Energy Harvesting Systems via Peristaltic Micro‐Pumps in Targeted Therapies: Comparative Analysis of Hybrid RSM–ANN Model

ABSTRACT

The movement of fluid caused by the contraction and relaxation of flexible conduit walls is known as peristalsis, and it is a phenomenon seen in biological systems such as blood and food flow. In industrial operations, microfluidics, and medicinal applications, understanding heat convection in peristaltic flows is essential. In order to develop medication delivery systems, optimize flow parameters for uniform distribution, and forecast performance under various physiological situations, it is imperative to model heat convection in peristaltic flow. Understanding heat transfer in blood arteries under peristaltic flow aids in temperature regulation in hyperthermia‐based cancer therapies, and artificial neural network models help with real‐time treatment parameter management. This work addresses the peristaltic flow of a Newtonian fluid in a symmetric channel that is governed by three nonlinear coupled partial differential equations. The novelty of the present study lies in combining numerical simulation with response surface methodology and artificial neural networks to analyze and predict pressure rise behavior in peristaltic flow through a porous symmetric channel. This hybrid approach improves prediction accuracy while efficiently capturing the nonlinear characteristics of the flow system. The approximations of long wavelength and low Reynolds number reduce these to nonlinear coupled ordinary differential equations. The solutions obtained using MATLAB's bvp4c routine yield several numerical findings. An empirical relation for is developed using the response surface methodology (RSM) and artificial neural networks (ANNs). The accuracies of the developed empirical relations are verified by calculating the coefficient of determination, , along with an analysis of variance. In this case, high  = 98.06% and adjusted = 96.32% values clearly show a good fit of the model. The sensitivity analysis shows that is maximum when the Darcy number, Brinkman number, and Grashof number increase. The primary computation includes mean squared error computation, error histogram generation, and a regression line in the case of ; it results in an MSE of around 1.1839e−09, thereby strongly correlated with the data and model.