Modeling and prediction of fatigue life of brass and EN24 steel using soft computing tool

This paper presents an approach for predicting the fatigue life of Brass and EN24 steel using Artificial Neural Network (ANN). The input required for the ANN model such as surface roughness, materials specifications etc have been obtained by conducting experiments on eighteen standard fatigue test specimens in the laboratory. The effect of cyclic loading range and surface roughness (Ra index) on to the fatigue life of brass and EN24 steel were discussed. Using multilayer feedback network (Levenberg–Marquardt), the ANN model was developed to predict the parameters like maximum stress and number of cycles. The ‘nntool’ in MATLAB toolbox has been used for training and testing the data in ANN model. The ANN model with five input neurons, one hidden layer with three neurons and two output neurons is used to get accurate results. The coefficient of regression (R2) for the ANN model is 0.999. Thus, the comparison between the experimental results and predicted values using ANN was agreeable and good correlation between the results were achieved. From the experimental study, it is concluded that the change in the Ra index from 2 to 0.8 μm, increases the fatigue life by 20% which shows that surface roughness also plays important role in finding the fatigue behaviour of the materials.