Modelling and parameter identification of electromechanical systems for energy harvesting and sensing

Kefal, Adnan and Maruccio, Claudio and Quaranta, Giuseppe and Oterkus, Erkan (2018) Modelling and parameter identification of electromechanical systems for energy harvesting and sensing. Mechanical Systems and Signal Processing. pp. 1-35. ISSN 0888-3270 (In Press)

[img]
Preview
Text (Kefal-etal-MSSP2018-Modelling-and-parameter-identification-of-electromechanical-systems)
Kefal_etal_MSSP2018_Modelling_and_parameter_identi_cation_of_electromechanical_systems.pdf
Accepted Author Manuscript
License: Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 logo

Download (6MB)| Preview

    Abstract

    Advanced modelling of electro-mechanical systems for energy harvesting (EH) and sensing is important to develop reliable self-powered autonomous electronic devices and for structural health monitoring (SHM). In this perspective, a novel computational approach is here proposed for both real-time and off-line parameter identification (PI). The system response is governed by a set of four partial differential equations (PDE) where the three displacement components and the electrical potential are the unknowns. Firstly, the finite element (FE) method is used to reduce the PDE problem into a set of ordinary differential equations (ODE). Then, a state- space model is derived with the aim to limit the PI problem to a subset of unknowns. After that, an identification error is introduced and the Lyapunov theory is used to derive the PI algorithm. The numerical implementation is based on a sensitivity analysis feedback block. The overall proposed computational strategy is robust and results in an exponential asymptotic convergence. The accuracy of the PI method is demonstrated by analysing the time–domain response of an array of piezoelectric bimorphs subjected to low–frequency structural random vibrations. The selected case–study is an existing cable–stayed bridge, for which an extensive dynamic monitoring campaign has provided the experimental data. Once time histories of the device response are obtained through time–dependent dynamic FE simulations, the PI algorithm is used to determine the unknown lumped coefficients of the state-space model. The comparison between FE method and lumped parameters model in terms of tip displacement and output voltage demonstrates the superior predictive capability of the new PI algorithm. As a result of the sensitivity analysis, guidelines to assess the optimal array configuration are also provided.