Design and reliability-based optimization of the piezoelectric flex transducer

Abstract

In recent years, the rapid development of low power consuming devices has resulted in a high demand for mobile energy harvesters. The main contribution of this thesis is to optimize the novel piezoelectric energy harvesting device called the piezoelectric flex transducer, which was developed by other researchers for the purpose of harvesting biokinetic energy from human gait. The optimization uses both conventional and reliability-based optimization approaches in order to improve the electrical power generation from the device. First, the piezoelectric flex transducer is modeled by using the finite element method with the finite element analysis software ANSYS APDL. Seven geometric parameters of the piezoelectric energy harvester are considered as design variables. A set of designs with different design variables are generated by the Design of Experiment technique, the generated designs are analyzed by the finite element model and the surrogate models that representing the behavior of the FEM are built by these inputs and the results of the FEA. Conventional optimization, taking into consideration different safety factors, is driven by the von mises stress of the device and is then searched by a mathematical algorithm with the assistance of surrogate models. To improve the efficiency of the surrogate modeling, a multi-level surrogate modeling approach for fast convergence will be introduced and the method will be demonstrated by optimizing the PFT device. As the optimal design is subject to a low stress safety factor, which may be unreliable with the uncertainties of the real-world, the reliability and sensitivity of the optimal design are analyzed. A Monte Carlo simulation is employed to analyse how the electrical power output has been affected by the input parameters with parametric uncertainties. The design parameters of a set of designs are perturbed around the optimal design parameters in order to imitate the optimal design under parametric uncertainties. The effects of parametric uncertainties are then evaluated by the constructed surrogate models. The method for improving the product reliability will be demonstrated.