Author/Authors :
Farokhzad، Saeid نويسنده Ph.D Student of Mechanical Engineering of Agricultural Machinery, University of Urmia, Urmia, Iran , , Asadi Asad Abad، Mohammad Reza نويسنده Department of Mechanical Engineering, Buinzahra branch, Islamic Azad University, Buinzahra, Iran , , Ranjbar Kohan، Mohammad نويسنده Department of Mechanical Engineering, Buinzahra branch, Islamic Azad University, Buinzahra, Iran ,
Abstract :
ABSTRACT: A neural network simulator built for prediction of faults in gearbox is discussed. A back propagation learning algorithm has been employed. The layers are constituted of nonlinear neurons and an input vector normalization scheme has been built into the simulator. Experiments are conducted on a procedure recognizes gears faults of a typical gearbox of Massy Ferguson tractor to generate training and test data. Feature vector is one of the most significant parameters to design an appropriate neural network. Features to apply as input to ANN were extracted from analysis of vibration signals in time-frequency domain by means of wavelet transform method. The classified network outputs are worn, broken-teeth of gear and faultless condition. In developing the ANN models, different ANN architectures, each having different numbers of neurons in hidden layer, were evaluated. The optimal model was selected after several evaluations based on minimizing of mean square error (MSE) and correct classification rate (CCR). Network training is carried out for a variety of inputs. The adaptability of deferent architectures is investigated. The results show that the designed system is capable of classifying records with 93.65% accuracy with one hidden layers of neurons in the neural network. The networks are validated for test data with unknown faults.
