Optimize output of a piezoelectric cantilever by machine learning ensemble algorithms

Due to fast development of various electronic products in our life, the demand for sustainable energy supply devices has increased. Vibration energy harvest systems made by piezoelectric materials such as lead zirconate titanate (PZT) have gradually attracted widespread attention. For specific usage scenarios, we need to establish the relationship between geometry parameters and output voltage/power quickly, which is a difficult job. Here, we demonstrated that through processing 2430 sets experimental output voltage/power data generated by PZT cantilevers, we could find out the relationship between the output, the physical parameters and working frequency via machine learning algorithms quickly. Three machine learning ensemble algorithms (gradient boosting regression tree, random forest and extreme gradient boosting) are used to process these experimental data and the optimal algorithm is found. Our work showed that machine learning ensemble algorithm can help us design energy harvest systems efficiently.

Automated summary

This study demonstrates the relationship between the output, physical parameters, and working frequency of piezoelectric materials in vibration energy harvest systems through experimental data and machine learning algorithms. It provides assistance in designing efficient energy harvest systems.