pith. sign in

arxiv: 2509.01775 · v2 · pith:666QV65Knew · submitted 2025-09-01 · ❄️ cond-mat.mes-hall

Multi-objective Evolutionary Algorithms (MOEAs) in PMEDM -- A Comparative Study in Pareto Frontier

Mohsen Asghari Ilani , Yaser Mike Banad This is my paper

Pith reviewed 2026-05-18 19:22 UTC · model grok-4.3

classification ❄️ cond-mat.mes-hall
keywords powder-mixed EDMmachine learningXGBoostmulti-objective evolutionary algorithmsPareto frontNSGA-IIprecision manufacturingelectrode vibration
0
0 comments X

The pith

XGBoost outperforms other ML models on PMEDM data with powder and vibration features, and MOEAs optimize the resulting Pareto front for manufacturing trade-offs.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper evaluates four machine learning models on datasets from powder-mixed electrical discharge machining that include features from powder addition and electrode vibration. It reports that XGBoost delivers the highest accuracy, followed by AdaBoost, deep neural networks, and ElasticNet, based on accuracy, MSE, RMSE, and MAE. The work then applies four multi-objective evolutionary algorithms—NSGA-II, NSGA-III, UNSGA-III, and C-TAEA—to search for optimal points on the Pareto front. A sympathetic reader would care because these steps aim to make EDM processes more precise, cost-effective, and reliable for high-tolerance parts without direct tool contact.

Core claim

XGBoost achieves superior accuracy on the enriched PMEDM datasets, and the integration of NSGA-II, NSGA-III, UNSGA-III, and C-TAEA optimizes the Pareto front to attain optimal solutions that balance multiple objectives in the machining process.

What carries the argument

The Pareto-front optimization performed by multi-objective evolutionary algorithms (NSGA-II and related methods) applied to predictions from the leading ML model.

If this is right

  • The optimized Pareto front supplies concrete parameter sets that trade off competing goals such as removal rate against surface quality in PMEDM.
  • Superior ML accuracy reduces the need for extensive trial-and-error runs during process setup.
  • The combined ML-plus-MOEA pipeline yields cost-effective and time-efficient operating points for precision manufacturing.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same pipeline could be tested on other non-contact machining methods that involve powder or vibration to check for comparable gains.
  • Real-time sensor feedback loops might close the gap between model predictions and shop-floor variability.
  • Extending the objective set to include tool wear or power consumption would produce more complete trade-off surfaces.

Load-bearing premise

The performance metrics obtained on the collected PMEDM datasets with added powder and vibration features will generalize to unseen real-world machining conditions.

What would settle it

Physical PMEDM experiments that apply the ML-predicted parameters and MOEA-selected solutions from the Pareto front and measure whether they produce statistically better material removal rates, surface finish, or energy use than standard unoptimized settings.

read the original abstract

Electrical discharge machining (EDM) is a crucial process in precision manufacturing, leveraging electro-thermal energy to remove material without electrode contact. In this study, we delve into the realm of Machine Learning (ML) to enhance the efficiency and precision of EDM, particularly focusing on Powder-Mixed Electrical Discharge Machining (PMEDM) with the integration of a vibration system. We comprehensively evaluate four leading ML models - Deep Neural Network (DNN), Extreme Gradient Boosting (XGBoost), Adaptive Gradient Boosting (AdaBoost), and ElasticNet, against a pool of ML models, employing various evaluation metrics including Accuracy, Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). Our evaluations, conducted on datasets enriched with features derived from powder addition and electrode vibration, reveal XGBoost superior accuracy, followed by AdaBoost, DNN, and ElasticNet. Furthermore, through the integration of Multi-Objective Evolutionary Algorithms (MOEAs) such as NSGA-II, NSGA-III, UNSGA-III, and C-TAEA, we explore and optimize the Pareto front to attain optimal solutions. Our findings underscore the transformative potential of ML and optimization techniques in advancing EDM processes, offering cost-effective, time-efficient, and reliable solutions for precision manufacturing applications.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript evaluates four ML models (DNN, XGBoost, AdaBoost, ElasticNet) on PMEDM datasets augmented with powder-addition and electrode-vibration features, claiming XGBoost yields the highest accuracy followed by AdaBoost, DNN and ElasticNet according to Accuracy, MSE, RMSE and MAE. It then applies the MOEAs NSGA-II, NSGA-III, UNSGA-III and C-TAEA to optimize the Pareto front and obtain optimal machining solutions.

