Prediction of Axle Loads Induced by Sugarcane Transport Vehicles using Statistical and Neural-Network Models

Abstract

Abstract Data were collected for trailers transporting single bundle and loose sugarcane, respectively, to develop models for predicting axle loads induced by sugarcane transport vehicles. Based on a statistical approach, linear- regression analysis was performed on the data to relate the axle load with the payload for each trailer. A backpropagation neural network was trained to predict the induced axle loads, with a network consisting of two, eight and two processing units in the input, hidden and output layers, respectively. Input to the network were payloads and empty trailer axle loads. The output corresponded to the measured trailer and tractor rear-axle loads. Using a ±5% residual error interval, the statistical and neural-network models attained over 85% prediction for trailer axle loads. The statistical and neural-network models achieved 85% prediction for tractor rear-axle load induced by loose sugarcane trailers. On the other hand, the neural-network model achieved 70% prediction as compared with 65% prediction achieved by the statistical model for tractor rear-axle load induced by single-bundle trailers. Since the neural network and statistical models had residual error standard deviations of 2·6 and 4·0%, respectively, for single bundle trailers, and 1·2 and 2·1%, respectively, for loose sugarcane trailers, the former model attained more uniform prediction for trailer axle loads than the later model.