⚡ electrical engineering

Physics-Guided Predictive Modeling of Tool Wear in Electrical Discharge Machining via Dimensional Analysis: A Buckingham Π Framework with TOPSIS-Based Electrode Selection

This paper presents a physics-guided predictive model for tool wear in electrical discharge machining by deriving a closed-form power-law equation via the Buckingham Π theorem, validating it with experimental data, and utilizing its sensitivity exponents to analytically rank electrode materials using the TOPSIS method.

Original authors: Rakesh Nath, Vinay Sharma, SOMAK DATTA

Published 2026-09-17

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Rakesh Nath, Vinay Sharma, SOMAK DATTA

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the world of heavy industry, where aerospace parts, medical implants, and automotive components are forged from materials too hard for traditional cutting tools, a different kind of machining takes over. This process, known as electrical discharge machining, does not use a sharp blade to slice away metal. Instead, it uses a shaped tool and a workpiece submerged in a special fluid, separated by a tiny gap. When a high voltage is applied, electricity jumps across this gap in the form of a spark. These sparks are incredibly intense, creating a localized plasma channel that reaches temperatures between 8,000 and 12,000 degrees Celsius. This heat is so extreme that it instantly melts and vaporizes tiny craters of the workpiece, allowing engineers to carve out complex shapes with precision. However, this violent process comes with a cost. The tool itself, usually a block of metal or graphite, is not immune to the heat. It slowly erodes, losing material with every spark. This wear changes the tool's shape, ruining the accuracy of the part being made and forcing frequent, expensive replacements. For decades, engineers have struggled to predict exactly how fast a tool will wear out, often relying on trial and error or complex computer simulations that lack a clear physical explanation.

A team of researchers from the Birla Institute of Technology in India has approached this problem by stripping it back to its fundamental physical laws. Rather than trying to fit a curve to a pile of experimental data, they used a classic method of physics called dimensional analysis to understand the relationship between the variables at play. They identified eight key factors that control how quickly a tool wears down: the strength of the electric current, the voltage across the gap, the duration of each spark, the frequency of the sparks, and three specific properties of the tool material itself—its density, its melting point, and its ability to conduct heat. By grouping these variables into three dimensionless numbers, the researchers created a simplified model that describes the physics of the wear process without needing to know every microscopic detail of the spark. This approach allowed them to derive a clear, closed-form equation that predicts the rate of tool wear based on the operating conditions and the material chosen.

The team tested this new model using a set of nine controlled experiments where a copper tool was used to machine mild steel. They adjusted the current, voltage, and spark duration across a specific range and measured the actual wear. When they compared their model's predictions to the real-world measurements, the results were strikingly close. The model yielded an R² value of 0.969 and a mean absolute percentage error (MAPE) of 10.1%. This level of accuracy is significant because it proves that the underlying physics of the process can be captured in a simple, understandable formula. The model revealed that the tool's ability to dissipate heat and its resistance to melting are the most critical factors in slowing down wear. Specifically, the analysis showed that doubling the thermal conductivity or the melting point of the tool material would nearly halve the rate at which the tool erodes. This finding provides a clear, physical reason why certain materials perform better than others, moving beyond guesswork to a science-based understanding.

With a reliable model in hand, the researchers turned to a practical question that plagues every machine shop: which tool material should be chosen for a specific job? Traditionally, selecting the best electrode involves weighing conflicting factors like cost, durability, and electrical properties, often relying on the subjective judgment of an expert. The researchers replaced this subjectivity with a rigorous, physics-based decision-making process. They used the sensitivity of their wear model to assign weights to different material properties. Because the model showed that thermal conductivity and melting point have the strongest influence on wear, these two properties were given the highest importance in the selection process. They then applied a mathematical ranking method to eleven common electrode materials, including copper, tungsten, silver, and various alloys. The result was a definitive ranking that placed a copper-tungsten alloy at the top, followed closely by pure tungsten and a silver-tungsten mix. These materials, known for their ability to withstand extreme heat and conduct it away quickly, emerged as the clear winners, while softer, cheaper materials like brass ranked last. This method removes the guesswork, ensuring that the choice of tool is driven by the actual physics of the machining process rather than intuition.

To ensure their findings were not just a lucky match for their specific experiments, the team tested their model against a completely separate set of data from a different study, which used a copper tool on a different type of steel with different machine settings. While the model's accuracy dipped slightly when predicting outside the range of its original training data, it still captured the general trend of the wear behavior. The few instances where the model missed the mark occurred when the machine was run at conditions far more extreme than those used in the initial tests, such as very short spark durations or very high currents. This behavior is expected, as the model is built on the physics observed within a specific operating window. Nevertheless, the fact that the model could predict wear on a different machine with a different material setup without needing to be re-tuned demonstrates its robustness. The researchers suggest that this framework could serve as the core of a "digital twin," a virtual replica of the machining process that updates in real-time. By feeding live data from the machine into the model, operators could monitor tool wear as it happens and predict exactly when a tool needs to be replaced, preventing costly errors and downtime.

The work presented here offers a fresh perspective on an old industrial challenge. By returning to first principles and using the timeless logic of dimensional analysis, the researchers have created a tool that is both simple and powerful. They have shown that the complex, chaotic process of electrical discharge machining can be understood through a few key physical relationships. The resulting model does not just predict numbers; it explains why certain materials last longer and provides a clear, objective path for selecting the right tool for the job. While the model has limits when pushed beyond its tested range, its ability to generalize across different materials and conditions suggests a promising future for more efficient, predictable, and cost-effective manufacturing. The path forward involves expanding the data to cover even more extreme conditions and integrating this physics-based model into real-time monitoring systems, turning a theoretical insight into a practical tool for the factory floor.

