Surface roughness prediction and optimization in the REMF process using an integrated DBN-GA approach

Surface roughness is a crucial factor affecting the surface quality of workpieces in manufacturing industries. Thus, it important to provide an accurate performance prediction and optimal parameters reduce burden time costs during process. In this study, two predict models, namely multiple linear regression deep belief network (DBN) were performed accurately change rotational electromagnetic finishing (REMF). Compared statistical-based model, data-driven model based on DBN architecture was significantly considerable effect REMF Among considered DBN5 as [7, 14, 1] showed effective features nonlinear relationship between process response with highest determination coefficient (R2) 0.9340 lowest mean squared error (MSE) 1.3037 $$\times$$ 10−3 testing datasets. addition, genetic algorithm (GA) heuristic optimization technique adopted optimize input best derived model. It that maximum 0.530 at particle length 3 mm, diameter 0.7 weight 1.3 kg, liquid water quantity 1.0 l, speed 1323 rpm, working 35 min, initial 2.5478 m $$\upmu$$ .