Effects of synthetic data applied to artificial neural networks for fatigue life prediction in nodular cast iron

The prediction of fatigue life is essential in the development of products to avoid unexpected failures during their useful life. Although different linear and nonlinear damage accumulation approaches have been proposed, no model has been as universally used as Miner's linear damage rule due to its simplicity and life prediction results. Discrepancies in the prediction of fatigue life are present within the manufacturing process, which is generated from the material through the manufacturing process and during applied loads. Owing to new design application areas, such as in biomedical devices and the aerospace industry, among others, the development of new ways to reduce errors in predicting fatigue has become an increasing necessity. This paper addresses fatigue life prediction improvement when if performed through a combination of synthetic data an artificial neural networks (ANNs). The novelty of this work is based on the proposal and validation of virtual synthetic fatigue data as a complementary input parameter in the ANN. For the design of the ANN, 116 experimental results of nodular cast iron direction knuckles were analyzed. As seen during the validation process, the employment of synthetic data as input increased significantly the forecast of the ANN.