Data-Driven Modeling and Response Prediction of Cut-Out Type Piezoelectric Beams

Area of Science:

• Mechanical Engineering
• Electrical Engineering
• Materials Science

Background:

• Theoretical models for piezoelectric beams with limiters often lack accuracy due to contact-impact nonlinearity.
• Accurate modeling is crucial for optimizing energy harvesting efficiency.

Purpose of the Study:

• To develop a data-driven modeling approach using backpropagation neural networks (BPNN) for predicting the performance of piezoelectric energy harvesters.
• To improve the accuracy of output voltage and displacement predictions under nonlinear conditions.

Main Methods:

• Experimental data on amplitude-frequency and time-voltage responses were collected for various limiter gap and installation distances.
• A multi-layer BP neural network was trained using experimental data, with frequency or time as input and voltage/displacement as output.

Main Results:

• The BPNN model accurately predicted amplitude-frequency response curves for voltage and displacement across different parameter combinations.
• The model also accurately predicted transient voltage outputs under varying load resistances.

Conclusions:

• The data-driven BPNN approach offers a highly accurate method for modeling piezoelectric energy harvesters with contact-impact nonlinearity.
• This approach enhances the prediction capabilities for energy harvesting devices, particularly concerning distance parameters and load conditions.