Multi-objective Optimization Using Backpropagation Neural Network and Teaching–Learning-Based-Optimization Method in Surface Grinding Under Dry and Minimum Quantity Lubrication Conditions (MQL)

Abstract

AbstractThe present work focuses on the performance modeling of surface grinding to attain an optimum parameter setting for the minimum coefficient of friction and surface roughness. The experimental data were collected during the surface grinding process using dry conditions and minimum quantity lubrication (MQL) as a clean technology lubricant. The usage material was SKD 61 tool steel. The varied surface grinding parameters were the depth of cut and table speed, wherein each had three levels. The surface grinding operation was performed by using a full factorial design 2 × 3 × 3. Backpropagation neural network (BPNN) was first applied to obtain the modeling of the surface grinding experiment, the objective function, the predictions of coefficient of friction, and surface roughness. The objective function is then modified into a fitness function. Finally, this fitness function is utilized in multi-objective optimization using the teaching–learning-based optimization (TLBO) method to attain the surface grinding parameters’ levels that simultaneously produce a minimum coefficient of friction and surface roughness. Based on our experimental results, the combination of BPNN-TLBO can be applied to simultaneously minimize the coefficient of friction and surface roughness in the grinding of SKD 61 by implementing MQL and setting the feeding speed at 150 mm/s and the depth of cut at 0.01 mm. As the result, the minimum surface roughness is 0.376 μm, and the coefficient of friction is 0.333.KeywordsBackpropagation neural networkClean technologyMinimum quantity lubricationSurface grindingTeaching–learning-based optimization