Construction and Experimental Implementation of a Model-Based Inverse Filter to Attenuate Hysteresis in Ferroelectric Transducers

Andrew G. Hatch, Ralph C. Smith, Tathagata De, Murti V. Salapaka

TL;DR

This paper presents a model-based inverse filter to compensate for the inherent hysteresis and nonlinearities in ferroelectric transducers, crucial for precise control in applications like atomic force microscopy. By developing a homogenized energy model and its inverse, they demonstrate significant improvements in tracking accuracy, especially at higher frequencies where hysteresis is more pronounced.

What Problem Does This Paper Address?

Ferroelectric materials, like PZT (lead zirconate titanate), are fantastic for actuators and sensors due to their electromechanical coupling, enabling high precision and broadband transduction. However, their unique crystal structure leads to significant hysteresis and nonlinearities. These effects, which are essentially memory effects where the output depends not just on the current input but also on its past history, make precise control challenging. While simple feedback control can handle these issues at low frequencies, it becomes insufficient at higher frequencies due to increased noise and reduced control authority. This paper aims to address this by developing a method to "linearize" the transducer's behavior, making it easier to control with standard linear techniques.

Methods

The core of this work lies in developing and implementing a model-based inverse filter. The process involves:

  • Homogenized Energy Model: They start by using a homogenized energy framework to create a macroscopic constitutive model that captures the electromechanical behavior of ferroelectric materials, including hysteresis. This model is derived from fundamental energy relations and accounts for material non-homogeneities.
  • Inverse E-P Relation: A key step is deriving the inverse relationship between the electric field (E) and polarization (P). This is crucial for creating the inverse filter. They develop an algorithm that efficiently computes the field required to achieve a desired polarization.
  • Lumped Transducer Model: The constitutive model is then used to create a lumped model for a specific transducer, in this case, a stacked PZT actuator used in an atomic force microscope (AFM) stage. This model captures the dynamic behavior of the transducer.
  • Inverse Displacement-Field Algorithm: Recognizing that the displacement-field (u-E) relation is not always monotonic (especially at higher frequencies), they develop an extended inversion algorithm that accounts for dynamic effects and hysteresis to accurately map desired displacements to the required electric fields.
  • Open-Loop Control Implementation: The developed inverse filter is experimentally implemented as an open-loop controller for an AFM stage. This means the filter directly calculates the input signal to the transducer to achieve a desired output, without relying on feedback to correct errors.

Datasets

The paper uses experimental data collected from a PZT-based AFM stage. This data includes field-displacement measurements at various frequencies (0.279 Hz, 5.58 Hz, and 27.9 Hz) and amplitudes. This data is used to:

  • Validate the developed constitutive and transducer models.
  • Estimate model parameters through least squares fitting.
  • Evaluate the performance of the inverse filter in open-loop tracking experiments.

No new benchmark datasets are introduced; rather, existing experimental data from the AFM stage is utilized.

Key Insights and Ideas

  • Hysteresis is Frequency-Dependent: The paper clearly illustrates how the severity of hysteresis in ferroelectric transducers increases with frequency, posing a significant challenge for high-speed applications.
  • Inverse Filtering as a Linearization Tool: The core idea is that by accurately modeling the transducer's nonlinear behavior and then inverting that model, you can effectively "linearize" the system. This allows for the application of well-established linear control design techniques.
  • Energy-Based Models Offer Unification: The use of a homogenized energy framework provides a unified approach to modeling hysteresis across different ferroic materials (ferroelectric, ferromagnetic, ferroelastic), highlighting its broad applicability.
  • Dynamic Effects Crucial for High Frequencies: At higher frequencies, dynamic (inertial) effects become significant and must be accounted for in the inverse model to achieve accurate tracking.
  • Open-Loop Control with Inverse Filters is Powerful: The experimental results demonstrate that even without feedback, an accurately designed inverse filter can dramatically improve tracking performance in the presence of significant hysteresis.

Key Results

  • Significant Accuracy Improvement: In experiments at 27.9 Hz, the model-based inverse filter achieved tracking errors approximately ten times smaller than a simple linear filter that did not account for hysteresis.
  • Marginal Improvement at Low Frequencies: At low frequencies (0.279 Hz), where hysteresis is minimal, the inverse filter provided only marginally better accuracy than the linear filter. This highlights that the benefit of the inverse filter is most pronounced when hysteresis is significant.
  • Model Validation: The developed lumped transducer model, based on the homogenized energy constitutive relations, accurately characterized the frequency-dependent dynamics and hysteresis of the AFM stage, as shown by comparisons with experimental data.
  • Efficient Algorithms: The paper emphasizes the development of efficient algorithms for both the forward model evaluation and the inverse filter implementation, which are crucial for real-time applications.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that the accuracy of the inverse filter is ultimately limited by the accuracy of the underlying model. Modeling errors will still lead to some deviation in physical systems.
  • The experimental results show variability between experiments, suggesting that precise voltage application and potential feedback mechanisms are still necessary for ultimate control precision.
  • The current work focuses on a specific type of transducer (stacked PZT actuator).
  • Future Work:
    • Extension to Other Materials: The authors are investigating the extension of these inverse filtering techniques to ferromagnetic and ferroelastic materials.
  • Robust Feedback Control: A significant area of future research is the development of robust feedback control designs that leverage these inverse filters to linearize transducer dynamics, particularly for magnetic transducers.
  • Experimental Feedback Designs: Experimental implementation of feedback designs utilizing these inverse filters is currently under investigation.

Summary updated

Abstract

Hysteresis and constitutive nonlinearities are inherent properties of ferroelectric transducer materials due to the noncentrosymmetric nature of the compounds.In certain regimes, these effects can be mitigated through restricted input fields, charge-or current-controlled amplifiers, or feedback designs.For general operating conditions, however, these properties must be accommodated in models, transducer designs, and model-based control algorithms to achieve the novel capabilities provided by the compounds.In this paper, we illustrate the construction of inverse filters, based on homogenized energy models, which can be used to approximately linearize the piezoceramic transducer behavior for linear design and control implementation.Attributes of the inverse filters are illustrated through numerical examples and experimental open loop control implementation for an atomic force microscope stage.i
BibTeX
@article{hatch2006construction,
  author = {Andrew G. Hatch and Ralph C. Smith and Tathagata De and Murti V. Salapaka},
  title = {{Construction and Experimental Implementation of a Model-Based Inverse Filter to Attenuate Hysteresis in Ferroelectric Transducers}},
  journal = {IEEE Trans. Control. Syst. Technol},
  year = {2006},
  doi = {10.1109/tcst.2006.883195},
}
Figure