Predicting Surface Roughness using Keras DNN Model

In the manufacturing industries, computer vision techniques play an important role in predicting the surface roughness of the workpiece in turning operations. The regression method has become more popular in the implementation of deep learning as it is the most basic application and. In this study, a regression with Keras deep neural network model is used to accurately evaluate the correlation between surface image characteristics and actual surface roughness using the machining parameters and the gray level of the surface image. The obtained results showed that the Keras Deep Neural Network model with a training data set of 57 values has an average accuracy of 80.52 % for predicting surface roughness in turning operations.