Prediction of surface roughness values of aluminum alloys cut by wire electro erosion machining by machine learning methods

Intensively using the technology, a large data pile was formed, which was increasing rapidly every day. This profits labor and time in many areas as the collection, storage, and processing of large digital data. Since data is meaningless and complex unless processed, they only consume unnecessary storage space and can be turned into meaningful data when processed by data mining or machine learning methods. These stacks are made meaningful and save time for institutions, organizations or individuals. Errors and time factors are minimized by using intelligent systems that will be developed with today's important learning methods such as data mining, machine learning or deep learning. In this study, the concepts are examined considering the importance of smart systems realized with data mining and machine learning. Al7075 alloy, which is one of the aluminum alloys that have an important place in the industry, especially in the aerospace and automotive sectors, has been processed in different parameters by wire electro-erosion processing (WEDM) method and the resulting surface roughness has been predicted by data mining methods. In this research, voltage, pulse duration, pressure, and wire feeding are taken as input parameters. After the WEDM process, obtained a data set with input and output parameters, and surface roughness values were predicted with ELM, W-ELM, SVR, and Q-SVR methods and the results were evaluated. As a result, W-ELM, which has the highest performance among the methods used, predicted the surface roughness with an R^2 value of 0.9720. As a consequence of the work carried out, predicted that the aluminum alloy has great potential in the manufacturing industry, will achieve high gains in labor, time, and cost.