Comparative Experimental Analysis of ML Based Surrogate Models for Predicting MR and PRR In WEDM of Shape Memory Alloys

DOI

10.2991/978-94-6239-697-5_16How to use a DOI?

Keywords

WEDM; Shape Memory Alloys; Surrogate Modelling; ANFIS; GPR; SVR

Abstract

Wire electro discharge machining i.e. WEDM of shape memory alloys (SMA) involves complex experimental processes, making optimization time-consuming and expensive. To address this, machine learning-based models trained using limited experimental data are employed as surrogate systems to predict machining responses. In this study, ML models most suitable for EDM output prediction, i.e., Adaptive-Neuro-Fuzzy-Inference-System (ANFIS), Support-Vector-Regression (SVR) and Gaussian-Process-Regression (GPR) were comparatively evaluated as surrogate models for predicting product removal rate (PRR) and material roughness (MR) during WED machining of Ti₅₀Ni₄₀Co₁₀ SMA. Experimental data were divided into 70% training and 30% testing sets. Model performances were assessed using square root of mean error i.e. RMSE, R² error or the determination coefficient error, and average percent accuracy. Analysis presents that, among all models, GPR shows meaningful confidence intervals and achieved the highest accuracy (85.78%) for MR prediction due to the smooth behaviour of MR, and ANFIS demonstrated the most superior performance accuracy (65.55%) for PRR prediction despite localized nonlinear relationships and a small experimental dataset. Both SVR and GPR showed limited robustness for PRR prediction due to its highly nonlinear nature. ANFIS avoided overfitting in PRR prediction by reducing fuzzy rules but captured limited details for modelling material roughness variations, leading to lower MR prediction accuracy. The study concludes that ANFIS and GPR are best suited as surrogate modelling frameworks for PRR and MR prediction during WEDM of Ti₅₀Ni₄₀Co₁₀ shape memory alloys. These frameworks can be combined into a hybrid model to make a complete surrogate of the WEDM process of SMAs. The study shows how machine learning–based surrogate models can significantly reduce experimental effort while maintaining reliable prediction accuracy.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Vidhi Bhateja
AU  - Hargovind Soni
PY  - 2026
DA  - 2026/06/04
TI  - Comparative Experimental Analysis of ML Based Surrogate Models for Predicting MR and PRR In WEDM of Shape Memory Alloys
BT  - Proceedings of the Conference on Bridging Engineering Disciplines with AI and Machine Learning (BEDAIML 2026)
PB  - Atlantis Press
SP  - 180
EP  - 194
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-697-5_16
DO  - 10.2991/978-94-6239-697-5_16
ID  - Bhateja2026
ER  -