Design and FEA of Composite Spur Gears: A Comparative Study Using ANFIS, XGBoost, and Variational Quantum Regression

Authors

• P. H. J. Venkatesh, Assistant Professor, Department of Mechanical Engineering, Vignan's Institute of Information Technology (A), Visakhapatnam, Andhra Pradesh, IndiaFollow
• Pothala Raghu, Assistant Professor, Department of Mechanical Engineering, Vignan's Institute of Engineering for Women (A), Kapujaggaraju Peta, Behind VSEZ, Visakhapatnam, Andhra Pradesh, IndiaFollow
• Mohd Rafeeq Ur Rahman, Assistant Professor, Department of Mechanical Engineering, J.B. Institute of Engineering and Technology (A), Hyderabad, Telangana 500075, IndiaFollow
• Saikumar Amda, Assistant Professor, Department of Mechanical Engineering, J.B. Institute of Engineering and Technology (A), Hyderabad, Telangana 500075, IndiaFollow
• T. S. Vamsi Krishna, Assistant Professor, Department of Mechanical engineering GVP college of engineering (A) VisakhapatnamFollow

Corresponding Author

P. H. J. Venkatesh

Authors ORCID

P. H. J. Venkatesh: https://orcid.org/0000-0003-0030-6790

Document Type

Article

Keywords

Spur gear, Metal matrix composite, Finite element analysis, ANFIS, XGBoost, Variational quantum regression, Surrogate modelling, SAE 8620

Abstract

The parametric FEA dataset for an automotive involute spur-gear pair in three materials, train an ANFIS, XG Boost, and four-qubit Variational Quantum Regression (VQR) surrogate on that dataset, and compare their accuracy, uncertainty, and computational cost against the FEA ground truth. The study also ranks the three materials using an auditable multi-criteria decision procedure. The gear pair was sized using the Lewis equation, modelled in CATIA V5, and solved in ANSYS Workbench for SAE 8620 low-alloy steel, a Steel Matrix Composite (SMC), and a Low-Cost Iron Metal-Matrix Composite (Fe-MCS). The nominal duty is 350 N-m of wheel torque, derived from a 140-kW engine at 6,000 rpm, and transmitted through a 2-module, 3 mm transmission-ratio gear pair. A 240-case Latin-Hypercube dataset of equivalent (von-Mises) stress, maximum principal stress, total deformation, and maximum shear stress was used for training, and the three surrogates were compared with the FEA ground truth using the R², RMSE, and MAPE metrics. XG Boost achieved the highest accuracy (R² = 0.998, MAPE = 0.45%), whereas ANFIS demonstrated high rule-based interpretability (R² = 0.991, MAPE = 1.12%). The VQR, a four-qubit parameterized quantum circuit, remained within engineering tolerance (R² = 0.965, MAPE = 3.25%) but was limited by the bounded observable of the Noisy-Intermediate-Scale-Quantum (NISQ) computation. All three surrogates reduced the single-query inference time from 45 minutes (FEA) to milliseconds. SMC exhibited the lowest equivalent stress (5,183 MPa, 19% less than that of Fe-MCS) and the greatest anticipated safety factor, whereas Fe-MCS was the least expensive but exhibited the poorest mechanical performance. The proposed AI-ML-QC framework is among the first to compare a quantum surrogate with its classical counterparts on a single composite-gear design benchmark under identical training, testing and statistical conditions, and it provides a transferable methodology for the accelerated data-driven optimization of other mechanical components.

How to Cite This Article

Venkatesh, P. H. J.; Raghu, Pothala; Rahman, Mohd Rafeeq Ur; Amda, Saikumar; and Krishna, T. S. Vamsi (2026) "Design and FEA of Composite Spur Gears: A Comparative Study Using ANFIS, XGBoost, and Variational Quantum Regression," Babylonian Journal of Mechanical Engineering: Vol. 4: Iss. 1, Article 5.