Technical Summary: Physics-Guided Predictive Modeling of Tool Wear in EDM

Problem Statement
Electrical Discharge Machining (EDM) is a critical non-traditional process for shaping hard, electrically conductive materials in aerospace, automotive, and biomedical sectors. However, electrode wear remains a progressive phenomenon driven by localized plasma temperatures of 8,000–12,000 °C. This wear degrades dimensional accuracy, compromises surface integrity, and increases production costs due to frequent electrode replacement. While existing literature offers empirical regression models, machine learning approaches, and analytical thermal models, there is a lack of quantitative, physics-based models for Tool Wear Rate (TWR) that are both closed-form and physically interpretable. Furthermore, electrode selection often relies on subjective expert weighting in Multi-Criteria Decision Making (MCDM) frameworks.

Methodology
The study employs a dimensional analysis framework based on the Buckingham Π theorem to derive a physics-guided predictive model for TWR in die-sinking EDM.

Dimensional Analysis: The authors identified eight governing variables (TWR, discharge current $I$ $I$ , gap voltage $V$ $V$ , pulse-on time $t_{on}$ $t_{o n}$ , spark frequency $f$ $f$ , electrode density $\rho$ $ρ$ , melting temperature $T_m$ $T_{m}$ , and thermal conductivity $\kappa$ $κ$ ) and five fundamental dimensions ( $M, L, T, A, \Theta$ $M, L, T, A, Θ$ ). This reduced the system to three independent dimensionless groups ( $\Pi_1, \Pi_2, \Pi_3$ $Π_{1}, Π_{2}, Π_{3}$ ):
- $\Pi_1$ : A dimensionless TWR group.
- $\Pi_2$ : A pulse duty ratio group ( $f \times t_{on}$ ).
- $\Pi_3$ : A thermal material group incorporating $\kappa$ , $T_m$ , and $\rho$ .
Model Calibration: A closed-form power-law model ( $\Pi_1 = C \cdot \Pi_2^\alpha \cdot \Pi_3^\beta$ ) was derived. The model was calibrated using a nine-run Taguchi L9 experimental dataset involving a copper electrode on mild steel ( $I=2\text{--}6$ A, $V=40\text{--}80$ V, $t_{on}=50\text{--}200$ $\mu$ s). Log-linear ordinary least-squares regression was used to determine the coefficients.
TOPSIS Integration: The log-log sensitivity exponents derived from the calibrated model were used to analytically determine weight factors for a TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) framework. This replaced subjective expert assignment with physics-grounded weights for electrode material selection.
Validation: The model was validated against an independent dataset from Bose & Mahapatra (2015) involving a copper electrode on AISI H-13 tool steel, covering operating conditions outside the calibration range.
Digital Twin Proposal: A conceptual architecture was proposed combining the calibrated physics model with Bayesian adaptive recalibration for real-time wear monitoring.

Key Results

Model Performance: The calibrated model achieved a coefficient of determination ( $R^2$ ) of 0.969 and a Mean Absolute Percentage Error (MAPE) of 10.1% on the calibration dataset.
Physical Interpretability: The derived thermal exponent $\beta = -0.990$ aligns with EDM erosion theory, confirming that higher electrode thermal conductivity and melting points significantly reduce TWR. The duty ratio exponent $\alpha = -0.119$ indicates that higher duty ratios lower TWR by spreading energy across more pulses.
Sensitivity Analysis: The model identified discharge current ( $I$ ) and gap voltage ( $V$ ) as having the strongest positive sensitivity (exponent $\approx +1.392$ ), while thermal conductivity and melting temperature showed strong negative sensitivity ( $\approx -0.990$ ).
Electrode Selection: Using the physics-derived weights, the TOPSIS analysis ranked Copper-Tungsten (CC = 0.611), Tungsten (CC = 0.606), and Silver-Tungsten (CC = 0.569) as the top three materials. Copper ranked fifth (CC = 0.504) due to its lower melting point, despite high thermal conductivity.
Validation Limits: Independent validation yielded a MAPE of 18.5% with an $R^2$ of 0.367. Four outliers occurred under conditions outside the calibration range (specifically $t_{on} < 50$ $\mu$ s and $I > 6$ A), highlighting the known limitations of extrapolating power-law models.

Significance and Contributions
The paper claims four primary contributions:

Closed-Form Physics Model: It provides a physically interpretable, closed-form model for EDM tool wear where all exponents are derived from first-principles dimensional analysis rather than purely empirical fitting.
Thermal Exponent Validation: The model yields a physically consistent thermal exponent ( $\beta = -0.990$ ), quantitatively demonstrating the protective role of thermal conductivity and melting point, a capability often missing in data-driven methods.
Objective MCDM: It introduces a reproducible, physics-grounded procedure for TOPSIS electrode selection, eliminating the subjectivity inherent in expert-assigned weights by deriving them directly from model sensitivity exponents.
Transparency: The study offers full openness, reporting the complete calibration dataset, model parameters, and prediction errors to facilitate reproducibility.

The authors conclude that while the model captures the main TWR trends, its quantitative accuracy decreases when extrapolated beyond the calibration domain. They propose the framework as a guide for future experimental work and as a foundational engine for digital twin architectures in EDM, pending full industrial validation with real-time sensor integration.

Technical Summary: Physics-Guided Predictive Modeling of Tool Wear in EDM

More like this