Identification of pre-sliding friction dynamics

Area of Science:

• Tribology and Materials Science
• Nonlinear Dynamics and Control Systems

Background:

• Pre-sliding friction exhibits complex hysteretic nonlinear behavior dependent on displacement.
• Accurate modeling of this behavior is crucial for understanding and controlling mechanical systems.

Purpose of the Study:

• To model the hysteretic nonlinear dependence of pre-sliding friction force on displacement.
• To compare the efficiency and accuracy of various physics-based and black-box identification methods.
• To investigate the potential of ensemble modeling for improved friction prediction.

Main Methods:

• Physics-based models (e.g., Maxwell-slip models).
• Black-box models including NARX models, neural networks, nonparametric models, and dynamical networks.
• Validation using an experimental time series for predicting friction force from displacement.

Main Results:

• All tested models demonstrated good prediction capability for pre-sliding friction.
• Varying degrees of accuracy were observed among the different identification methods.
• Ensemble modeling using the best individual models yielded superior prediction results.

Conclusions:

• Multiple modeling approaches can effectively capture pre-sliding friction behavior.
• Ensemble methods offer a significant advantage for enhancing prediction accuracy in friction modeling.
• The findings contribute to more robust friction force prediction in mechanical systems.