Process parametric and performance characteristics study of WED machined Ni–Ti–Hf high-temperature shape memory alloys: an experimental and artificial intelligence approach

Abstract In recent years, to meet the shortcomings of the conventional machining of Ni–Ti–Hf shape memory alloy (SMA), Wire Electric Discharge Machining (WEDM) as one of the unconventional machining methods, has emerged as the preferred method for processing SMAs. Therefore, in this study, according to the Response Surface Methodology-based Central Composite Design layout, WED machining of Ni–Ti–Hf SMA is carried out using the control parameters like spark time ( S ON ), spark pause time ( S OFF ), gap voltage ( V g ), and dielectric flow rate ( F DL ). A General Regression Neural Network (GRNN) model was used to predict the critical WEDM responses: material removal rate (MRR), surface roughness (Ra), and kerf width (KW). The GRNN model closely agrees with the experimental WEDM responses, resulting in a Mean Absolute Percentage Error below 2%. Field Emission Scanning Electron Microscopy result revealed a recast layer thickness of 10.64 <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" overflow="scroll"> <mml:mrow> <mml:mo>±</mml:mo> </mml:mrow> </mml:math> 2.06 µ m and 38.19 <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" overflow="scroll"> <mml:mrow> <mml:mo>±</mml:mo> </mml:mrow> </mml:math> 9.55 µ m for the samples with the lowest surface roughness (Ra) and highest MRR, respectively. The shape recovery test result shows a less than 4% reduction in recovery ratio post-WEDM. Further, electrochemical corrosion studies revealed that owing to these surface defects, the corrosion rate increased with higher discharge energy. The corrosion rate of the base material, low Ra sample, and high MRR sample were 0.0013 mm yr −1 , 0.0128 mm yr −1 , and 0.0334 mm yr −1 , respectively.