Optimizing gear NVH performance using genetic algorithm-based on tooth profile modification and a comparative analysis of simulation and experimental outcomes

Gear systems often suffer from excessive Noise, Vibration, and Harshness (NVH), which negatively affect performance, durability, and user comfort. Existing optimization approaches, particularly those based only on Static Transmission Error (STE), are limited because they rely solely on computational predictions without rigorous experimental validation. This study presents a sensor-based experimental validation framework of a Genetic Algorithm (GA)-based Tooth Profile Modification (TPM) optimization for reducing gear system Noise, Vibration, and Harshness (NVH). The GA optimizes the Dynamic Load Factor (DLF) by considering parameters such as Dynamic Transmission Error (DTE) and Static Transmission Error (STE), gear mesh stiffness, meshing load forces, dynamic tooth backlash, and damping forces. The optimized TPM design is validated through a metrologically traceable measurement methodology using a tri-axial accelerometer and Class 1 free-field microphone, deployed according to ISO 10816–3 and ISO 3746, to capture vibration acceleration, velocity, displacement, and airborne noise from a spur gear test rig. Calibration verification, repeatability testing, and spectral analysis of RMS values were performed, with measurement accuracy and reproducibility quantified through repeated trials. Uncertainty evaluation followed the Guide to the expression of Uncertainty in Measurement (GUM). Results show a statistically significant reduction in RMS vibration acceleration, velocity, displacement, and airborne noise compared to the unmodified gear pair, both exceeding expanded uncertainty bounds. The proposed framework bridges gear design optimization and measurement science, supporting traceable, reproducible NVH performance assessment. In contrast to prior studies that apply GA solely based on STE as a computational design tool, The novelty of this work lies in integrating a GA based on DLF optimization with a fully traceable, uncertainty-quantified experimental validation framework, ensuring that predicted improvements are verified in compliance with Vocabulary International Metrology (VIM) and GUM standards.