Surface roughness prediction of machined aluminum alloy with wire electrical discharge machining by different machine learning algorithms

Abstract

Aluminum alloys are preferred in aviation, aerospace and automotive industries because of their high strength and durability compared to their lightness. Precision production of parts is very important in such industries. Therefore, precision machining of aluminum, which is difficult to manufacture with traditional methods, with non-traditional methods such as wire electrical discharge machining (WEDM), is a very popular approach. Surface roughness has an impact on the important properties of materials such as strength, wear resistance and fatigue strength. Experimental determination of surface roughness of surfaces machined with WEDM is time consuming and costly. These cost and time losses can be eliminated by predicted surface roughness with machine learning algorithms. In this study, Al7075 aluminum alloy was machined with different parameters (voltage, pulse-on-time, dielectric pressure and wire feed) with WEDM. Each parameter is at 3 levels, so 81 experiments were carried out. The surface roughness of the machined surfaces was measured by surface profilometer. The lowest surface roughness was 2.490 μm machined at 8 V voltage, 8 μs pulse on-time, 25 bar dielectric pressure and 2 mm/min wire feed. The experiments for machining of Al7075 via WEDM were modeled by machine learning methods. Four different models of two different methods were used for the prediction of surface roughness values of machined samples with WEDM. These models were ELM, W-ELM, SVR and Q-SVR. All of the models were applied to the data set and the W-ELM model was the best performing model with the value of 0.9720 R2. Thus, the W-ELM model has excellent potential in manufacturing industry which produced parts with WEDM.