Research on Establishing a Predictive Model for Surface Roughness of Al6061 Aspherical Reflector in Single Point Diamond Turning Using Back Propagation Artificial Neural Networks
Year: 2026
Type: article
Abstract: The paper presents establishing of a predictive model for surface roughness value of Al6061 material aspherical surface in ultra-precision turning using a single point diamond tool (SPDT). The model employs the structure of a back propagation artificial neural network (ANN). Three cutting parameters are considered, including spindle speed, feed rate, and depth of cut. An experimental matrix was established based on data from 30 actual experiments measuring surface roughness values of aspherical surfaces under corresponding machining conditions for the given parameter sets. Through the evaluation of six specific neural network structures (based on the number of layers and neurons in each layer), an optimal neuron ratio between layers was determined to optimize the predictive model. The ANN structure 3-5-15-1 yielded the best prediction results, as demonstrated by evaluation metrics such as: Coefficient of Determination (R2 equal 0.9999), Mean Square Error (MSE equal 2.6e-4), Root Mean Square Error (RMSE equal 0.0163) and Mean Absolute Percent Error (MAPE equal 0.6949%).Validation experiments, involving six training sessions using MATLAB software, confirmed the high feasibility of the predictive model. This was evidenced by the minimal error (1-2%) between the predicted surface roughness values and the experimental measured roughness values. This research is directly applied to predic surface roughness value of Al6061 aspherical surface in SPDT and serves as foundation for similar studies on different material or surface geometries of machined parts.
Language: English
Subfield: Mechanical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W7160495338