Stiffness optimization of electric spindle performance based on multi-layer perceptron integrated Bayesian.

Area of Science:

• Mechanical Engineering
• Manufacturing Technology
• Computational Engineering

Background:

• Electric spindle performance optimization is crucial for precision machining.
• Traditional methods face challenges with high-dimensional parameter spaces and nonlinear dynamics.
• Existing approaches may lack efficiency and accuracy in complex optimization tasks.

Purpose of the Study:

• To propose a novel hybrid approach for electric spindle performance optimization.
• To model complex relationships between design parameters and performance metrics.
• To efficiently identify optimal electric spindle configurations for enhanced accuracy and efficiency.

Main Methods:

• Integration of Multi-Layer Perceptron (MLP) for modeling nonlinear relationships.
• Application of Bayesian Optimization (BO) for efficient design space navigation.
• Utilizing ANSYS for finite element simulation of electric spindle stress and deformation.

Main Results:

• Identified key design parameters influencing spindle stress-deformation: overhang length, support span, and bearing stiffness.
• Achieved a 25.7 N/µm increase in static stiffness compared to multi-objective genetic algorithms (MOGA).
• Reduced spindle deformation by 0.2 µm, demonstrating superior performance.

Conclusions:

• The proposed MLP-BO framework offers superior optimization efficiency, accuracy, and global search capability.
• This hybrid approach provides valuable insights for designing high-performance electric spindles.
• The methodology is applicable to other precision manufacturing systems requiring complex optimization.