Evaluation the performance of artificial intelligence algorithms to predict surface roughness in Turning processes

Abstract

An important characteristic of products must be the degree of smoothness of their surfaces, as many practical applications need surfaces on a high degree of smoothness in order to perform the required function as best as possible. Surface roughness is important in being a basic indicator for measuring material quality. In manufacturing and production processes, it is very important but difficult to know the surface roughness by relying on inputs (cutting speed _ feedrate _ depth of cut). Some studies have indicated that by increasing speed or decreasing depth of cut the surface roughness is improved and therefore the impact of cutting parameters (cutting speed _ feed rate _depth of cut) on metal has been studied AISI1040 to know the impact of cutting parameters in the turning process, this metal was selected because of its important and many uses in the industry, the most important of which are: bearings _ crankshaft etc. It was found that the feedrate is only by increasing the surface rough  is increased while the cutting speed and feeding rate cannot know its optimal values for the smoothness of a good surface. In order to find out the surface roughness before operations, we were able to use machine learning algorithms (linear regression _ decision tree_ neural networks_ random forest) to predict surface roughness. Real values obtained from the computer-aided design and production laboratory at the Faculty of Mechanical and Electrical Engineering at Damascus University were used; Neural networks have been shown to give the best prediction compared to the decision tree, linear regression and random forest, With the possibility of using this model to predict the surface roughness in the future.