Tractor Drawbar Performance Prediction Using Artificial Neural Network

Abstract

Prediction of tractor drawbar performance can lead to simulation and optimization of tractor performance, allowing optimum setting of different parameters as well as guiding manufacturer in decision-making for design of new tractors. Twenty different input parameters were selected for drawbar performance prediction. The data used as input to train the network was collected from 141 tractor test reports tested between 1997 and 2013 at the Central Farm Machinery Training and Testing Institute, Budni (M.P.). A back propagation artificial neural network (ANN) was developed using Neural Network Toolbox in Matlab software. Matrix of 1140×20 and 1140×1 was made as input and target values for drawbar prediction in the ANN. The optimum structure of neural network was determined by trial-and-error method, and 30 different structures were evaluated. Highest performance was obtained for the network with two hidden layers, each having 35 neurons, and employed Levenberg-Marquardt training algorithm. Coefficient of determination (R2) and Mean square error (MSE) for this neural network was 0.994 and 1.284, respectively.