Artificial neural network modeling of creep behavior in a rotating composite disc

Purpose - This paper aims to explore the capabilities of artificial neural network (ANN) for predicting the creep response of a rotating Al-SiC(p) composite disc operating at elevated temperature. Design/methodology/approach - Mathematical modeling of the steady state creep behavior, as described by Sherby's law, of a rotating disc made of isotropic aluminium-silicon carbide particulate composite has been carried out. The creep response has been calculated for various combinations of particle size, particle content and temperature by extracting creep parameters from the limited experimental creep data available on similar material. The results thus obtained are used to train the ANN based on back propagation learning algorithm with particle size, particle content and temperature as input and stress and strain rates as output parameters. The trained network is used to predict the stresses and strain rates in the disc for the data set not covered in the training of network. The predictions obtained from the ANN model have been compared with the corresponding analytical values. Findings - A nice agreement between the ANN predicted and analytical values of the creep stresses and strain rates has been observed. Originality/value - ANN can be used as a reliable tool for investigating the effect of operating temperature and, reinforcement-size and -content, on the creep behavior of a rotating composite disc to reach at optimum design code.