Micro-electrical discharge machining (micro-EDM) is a potential non-contact method for fabrication of biocompatible micro devices. This paper presents an attempt to model the tool electrode wear in micro-EDM process using multiple linear regression analysis (MLRA) and artificial neural networks (ANN). The governing micro-EDM factors chosen for this investigation were: voltage (V), current (I), pulse on time (Ton) and pulse frequency (f). The proposed predictive models generate a functional correlation between the tool electrode wear rate (TWR) and the governing micro-EDM factors. A multiple linear regression model was developed for prediction of TWR in ten steps at a significance level of 90%. The optimum architecture of the ANN was obtained with 7 hidden layers at an R-sq value of 0.98. The predicted values of TWR using ANN matched well with the practically measured and calculated values of TWR. Based on the proposed soft computing-based approach towards biocompatible micro device fabrication, a condition for the minimum tool electrode wear rate (TWR) was achieved.
|Journal||Journal of Machine Engineering|
|Publication status||Published - 2017|
- Micro devices
- Electrical discharge machining
- Multiple linear regression
- Artificial neural networks