.png){kind=icon}
.png){kind=icon}
Return
Machine learning based design optimization method for wave generators of harmonic drives considering multiple performance indicators
.png){kind=icon}
PRE
AI
DOI：10.1007/s12206-025-1238-6.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In the harmonic drive (HD) system, the wave generator (WG) plays a crucial role in its function in the aspect of lifespan, deformation in flexspline (FS), and transmission accuracy. However, conventional elliptical WG profiles remain dominant, leading to restricted functionality. This paper presents a machine learning (ML) based optimization framework for WG profiles, using support function. Three performance indicators-FS stress, engagement ratio, and WG curvature-are evaluated. Finite element model (FEM) is built to simulate stress distribution and deformation patterns. Surrogate models using different ML methods trained on 331 datasets overcome computational bottlenecks, with the XGBoost network showing the highest accuracy. Genetic algorithm optimization reduces stress by 21.70 %, boosts engagement by 12.42 %, and lowers curvature by 62.49 % versus traditional elliptical profiles. Tooth profile modification mitigates interference. The scalable framework benefits precision transmission systems.
Keywords:
Harmonic drive
Machine learning
Optimization design
Performance evaluation
Wave generator
Journal
IF:
1.7
Papers:
601
Citations:
1.2W