزة – جمهورية مصر العربية.
|
| Availability number |
|
| Organization Name |
Agricultural Engineering Research Institute (AENRI)
|
| serial title |
تم النشر بالمؤتمر العاشر للجمعية المصرية للهندسة الزراعية فى 16-17 أكتوبر 2002 م.، عدد خاص
|
| Department |
Agriculture Power and Energy
|
| Author(s) from ARC |
|
| External authors (outside ARC) |
|
| Agris Categories |
Agricultural engineering
|
AGROVOC
TERMS |
Energy.
Network analysis.
|
| Proposed Agrovoc |
Plowing;
|
| Publication Type |
Conference/Workshop
|