Cutting the Titanium alloys is a complicated task which cannot be performed by traditional methods and modern machining processes, such as Wire electro-discharge machining (WEDM) process which are mainly used for this purpose. As a result of the high price of the Ti-6Al-4V alloy, proper tuning of the input parameters so as to attain a desired value of the surface roughness is an important issue in this process. For this purpose, it is necessary to develop a predictive model of surface roughness based on the input process parameters. In this paper, The Taguchi method was used for the design of the experiment. According to their effectiveness, the input parameters are pulse-on time, pulse-off time, wire speed, current intensity, and voltage; and the output parameter is surface roughness. However, a predictive model cannot be defined by a simple mathematical expression as a result of the complicated and coupled multivariable effect of the process parameters on the surface roughness in this process. In this study, application of the relevance vector machine as a powerful machine learning algorithm for modeling and prediction of surface roughness in wire electro-discharge machining for Ti-6Al-4V titanium alloy has been investigated. The predicting result of model based on the root means square error (RMSE) and the coefficient of determination (R2) statistical indices, prove that this approach provides reasonable accuracy in this application. پرونده مقاله

The Electro-discharge coating process is an efficient method for improvement of the surface quality of the parts used in molds. In this process, Material Transfer Rate (MTR), an average Layer Thickness (LT) are important factors, and tuning the input process parameters to obtain the desired value of them is a crucial issue. Due to the wide range of the input parameters and nonlinearity of this system, the establishment of a mathematical model is a complicated mathematical problem. Although many efforts have been made to model this process, research is still ongoing to improve the modeling of this process. To this end, in the present study, three powerful machine learning algorithms, namely, Relevance Vector Machine (RVM), Extreme Learning Machine (ELM) and the Least Squares Support Vector Machine (LS-SVM) that have not been used to model this process, have been used. The values R2 above 0.99 for the training data and above 0.97 for the test data show the high accuracy and generalization capability degree related to the LS-SVM models, which can be applied for the input parameters tuning in order to attain a preferred value of the outputs. پرونده مقاله