RF-SVR-based prediction methodology for metal tube-bending rebound: Handling non-uniformity and limited sample...

Area of Science:

• Mechanical Engineering
• Materials Science
• Data Science

Background:

• Predicting the rebound angle of tubes is crucial in various engineering applications.
• Non-uniform and small-sample datasets pose significant challenges for traditional prediction algorithms.
• Existing methods often struggle with data variability and limited sample sizes.

Purpose of the Study:

• To develop an advanced prediction algorithm for determining the rebound angle of non-uniform and small-sample tubes.
• To address the limitations of current methods in handling data heterogeneity and scarcity.
• To improve the accuracy and reliability of rebound angle predictions in metal tube bending.

Main Methods:

• Proposed a novel algorithm combining Random Forest (RF) and Support Vector Regression (SVR), termed RF-SVR.
• Employed polynomial feature generation to manage non-uniform data.
• Utilized RF for feature selection to identify influential rebound characteristics.
• Applied SVR for the final regression prediction of the rebound model.

Main Results:

• The RF-SVR algorithm demonstrated superior performance compared to traditional RF-BP and SVR methods.
• Achieved significantly higher prediction accuracy on datasets characterized by non-uniformity and small sample sizes.
• Successfully retained critical features influencing the rebound angle through RF-based selection.

Conclusions:

• The RF-SVR approach offers a robust solution for predicting the rebound angle of metal tubes.
• This method effectively overcomes challenges associated with non-uniform and small-sample data.
• The study highlights the potential of hybrid machine learning models in engineering predictions.