Significance. If the metric definitions and numerical results can be clarified and shown to be robust, the work would illustrate a concrete workflow that couples supervised learning with multi-objective evolutionary optimization for a practical manufacturing process. Such an end-to-end demonstration could be useful for process engineers seeking data-driven parameter selection in non-conventional machining.

major comments (2)
  1. [Abstract] Abstract: the simultaneous reporting of Accuracy together with MSE, RMSE and MAE is internally inconsistent for the continuous regression targets that dominate PMEDM (MRR, TWR, surface roughness). Accuracy is appropriate only after explicit discretization or for a classification reformulation; neither is described. This mismatch directly weakens the central claim that XGBoost is superior.
  2. [Evaluation and Results sections] Evaluation and Results sections: no numerical values, dataset cardinality or provenance, train/test split details, cross-validation protocol, baseline comparisons beyond the four listed models, or statistical significance tests are supplied. Without these quantities the reported ranking and the subsequent MOEA Pareto-front claims cannot be verified or reproduced.
minor comments (2)
  1. [Abstract] Clarify whether 'a pool of ML models' refers to additional algorithms beyond the four explicitly named or is simply rhetorical.
  2. [Methods] Standardize algorithm nomenclature (e.g., 'Adaptive Gradient Boosting' for AdaBoost) and supply the precise hyper-parameter search ranges used for each model.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback. We address each major comment point by point below, indicating where revisions will be made to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the simultaneous reporting of Accuracy together with MSE, RMSE and MAE is internally inconsistent for the continuous regression targets that dominate PMEDM (MRR, TWR, surface roughness). Accuracy is appropriate only after explicit discretization or for a classification reformulation; neither is described. This mismatch directly weakens the central claim that XGBoost is superior.

    Authors: We appreciate the referee for identifying this inconsistency. The primary targets (MRR, TWR, surface roughness) are continuous regression variables, for which MSE, RMSE and MAE are the suitable metrics. The appearance of Accuracy in the abstract and results was an inadvertent inclusion and does not reflect any discretization or classification step performed in the study. In the revised manuscript we will remove all references to Accuracy, ensuring that the performance ranking of XGBoost is presented exclusively on the basis of the regression metrics. This change will eliminate the internal inconsistency and reinforce the validity of the model comparison. revision: yes

  2. Referee: [Evaluation and Results sections] Evaluation and Results sections: no numerical values, dataset cardinality or provenance, train/test split details, cross-validation protocol, baseline comparisons beyond the four listed models, or statistical significance tests are supplied. Without these quantities the reported ranking and the subsequent MOEA Pareto-front claims cannot be verified or reproduced.

    Authors: We agree that the submitted manuscript omits several details required for reproducibility. In the revised version we will augment the Evaluation and Results sections with: the exact cardinality and provenance of each PMEDM dataset; the train/test split ratios and cross-validation protocol employed; the complete numerical values of all metrics for the four models; additional baseline comparisons where appropriate; and statistical significance tests (e.g., paired t-tests) supporting the reported ranking. These additions will permit independent verification of both the supervised-learning results and the subsequent MOEA Pareto-front optimizations. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical ML and MOEA comparisons are self-contained

full rationale

The paper reports an empirical study comparing off-the-shelf ML models (DNN, XGBoost, AdaBoost, ElasticNet) and MOEAs (NSGA-II, NSGA-III, UNSGA-III, C-TAEA) on PMEDM datasets with added powder/vibration features. Performance is assessed via standard metrics (Accuracy, MSE, RMSE, MAE) and Pareto-front optimization. No equations, derivations, fitted-parameter predictions, self-citations, or ansatzes appear in the abstract or described content. Claims rest on direct experimental evaluation rather than any reduction of outputs to inputs by construction, satisfying the default expectation of no significant circularity.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 0 invented entities

The central claims rest on the representativeness of the collected PMEDM datasets and on the assumption that standard ML training procedures produce generalizable predictors for electro-thermal removal processes.

free parameters (1)
  • ML model hyperparameters
    Typical hyperparameters for DNN, XGBoost, AdaBoost, and ElasticNet are tuned on the data but not reported in the abstract.
axioms (1)
  • domain assumption The enriched feature set derived from powder addition and electrode vibration sufficiently captures the relevant process physics for accurate prediction.
    Invoked when the authors state that evaluations on these datasets reveal model performance differences.

pith-pipeline@v0.9.0 · 5771 in / 1343 out tokens · 62060 ms · 2026-05-18T19:22:39.634450+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

15 extracted references · 15 canonical work pages

  1. [1]

    Introduction Electro-Discharge Machining (EDM) has emerged as an advanced manufacturing technique designed to address shortcomings of conventional methods by accommodating complex geometries and high material hardness, while simultaneously reducing material consumption and production lead times. Its ability to enhance performance, reliability, and customi...

    work page 2009

  2. [2]

    It comprises several components, including raw dataset features, featurization processes, the employed ML models and MOEs, Pareto frontier analysis, and targe t regression

    Methodology Figure 1 showcases the utilization of both ML models and Multi -objective Evolutionary Algorithms (MOEAs). It comprises several components, including raw dataset features, featurization processes, the employed ML models and MOEs, Pareto frontier analysis, and targe t regression. This section delves into the processes of dataset collection and ...

