Representation of control systems for preliminary Space Station design

Abstract

Introduction T field of control/structure interaction combines the two disciplines of structural dynamics and control engineering. This pairing can also imply combining two sets of software tools, one for each discipline. For applications where the control systems can be represented as linear dynamic components, however, it is often possible to analyze the control/structure interaction problem within a structural dynamics code. Proportional plus derivative controllers with collocated sensors and actuators provide a trivial case that can be represented by springs and dampers. This paper extends the representation to arbitrary multiple-input multipleoutput linear control systems. When the approach is applicable, it offers some advantages over the more common approach of transferring modal data to a control system simulation code. Elimination of excess software accelerates turnaround time for simulations, while reducing training time. A more significant advantage is that the structural dynamics routine treats the problem in physical coordinates, which allows the plotting of the closed-loop mode shapes. This can greatly increase the engineer's physical insight into the problem. Analyses that can be performed using this approach include the evaluation of closed-loop eigenvalues and eigenvectors (stability analysis). Transient response to physical forcing functions can be calculated along with openor closed-loop transfer functions (Bode plots). Although most structural dynamics routines do not offer a root-locus capability, this can be performed by manually varying a parameter and recalculating closed-loop eigenvalues. In summary, most of the linear analyses available to the control engineer can be performed directly in a structural dynamics routine. This paper is organized as follows: First, the method for incorporating control systems is described; next, examples of PID attitude control and manipulator control on a realistic space station model are presented, followed by some conclusions. The overall software approach is described in greater detail in Refs. 1-3.