Self-Powered Systems for Structural Control: Theory and Experiment

Abstract

A self-powered system is a control actuation technology that derives all energy to power its operations from the dynamic response of the plant in which it is embedded. In the context of structural control of buildings or bridges, such systems are attractive as they do not rely on any external power supply, which may be unreliable during extreme events such as earthquakes. The simplest example of a self-powered control system consists of an electromechanical transducer embedded within a vibratory system and electrically connected to an energy storage subsystem. In this configuration, the transducer can absorb energy from the vibrating structure, store that energy, and then intelligently re-inject that stored energy at a future time to reduce the vibratory response. More generally, a network of several transducers could be embedded within a single structure (or in different structures) and share power with each other via a single storage system. Self-powered systems have great potential in a variety of applications for which energy-autonomy is desired or external power supplies are unreliable. However, the synthesis of feedback control laws for use with self-powered systems is challenging, as they must account for parasitic losses in the transducers, power electronics, and energy storage subsystem.

The first contribution of this dissertation is the establishment of an explicit feasibility condition for dynamic feedback laws which accounts for parasitic dissipation. This condition is general in the sense that it can accommodate time-invariant, time-varying, and infinite-dimensional controllers. We show how this criterion can be used to identify a Pareto surface of the least-efficient loss parameters needed to realize a linear, passive, colocated feedback controller using self-powered hardware. In essence, this allows the designer to determine how efficient must the transducers and energy storage subsystem be in order to successfully implement a given controller. The procedure requires formulating and solving a quasiconvex optimization, which we demonstrate via several numerical examples.

We also innovate tractable methodologies for the design of linear and nonlinear controllers, which adhere to the self-powered feasibility criterion. The scope of the techniques is restricted to linear-time-invariant (LTI), passive plants with exogenous disturbances modeled as stationary Gauss-Markov processes. We consider both single and multi-objective performance measures. Our nonlinear control framework makes use of a methodology known as Performance-Guaranteed Control, which although sub-optimal, has the distinct advantage of guaranteeing to improve upon the performance of any colocated, LTI self-powered controller.

Finally, we provide the first experimental validation of a bench-scale, prototype self-powered structural control system using real-time hybrid simulation (RTHS). RTHS is a cyber-physical structural testing method that interfaces numerical models with physical experiments in real time. In the RTHS method, a structural system is partitioned into so-called numerical and physical substructures. Actuators and sensors are used to enforce both displacement compatibility and force equilibrium at the substructure coupling points. In this dissertation, the numerical substructure consists of a linear building model subjected to a stochastic ground acceleration disturbance. The physical substructure (i.e., self-powered control system) consists of a permanent-magnet synchronous machine coupled with a linear ball-screw actuator, a three-phase inverter, DC-DC power converter, and storage capacitor.