Application of Bayesian regularized neural networks for simulating blood nanofluid flow in stenotic arteries

In this study, a unique hyperbolic tangent sigmoid function, which can be divided into velocity and temperature components, is designed for the numerical treatment of the blood nanofluid model (BNM) in stenotic arteries. The proposed model uses 2 hidden layers one utilize 15 and second employ 30 number of neurons. Both hidden layers are created via a procedure known as the hyperbolic tangent sigmoid transfer function. The proposed neural network (NN) model for BNM has been further optimized using Levenberg–Marquardt backpropagation to validate multi-layer NN. A deterministic nanofluid blood model is developed and solved using Adam Bashforth method to generate the required datasets to train, test and validate the proposed NN model. The following proportion of datasets has been utilized 70% for training, 25% for testing and 5% for validation purposes, respectively. It is observed that through the MSE (mean square error) and error histograms that model is highly accurate and stable.