Genetic programming approach to predict the performance characteristics of WEDM taper cutting process

Abstract

One of the unique applications of the wire electrical discharge machining (WEDM) method is taper cutting process. This process is ideal for numerous intricate, difficult to machine materials with complex profiles, deep slots with tight corners and features at different angles, those are basically used in defense and aerospace applications. According to the literature, numerous mathematical models based on physics and data may be created to identify the optimum parameter settings throughout the taper cutting process. However, it is a challenging process since it necessitates extensive understanding of how to do tapering operations with WEDM. As a result, a genetic programming (GP) model is created in this study to anticipate the angle inaccuracy during WEDM taper cutting. Six process parameters including workpiece thickness, taper angle, pulse duration, discharge current, wire speed and wire tension were considered at three level each for minimizing the angular error and surface roughness during WEDM taper cutting process. The L₂₇ orthogonal array of Taguchi’s design of experiment is utilized to collect data on the procedure. Three assessing criteria are used to assess the correctness of the suggested model: root mean square error (RSME), mean absolute percentage error (MAPE) and co-efficient of determination (R²). In compared to typical prediction approaches, the results suggest that MGGP is a successful strategy. The model is cost effective and time saving way to investigate angular error and surface roughness in taper cutting before performing the actual machining process.

Introduction

Many complicated manufacturing processes exist in today’s industrial environment where the cost and quality are the main driving forces. Among them, taper cutting utilizing wire electrical discharge machining process is a sophisticated engineering process. In this process the wire is deformed at the time of angular cutting, resulting variations in the actual angle of machined items [1]. As a result, the accuracy of the machined component is compromised. Hence, the realm of taper cutting operation in WEDM, effective process parameter selection is critical challenge. However experimental process optimization of any machining process is expensive, time consuming and demands high degree of technical expertise owing to the complex, nonlinear nature of input–output variables of the process. In the other way we may acquire output values for preset input parameter values either empirically, mathematically, or numerically. However, employing mathematical and numerical models to map the behavior and interrelation of process factors is also difficult. Because this procedure is time intensive and demands a high degree of technical competence, solving such numerical and mathematical models is not always a straightforward task. Hence, statistical models are unable to forecast performance measurements due to the multiple factors involved and the non-linearity of the connection between the parameters. Only when the number of variables is minimal and their effects can be characterized by simple connections will statistical approaches be appropriate [2]. To overcome this constraint, multiple academics have suggested several artificial intelligence strategies for correlating input parameters with WEDM performance measures. [3], [4], [5]. Saha et al. proposed a multi variable regression model and a feed- forward neural network model for WEDM process to predict the performance measures such as cutting speed and surface roughness considering pulse on time, pulse off time, peak current and capacitance as input variables [6]. Aich and Banarjee used support vector machine regression to create a process model for the EDM process and used particle swarm optimization to find the optimal parametric combination [7]. Conde et al. proposed a way to predict the accuracy of components produced by WEMD by using an Elman-based Layer Recurrent Neural Network (LRNN). They have also suggested an algorithm for designing wire paths of variable radius to correct the deviations in the machined part through software [8]. Majumdar and Maity [9] predict and optimize the WEDM performance measures such as surface roughness and micro hardness using general regression neural network (GRNN) and MOORA fuzzy approach considering nitinol as workpiece material. Caydas et al. developed an adaptive neuro-fuzzy inference system (ANFIS) model for the white layer thickness (WLT) and the average surface roughness of WEDM process considering pulse duration, open circuit voltage, dielectric flushing pressure and wire feed rate as model’s input feature [10]. Nayak and Mahapatra have developed a support vector regression (SVR) model to predict the angular error in taper cutting using wire electrical discharge machining process [11]. However, the use of artificial intelligent techniques in the field of taper cutting employing the WEDM is quite limited. As a result, geometric programming (GP) model is proposed in this paper to predict angular error and surface roughness during taper cutting in WEDM process. The root meat square error (RMSE), mean absolute percentage error (MAPE) and co-efficient of determination (R²) are used to asses the model’s performance.