Thermal modeling of Johnson–Segalman nanofluid in blade coating process: a comparative study with machine learning framework

Abstract This study simulates the blade coating process of Johnson–Segalman (JS) nanofluid using machine learning, accounting for magnetic field, thermophoresis, and Brownian motion effects. Expressions are simplified by lubrication approximation theory (LAT) and numerically solved by shooting technique. A supervised neural network employing the Levenberg–Marquardt backpropagation (LMBP-SNN) algorithm was trained, tested, and validated using these numerical solutions by using regression plots and mean squared error (MSE) analysis. It is evident that there is a strong correlation between the predictions of the ANN and the numerical model. However, it should be noted that the ANN-LM model demonstrated outstanding accuracy that reached a minimum mean squared error 9.946 × 10 −8 at epoch 310, indicating good convergence stability and strong generalization ability. Results indicate increasing Weissenberg number (We = 0.4–0.8) leads to a reduction of 15 % in axial velocity as a result of increased elasticity resistance in the middle of the coating zone. The results show that the non-Newtonian parameter and the Hartmann number are the primary governing elements in decreasing the coating thickness, improving coating efficiency and shelf life of web.