MODEL FOLLOWING AND HIGH ORDER AUGMENTATION FOR ROTORCRAFT CONTROL, APPLIED VIA PARTIAL AUTHORITY
Open Access
- Author:
- Spires, James Michael
- Graduate Program:
- Aerospace Engineering
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- April 25, 2017
- Committee Members:
- Joseph Horn, Dissertation Advisor/Co-Advisor
Joseph Horn, Committee Chair/Co-Chair
Edward Smith, Committee Member
Jacob Langelaan, Committee Member
Chris Rahn, Outside Member - Keywords:
- rotorcraft control
model following control
dynamic decoupling
explicit model following
H2
Hinf
dynamic inversion
nonlinear dynamic inversion
handling qualities
partial authority control
feed forward
turbulence rejection
piggyback compensation - Abstract:
- This dissertation consists of two main studies, a few small studies, and design documentation, all aimed at improving rotorcraft control by employing multi-input multi-output (MIMO) command-model-following control as a baseline, together with a selectable (and de-selectable) MIMO high order compensator that augments the baseline. Two methods of MIMO command-model-following control design are compared for rotorcraft flight control. The first, Explicit Model Following (EMF), employs SISO inverse plants with a dynamic decoupling matrix, which is a purely feed-forward approach to inverting the plant. The second is Dynamic Inversion (DI), which involves both feed-forward and feedback path elements to invert the plant. The EMF design is purely linear, while the DI design has some nonlinear elements in vertical rate control. For each of these methods, an architecture is presented that provides angular rate model-following with selectable vertical rate model-following. Implementation challenges of both EMF and DI are covered, and methods of dealing with them are presented. These two MIMO model-following approaches are evaluated regarding (1) fidelity to the command model, and (2) turbulence rejection. Both are found to provide good tracking of commands and reduction of cross coupling. Next, an architecture and design methodology for high order compensator (HOC) augmentation of a baseline controller for rotorcraft is presented. With this architecture, the HOC compensator is selectable and can easily be authority-limited, which might ease certification. Also, the plant for this augmentative MIMO compensator design is a stabilized helicopter system, so good flight test data could be safely gathered for more accurate plant identification. The design methodology is carried out twice on an example helicopter model, once with turbulence rejection as the objective, and once with the additional objective of closely following pilot commands. The turbulence rejection HOC is feedback only (HOC_FB), while the combined objective HOC has both feedback and feedforward elements (HOC_FBFF). The HOC_FB was found to be better at improving turbulence rejection but generally degrades the following of pilot commands. The HOC_FBFF improves turbulence rejection relative to the Baseline controller, but not by as much as HOC_FB. However, HOC_FBFF also generally improves the following of pilot commands. Future work is suggested and facilitated in the areas of DI, MIMO EMF, and HOC augmentation. High frequency dynamics, neglected in the DI design, unexpectedly change the low frequency behavior of the DI-plant system, in addition to the expected change in high frequency dynamics. This dissertation shows why, and suggests a technique for designing a pseudo-command pre-filter that at least partially restores the intended DI-plant dynamics. For EMF, a procedure is presented that avoids use of a reduced-order model, and instead uses a full-order model or even frequency-domain flight test data. With HOC augmentation, future research might investigate the utility of adding an H∞ constraint to the design objective, which is known as an equal-weighting mixed-norm H2/H∞ design. Because all the formulas in the published literature either require solution of three coupled Riccati Equations (for which there is no readily available tool), or make assumptions that do not fit the present problem, appropriate equal-weighting H2/H∞ design formulas are derived which involve two de-coupled Riccati Equations.