Fatigue damage effect approach by artificial neural network

This study is concerned with the fatigue strength behaviour of chassis components made of steel S420MC. Experimental results show differences when applying sequences of loads, but also when the effect of the operating temperature is taken into account for the prediction of the fracture in the component. Artificial neural networks are a suitable way to establish a relationship between the sequence effects and the fatigue life. To achieve this, the artificial neural network was trained to predict the damage on a rear axle-mounting bracket. Experimental tests were developed at constant and variable amplitudes, defined as load sequences. A series of experimental tests was performed with temperatures of 23 degrees C (room temperature), 35 degrees C and 45 degrees C to evaluate their effect with the different load sequences. Although the maximum temperature used in the experimental set up was only 3% of the melting temperature, differences in the damage to the component were found. The artificial neural network was trained and validated with 68 experimental results to predict the damage of different loading sequences. The artificial neural network demonstrated a higher prediction capability at some load sequences in comparison to the damage rule.