Frequency dependence prediction and parameter identification of rubber bushing

Area of Science:

• Mechanical Engineering
• Materials Science
• Computational Mechanics

Background:

• Rubber bushing dynamic mechanical properties exhibit nonlinear behavior influenced by frequency and amplitude.
• Understanding frequency dependence is crucial for accurate simulation of vehicle vibration.

Purpose of the Study:

• To predict the dynamic stiffness and loss factor of rubber bushings at higher frequencies (61-100 Hz) using a GA-BP neural network.
• To develop and validate a five-parameter mathematical model (FPM) for simulating rubber bushing behavior in engineering software.

Main Methods:

• A genetic algorithm-optimized BP neural network (GA-BP) was trained on test data from 1-60 Hz to predict properties at 61-100 Hz.
• Elastomer tests were conducted at amplitudes of 0.2 mm, 0.4 mm, and 0.6 mm.
• A five-parameter mathematical model (FPM) was developed and its parameters identified using the least squares method.

Main Results:

• The GA-BP neural network achieved prediction errors of less than 1% for dynamic stiffness and less than 3% for the loss factor.
• The FPM demonstrated fitting errors of less than 2% for dynamic stiffness and less than 3% for the loss factor when compared to test data.

Conclusions:

• The GA-BP neural network and FPM provide an efficient and cost-effective method for predicting rubber bushing dynamic mechanical behavior.
• These models reduce the need for iterative experiments and high computational costs, making them suitable for analyzing complex vehicle systems under various vibration conditions.