The application of artificial intelligence (AI) in modeling of various machining processes has

been the topic of immense interest among the researchers since several years. In this direction,

the principle of fuzzy logic, a paradigm of AI technique, is effectively being utilized

to predict various performance measures (responses) and control the parametric settings of

those machining processes. This paper presents the application of fuzzy logic to model two

non-traditional machining (NTM) processes, i.e. electrical discharge machining (EDM) and

electrochemical machining (ECM) processes, while identifying the relationships present between

the process parameters and the measured responses. Moreover, the interaction plots

which are developed based on the past experimental observations depict the effects of changing

values of different process parameters on the measured responses. The predicted response

values derived from the developed models are observed to be in close agreement with those

as investigated during the past experimental runs. The interaction plots also play significant

roles in identifying the optimal parametric combinations so as to achieve the desired

responses for the considered NTM processes.

Wire electrical discharge machining (WEDM) is a non-conventional material-removal process where a continuously travelling electrically conductive wire is used as an electrode to erode material from a workpiece. To explore its fullest machining potential, there is always a requirement to examine the effects of its varied input parameters on the responses and resolve the best parametric setting. This paper proposes parametric analysis of a WEDM process by applying non-parametric decision tree algorithm, based on a past experimental dataset. Two decision tree-based classification methods, i.e. classification and regression tree (CART) and Chi-squared automatic interaction detection (CHAID) are considered here as the data mining tools to examine the influences of six WEDM process parameters on four responses, and identify the most preferred parametric mix to help in achieving the desired response values. The developed decision trees recognize pulse-on time as the most indicative WEDM process parameter impacting almost all the responses. Furthermore, a comparative analysis on the classification performance of CART and CHAID algorithms demonstrates the superiority of CART with higher overall classification accuracy and lower prediction risk.