In the present work, a novel magnetically assisted externally supplied abrasive-based finishing (MESAF) process was modeled using a data-driven artificial neural network (ANN). A MESAF process uses a ferromagnetic brush and abrasive particles are supplied from an external source in the form of a suspension. The MESAF process overcomes the limitations of the traditional magnetic abrasive finishing process, particularly the restriction on the number of abrasive particles and their real time replenishment. The effect of process parameters on change in surface roughness ( \({\Delta R}_{a}\) ) has been experimentally analyzed, and a second-order regression equation was developed. In contrast to regression, which necessitates defining a specific functional form (linear, polynomial, etc.), ANNs learn directly from the input data, allowing them to be more flexible and adaptable to changes in the process parameters. A feed-forward back-propagation neural network (FFBPN) model with the Levenberg–Marquardt (LM) algorithm as the training method was used. Additionally, a support vector regression (SVR) model based on the support vector machine (SVM) approach was implemented to compare the performance of the ANN model with SVR. The maximum and average absolute errors in predicting \({\Delta R}_{a}\) , using the ANN modeling, were found to be 2.81 \(\%\) and 0.58 \(\%,\) respectively, for SVR they were 6.25 \(\%\) and 2.04 \(\%\) whereas, for regression modeling, they were 9.17 \(\%\) and 3.61 \(\%,\) respectively. An overall R2 using regression analysis was 0.9486. In contrast, the overall R2 of 0.98596 was observed with the FFBPN model, demonstrating that predictions obtained through ANN were closer than those obtained through the regression. The training function in the FFBPN model was also modified with other available training functions; however, the training function LM outperformed other training functions. Additionally, the surface texture achieved through the MESAF process was analyzed using SEM and atomic force microscopy (AFM) micrographs and is reported in the article.
