An Intelligent Approach for Predicting Precision Machining of Wire EDM Process Using Machine Learning

Electrical Discharge Machining (EDM) is widely used for precision machining of intricate parts in many industrial applications. The process output and machining performance of the EDM process are highly affected by the complexity of the EDM process and nonlinear effects. This paper presents a comparative, machine-learning-based study of Wire EDM parameter optimization for AISI D2 tool steel. Four progressively developed frameworks are evaluated: baseline data-driven, realistic data-driven, physics-constrained, and hybrid physics-informed optimization (the last two use physically motivated bounds and penalties on the predicted responses rather than first-principles simulation). The optimization targets simultaneous minimization of power consumption, surface roughness (Ra), tool wear rate (TWR, % wire mass loss), and kerf width while, in the two physics-informed frameworks, maximizing material removal rate (MRR) for 8 mm-thick workpieces. Process parameters investigated include pulse-on time (ON: 2–14 μs), pulse-off time (OFF: 10–16 μs), servo voltage (SV: 35–50 V), and wire tension (WT: 9–12 kg). All models were built from 16 experimental runs; because no independent test set exists, surrogate accuracy was assessed by leave-one-out cross-validation (LOOCV), which gave R2 of 0.74–0.98 for random-forest and quadratic surrogates and 0.25–0.63 for the three-nearest-neighbor model used in the dashboard. The reported optimization outcomes are model predictions and have not yet been confirmed experimentally. The baseline framework produced numerically large but physically inconsistent improvements; the realistic framework remained within the experimental range; the physics-constrained framework achieved a composite optimization score of up to 38.45 with all predictions inside the imposed bounds; and the hybrid framework, using confidence-weighted surrogates, predicted an MRR increase of about 183% (0.78 to 2.22 mm3/min) at ON = 2 μs, OFF = 16 μs, SV = 50 V, and WT = 12 kg, accompanied by higher power, TWR, Ra, and kerf. An intermediate confidence level (0.7) gave the most balanced Pareto trade-off. Within the measured data, low pulse-on times (2–6 μs) gave the lowest roughness (Ra 1.37–2.59 μm), which is relevant to finish machining. An interactive EDM dashboard was developed that combines real-time data-driven prediction with feasibility checks restricted to the observed data domain to support instant, interpretable process exploration. This comparative study offers manufacturers a framework for exploring EDM parameter trade-offs; experimental confirmation of the predicted optima is required before industrial adoption.

Authors

Institutions

Publication Details

Journal
Journal of Manufacturing and Materials Processing
Published
2026-11-12
DOI
https://doi.org/10.3390/jmmp10110505
Primary Topic
Advanced Machining and Optimization Techniques
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

An Intelligent Approach for Predicting Precision Machining of Wire EDM Process Using Machine Learning

Mohammad Younas, Shahana Bano, Taimoor Asim, Priyanshi Kumari et al.
Journal of Manufacturing and Materials Processing
Advanced Machining and Optimization Techniques
article

An Intelligent Approach for Predicting Precision Machining of Wire EDM Process Using Machine Learning

Mohammad Younas, Shahana Bano, Taimoor Asim, Priyanshi Kumari, Priyanshu Mishra
article en

Abstract

Electrical Discharge Machining (EDM) is widely used for precision machining of intricate parts in many industrial applications. The process output and machining performance of the EDM process are highly affected by the complexity of the EDM process and nonlinear effects. This paper presents a comparative, machine-learning-based study of Wire EDM parameter optimization for AISI D2 tool steel. Four progressively developed frameworks are evaluated: baseline data-driven, realistic data-driven, physics-constrained, and hybrid physics-informed optimization (the last two use physically motivated bounds and penalties on the predicted responses rather than first-principles simulation). The optimization targets simultaneous minimization of power consumption, surface roughness (Ra), tool wear rate (TWR, % wire mass loss), and kerf width while, in the two physics-informed frameworks, maximizing material removal rate (MRR) for 8 mm-thick workpieces. Process parameters investigated include pulse-on time (ON: 2–14 μs), pulse-off time (OFF: 10–16 μs), servo voltage (SV: 35–50 V), and wire tension (WT: 9–12 kg). All models were built from 16 experimental runs; because no independent test set exists, surrogate accuracy was assessed by leave-one-out cross-validation (LOOCV), which gave R2 of 0.74–0.98 for random-forest and quadratic surrogates and 0.25–0.63 for the three-nearest-neighbor model used in the dashboard. The reported optimization outcomes are model predictions and have not yet been confirmed experimentally. The baseline framework produced numerically large but physically inconsistent improvements; the realistic framework remained within the experimental range; the physics-constrained framework achieved a composite optimization score of up to 38.45 with all predictions inside the imposed bounds; and the hybrid framework, using confidence-weighted surrogates, predicted an MRR increase of about 183% (0.78 to 2.22 mm3/min) at ON = 2 μs, OFF = 16 μs, SV = 50 V, and WT = 12 kg, accompanied by higher power, TWR, Ra, and kerf. An intermediate confidence level (0.7) gave the most balanced Pareto trade-off. Within the measured data, low pulse-on times (2–6 μs) gave the lowest roughness (Ra 1.37–2.59 μm), which is relevant to finish machining. An interactive EDM dashboard was developed that combines real-time data-driven prediction with feasibility checks restricted to the observed data domain to support instant, interpretable process exploration. This comparative study offers manufacturers a framework for exploring EDM parameter trade-offs; experimental confirmation of the predicted optima is required before industrial adoption.

Journal of Manufacturing and Materials ProcessingVol. 10(11)
Indian Institute of Technology Kharagpur (IN), Robert Gordon University (GB), KIIT University (IN)
Advanced Machining and Optimization Techniques
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon