Model-based self-sensing algorithm for dielectric elastomer transducers based on an extended Kalman filter

Abstract Self-sensing enables electromechanical transducers of being simultaneously used for actuating and sensing without additional, external sensors. In case of dielectric elastomer (DE) transducers the shape-varying, strain-dependent capacitance is usually considered for this purpose and estimated based on the measured terminal voltage and current. In the literature most self-sensing algorithms for DE transducers require an explicit superimposition of an additional AC excitation signal. Thus, only power supplies providing the opportunity of superimposing this excitation signal can be used for this purpose. Within this contribution we propose a new approach based on an extended Kalman filter. It operates with any arbitrary driving signal and thus does not require a superimposed excitation. Here, the extended Kalman filter is used for both the estimation of the inner states and parameters of the nonlinear process. The accuracy and dynamics of the estimation results obtained with proposed self-sensing approach are demonstrated by comparing the estimated with the measured transducer strain. The measurements are conducted with two different power supplies showing the applicability and robustness of the proposed self-sensing approach ensuring reliable and accurate estimation results independent of the kind of excitation signal. Here, the utilized high voltage amplifier represents a common power supply for experimental studies in laboratory environment. In contrast, the bidirectional flyback-converter is an energy-efficient and economical opportunity for feeding of DE transducers in technologically sophisticated applications. However, due to its operation principle it does not provide the opportunity to superimpose particular sensing signals so that this new approach is required to enable the self-sensing capability.