Experimental and Machine Learning Study of a Modified Cymbal Piezoelectric Energy Harvester

Cymbal piezoelectric energy harvesters offer an effective platform for converting mechanical vibrations into electrical energy due to their ability to exploit both longitudinal (d₃₃) and transverse (d₃₁) piezoelectric coefficients. However, the design of flexible cymbal structures that ensure efficient stress transfer to polymer-based piezoelectric materials remains insufficiently explored. In this study, a bridge-like cymbal harvester incorporating polyvinylidene fluoride (PVDF) films as the active layer was designed, fabricated, and experimentally investigated. To support the design process and reduce the computational burden associated with evaluating multiple geometric configurations, we developed a novel machine learning methodology that integrates singular value decomposition (SVD) with metamodeling. This framework provides rapid predictions of resonance behavior and electrical response from key design parameters. The findings demonstrate the feasibility of PVDF-based cymbal harvesters for flexible energy harvesting applications and establish an efficient data-driven approach for guiding future design optimization.