Sensitivity Analysis of RV Reducer Rotation Error Based on Deep Gaussian Processes

Area of Science:

• Mechanical Engineering
• Robotics
• Precision Engineering

Background:

• Rotation error is a critical quality index for rotate vector (RV) reducers.
• Optimizing RV reducer design is complex due to numerous interacting factors affecting rotation error.

Purpose of the Study:

• To develop a high-precision predictive model for RV reducer rotation error.
• To identify and quantify the sensitivity and coupling effects of factors influencing RV reducer rotation error.

Main Methods:

• Optimal Latin Hypercube Sampling (OLHS) for data generation.
• Deep Gaussian Processes (DeepGP) for building a high-precision regression prediction model.
• Sobol sensitivity analysis (SA) for global analysis of influencing factors and their coupling.

Main Results:

• A prediction model with up to 95% accuracy was achieved using OLHS and DeepGP.
• Rotation error is primarily influenced by factors in the second-stage cycloidal pinwheel drive.
• The primary involute planetary part and planetary output carrier have minimal impact on rotation error.

Conclusions:

• The DeepGP model and OLHS approach are effective for predicting RV reducer rotation error.
• Key factors influencing rotation error are identified, with a focus on the second-stage drive.
• Significant coupling effects exist between pin gear/needle gear hole clearance (δJ) and needle gear hole position error (δt).