Use of a Convolutional Neural Network for Predicting Fuel Consumption of an Agricultural Tractor

The energy crisis and depleting fossil fuel resources have always been the focus of researchers. Fuel consumption agricultural tractors is not an exception. Researchers used different methods to predict consumption. With development artificial intelligence in last decade, all re-searchers’ attention has directed towards it. Deep learning a subset machine learning, which was inspired by data processing patterns human brain. deep method research due advantages high accuracy generalization. So far, no this In research, field experiments were carried out sandy clay loam soils model temporal specific tractor using convolutional neural network (CNN), while having some parameters such as soil type, conditions, tool parameters, operation pa-rameters. conducted within each texture factorial manner based on randomized complete block design (RCBD) with three replicates. For texture, various moisture levels (8–17% for dry 18–40% moist soils), forward speeds (1.2, 1.6, 1.8, 2.2 km h−1), working depths (30 50 cm), number passes (2 6), tire inflation pressure (20 25 psi) selected, cone index, dynamic load, content measured experimental section. designed networks instant CNN type. results indicated that developed Sgdm algorithm outperformed others, thus it selected modeling purposes. evaluated R2 MSE criteria. consumption, best obtained 8-510-510-1 architecture = 0.9729 0.0049. 8-100-95-1 also led prediction 0.9737 0.0054. low error compared previous studies indicate superiority order It observed from input include soil, tool, operational are effective Proper management depth, pressure, speed, can help optimize