A Hybrid Process Model for EDM Based on Finite-Element Method and Gaussian Process Regression
Wuyi Ming,Guojun Zhang,He Li,Jianwen Guo,Zhen Zhang,Yu Huang,Zhi Chen
DOI: https://doi.org/10.1007/s00170-014-5989-y
2014-01-01
Abstract:This paper proposed a hybrid intelligent process model, based on finite-element method (FEM) and Gaussian process regression (GPR), for electrical discharge machining (EDM) process. A model of single-spark EDM process has been constructed based on FEM method, considering the latent heat, variable heat distribution coefficient of cathode (f (c) ), and plasma flushing efficiency (PFE), to predict material removal rate (MRR) and surface roughness (Ra). This model was validated using reported analytical and experimental results. Then, a GPR model was proposed to establish relationship between input process parameters (pulse current, pulse duration, and discharge voltage) and the process responses (MRR and Ra) for EDM process. The GPR model was trained, tested, and tuned using the data generated from the numerical simulations. Through the GPR model, it was found that responses of EDM process can be accurately predicted for the chosen process conditions. Therefore, for the selection of optimum parameters, the hybrid intelligent model proposed in this paper can be used in EDM process.