    work page

  3. [3]

    Powder concentration (C)

    work page

  4. [4]

    Injection pressure (P)

    work page

  5. [5]

    Vibration frequency (F)

    work page

  6. [6]

    We employed different ML regression models to accurately predict MRR, EWR, and Ra through PMEDM with the electrode vibration system

    Amplitude of vibration (A) Inputs OutputsM Models M egression egression a egression MOEAs A II A III A A properties of these outputs. We employed different ML regression models to accurately predict MRR, EWR, and Ra through PMEDM with the electrode vibration system. Among the 13 models used were conventional ML models such as 'XGBoost', 'LGBM', 'AdaBoost'...

    work page

  7. [7]

    It examines the impact of adding powders and adjusting parameters of the vibration system to enhance preprocessing and cleaning datasets, thus refining features for ML model input. The addition of powders, ranging in concentration from 0 to 5 g/l (Figure 4a, b, c), demonstrates an increase in MRR and a decrease in both EWR and Ra compared to scenarios whe...

    work page

  8. [8]

    Initially, we undertake a regression task employing the four ML models and evaluate them using commonly adopted metrics: MAE, MSE, RMSE, and R -squared

    Results and discussion In this section, we assess the performance of four ML models on datasets that have not received extensive prior study, aiming to determine the most effective outputs for EDM, considering parameters such as MRR, EWR, and Ra, alongside variables like powder composition and electrode vibration. Initially, we undertake a regression task...

    work page

  9. [9]

    This suggests that UNSGA -III may offer comparable performance to these algorithms in terms of exploring and optimizing the Pareto front

    and ElasticNet (Figure 20) under UNSGA-III, [2.27933009, 0.36787944, 0.45151965] and [1.17967761, 0.48496365, 0.71883236], respectively, show similarities with the solutions obtained under NSGA -II and NSGA -III. This suggests that UNSGA -III may offer comparable performance to these algorithms in terms of exploring and optimizing the Pareto front. Overal...

    work page

  10. [10]

    This suggests that C-TAEA may offer comparable performance to other algorithms in exploring and optimizing the Pareto front

    and ElasticNet (Figure 24) under C -TAEA, [2.27933009, 0.36787944, 0.45151965] and [1.17329555, 0.48154035, 0.72506959], respectively, show similarities with the solutions obtained under other MOEAs. This suggests that C-TAEA may offer comparable performance to other algorithms in exploring and optimizing the Pareto front. Overall, comparing the performan...

    work page

  11. [11]

    Conclusion In conclusion, our study underscores the pivotal role of machine learning in enhancing the efficiency and precision of Electrical Discharge Machining (EDM), a complex process vital for precision manufacturing applications. By leveraging ML techniques, spec ifically evaluating four top -performing models - DNN, XGBoost, AdaBoost, and ElasticNet ...

    work page doi:10.1080/10426914.2020.1854 2021

  12. [12]

    A., & Khoshnevisan, M

    https://doi.org/10.1007/S00170-021- 07569-3/FIGURES/30 Ilani, M. A., & Khoshnevisan, M. (2022). An evaluation of the surface integrity and corrosion behavior of Ti-6Al-4 V processed thermodynamically by PM-EDM criteria. International Journal of Advanced Manufacturing Technology, 120(7–8), 5117–

    work page doi:10.1007/s00170-021- 2022

  13. [13]

    N., & Pande, S

    https://doi.org/10.1007/S00170-022- 09093-4/FIGURES/18 Joshi, S. N., & Pande, S. S. (2009). Development of an intelligent process model for EDM. International Journal of Advanced Manufacturing Technology, 45(3–4), 300–

    work page doi:10.1007/s00170-022- 2009

  14. [14]

    https://doi.org/10.1007/S00170-009- 1972-4/METRICS Mert Doleker, K., Odabas, O., Ozgurluk, Y ., -, al, Ece Sayn, F., Topalolu, G., Sharma, V ., Prakash Misra, J., & Singhal, P. (2019). Multi-Optimization of Process parameters for Inconel 718 While Die-Sink EDM Using Multi-Criterion Decision Making Methods. Journal of Physics: Conference Series, 1240(1), 0...

    work page doi:10.1007/s00170-009- 1972

  15. [15]

    K., & Gupta, R

    https://doi.org/10.1108/IJSI-09-2021- 0101/FULL/XML Sharma, N., Khanna, R., Sharma, Y . K., & Gupta, R. D. (2019). Multi-quality characteristics optimisation on wedm for ti-6al-4v using taguchi-grey relational theory. International Journal of Machining and Machinability of Materials, 21(1–2), 66–81. https://doi.org/10.1504/IJMMM.2019.09806 7 Sharma, P., C...

    work page doi:10.1108/ijsi-09-2021- 2021