Hysteresis nonlinearity in magnetic and smart material systems hinders the realization of their potential in sensors and actuators. The goal of this dissertation is to advance methods for modeling, identification, and control of hysteretic systems. These methods are applied to the inverse compensation, self-sensing feedback control, and robust control of vanadium dioxide (VO2) microactuators. As a novel smart material, VO2 undergoes a thermally induced insulator-to-metal transition and a structural phase transition, exhibiting pronounced hysteresis in electrical and mechanical domains. With the goal of obtaining accurate hysteresis models while maintaining a low model complexity, optimal compressions for two popular hysteresis models, namely the Preisach operator and the generalized Prandtl-Ishlinskii (GPI) model, are studied, where the Kullback-Leibler divergence and entropy, respectively, are adopted to quantify the information loss in model compression. While the optimal compression of the Preisach operator is realized using exhaustive search, dynamic programming is employed to optimally compress the GPI model efficiently. Both simulation and experimental results demonstrate that the proposed algorithms yield superior performances than typically adopted schemes. In order to identify the Preisach operator, existing work involves applying a complicated input sequence and measuring a large set of output data. We propose an efficient approach to identify the Preisach operator that requires fewer measurements. The output of the Preisach operator is transformed into the frequency domain, generating a sparse vector of discrete cosine transform (DCT) coefficients. The model parameters are reconstructed using a compressive sensing-based algorithm. The effectiveness of the proposed scheme is illustrated through simulation and experiments. A few new contributions have been made to the modeling and control of VO2 microactuators. In order to capture the non-monotonic curvature-temperature hysteresis of VO2 microactuators, physics-motivated models that combine a monotonic hysteresis operator for phase transition induced curvature and a memoryless operator for differential thermal expansion induced curvature are proposed. Effective inverse compensation schemes for the proposed non-monotonic hysteresis models are presented. The modeling and inverse compensation schemes are validated experimentally. Since external sensing systems are not desirable with micro devices, a self-sensing model is developed for VO2 microactuators to estimate the deflection from the resistance measurement. We exploit the physical understanding that each of the resistance and the deflection is determined by a hysteretic relationship with the temperature, which is modeled with a GPI model and an extended GPI model, respectively. The self-sensing model is obtained by cascading the extended GPI model with the inverse of the GPI model. The performance of the self-sensing scheme is experimentally evaluated with proportional-integral control. Finally, an H_infty robust controller is further developed, where a simple polynomial-based self-sensing scheme is adopted, as the emphasis is on accommodating the uncertainties produced by the hysteresis nonlinearity and the self-sensing error. The effectiveness of the proposed approach is demonstrated through experiments.
Read
