Study on an Assembly Prediction Method of RV Reducer Based on IGWO Algorithm and SVR Model

Area of Science:

• Mechanical Engineering
• Machine Learning
• Predictive Modeling

Background:

• Existing rotation error prediction methods are time-consuming and lack accuracy.
• Current methods fail to meet enterprise production beat and product quality requirements.

Purpose of the Study:

• To develop a novel method for predicting rotation error with improved accuracy and efficiency.
• To address the limitations of existing rotation error research methods in industrial applications.

Main Methods:

• An improved grey wolf algorithm (IGWO) was developed using optimal Latin hypercube sampling initialization, a nonlinear convergence factor, and dynamic weights.
• The IGWO algorithm was used to optimize support vector regression (SVR) model parameters.
• An IGWO-SVR prediction model was established to predict rotation error based on critical part manufacturing errors, using an RV-40E reducer as a case study.

Main Results:

• The improved grey wolf algorithm demonstrated superior parameter optimization performance.
• The IGWO-SVR method significantly outperformed existing methods, including back propagation (BP) neural networks and SVR models optimized by particle swarm and standard grey wolf algorithms.
• The IGWO-SVR model achieved a mean squared error of 0.026, a running time of 7.843 seconds, and a maximum relative error of 13.5%.

Conclusions:

• The IGWO-SVR method effectively predicts rotation error, meeting production beat and product quality requirements.
• The proposed method shows significant potential for application in RV reducer parts-matching models to enhance product quality and reduce costs.