High-speed spindle systems operating above 20,000 RPM pose significant design challenges due to strongly coupled dynamic, thermal, and manufacturing constraints. Conventional design approaches either rely on simplified analytical models with limited physical fidelity or computationally expensive numerical simulations that hinder full-system optimization. This paper presents the development, optimization, manufacturing, and experimental validation of a 40,000 RPM electro-spindle using an integrated, multi-stage, multi-objective optimization framework. The proposed methodology combines fast analytical dynamic modeling with reduced-order thermal analysis and a Teaching–Learning-Based Optimization (TLBO) algorithm. Spindle dynamics are modelled using an analytical bearing formulation and a Timoshenko-beam representation of the shaft, solved via Receptance Coupling Substructure Analysis (RCSA) to enable rapid evaluation of stiffness, stability, and critical-speed constraints. Thermal behavior is captured using POD-based thermal modal simulations, significantly reducing computational cost while preserving the dominant thermal characteristics required for optimization. The optimization process is structured into two stages: fundamental optimization, which establishes the shaft–bearing–motor architecture, and optimization for manufacturing, which incorporates preload strategies, lubrication and cooling layouts, sealing concepts, and manufacturability constraints. The optimization strategy guided decisions on bearing selection (71909 CE Front, 71907 CE Rear), bearing locations, dimensions for auxiliary components like sleeves, locknuts, and tail length. Although mid-design stages were presented, the optimization-guided decisions on auxiliary components led to potential improvements of 7% in natural frequency and 11% in mass reduction. The optimized spindle was manufactured and experimentally tested. The measured dynamic (fundamental spindle FRF response in the 1000 Hz range) and thermal responses (thermal deformation of 29 μm) show good agreement with model predictions and design targets, confirming the effectiveness of the proposed framework. The results demonstrate that integrating fast, physically representative models with manufacturability-aware optimization enables efficient and reliable design of high-speed spindles under realistic industrial constraints.
Afridi et al. (Tue,) studied this question.