Artificial Neural Network Modeling of Titanium Alloy Tribological Behaviour in Beta Solution Treated Condition

Abstract

In the present investigation artificial neural network (ANN) approach was used for the prediction of wear and friction properties of low cost near beta titanium alloy β solution treated condition. The input parameters are load, track diameter and β solution treated temperature and output parameters are %weight loss, coefficient of friction and temperature generated between the pin and disc. In order to get the best model, different parameters like number of layers, number of hidden neurons, and transfer functions are changed. The data obtained in sliding wear tests were divided into two sets training data and testing data. A neural network was trained using a training data set and was validated using test data. The best network for prediction of tribological properties of these β solution treated specimens was 3-[11]1-[9]2-3 layer recurrent with purelin transfer function and trainlm is training function. The percentage error for %weight loss, coefficient of friction and temperature are 2.8, 1.7 and 5.3 respectively.