A Review on Multi-objective Optimization Techniques of Wire Electrical Discharge Machining

In the present-day manufacturing environment, wire electrical discharge machining (WEDM) has become one of the most efficient non-conventional material removal processes to generate complicated 2D and 3D profiles on many of the difficult-to-cut engineering materials. Although the material removal rate of this process is comparatively low, but it can provide high dimensional accuracy and tolerance along with excellent surface integrity. To explore its maximum potential, it is advised to operate this process at the optimal combination of its various input parameters, which can only be derived using some optimization tools. The past researchers have already applied several multi-objective optimization techniques to resolve the issue. This paper comprehensively reviews and documents applications of four major multi-objective optimization tools, i.e. desirability function approach, grey relational analysis (GRA), multi-criteria decision making methods and metaheuristic algorithms considered for parametric optimization of WEDM processes. It also extracts information regarding type of the experimental design plan, work and wire materials, dielectric utilized, and WEDM parameters and responses considered. It is observed that Taguchi's L27 orthogonal array has been the most commonly deployed design plan, while medium and high carbon steels, and brass have been the most prevalent work and wire materials, respectively. Most of the researchers have preferred deionized water as the dielectric and GRA as the multi-objective optimization technique. During WEDM experiments, pulse-on time and pulse-off time have appeared as the two most significant input parameters; and surface roughness has been the most important response, followed by material removal rate. The outcome of this review paper would help the future researchers to have an idea regarding initial settings of different WEDM parameters and achievable response values. It would also act as a data support for subsequent utilization in developing machine learning-based prediction models.