2017 Theses Doctoral

# Frequency Response Based Repetitive Control for Periodic Coefficient Systems Motivated by Cam Followers

Cam follower systems are generally designed to operate at a fixed speed or a range of fixed speeds. However manufacturing defects, wear, or a change of design goals may require altering the camshaft speed to produce a follower trajectory which is not possible using a fixed speed. The follower trajectory may also be optimized for some performance criteria such as minimizing vibration and wear. Like most real world systems, the differential equations governing a cam follower system are nonlinear.

A common approach for controlling a nonlinear system is to first linearize the system about a nominal operating point, then apply linear control laws. In many cases, such as the cam follower system, one can create a trajectory and numerically solve the nonlinear system for the inputs required to follow it.

Linearizing about this solution creates a linear time varying system whose states are deviations from the desired solution. The speed trajectory in the cam follower system is periodic, which results in a linear system with periodic coefficients.

Repetitive control creates control systems that aim to converge to zero tracking error following a periodic command, or aim to completely cancel the effects of a periodic disturbance. Using the inverse of the steady state frequency response as a compensator has been shown to be very effective for linear time invariant systems. That idea is applied here to linear time periodic systems. The periodic state matrices lend themselves well to frequency domain representations, which can be used to construct a matrix form of the steady state frequency response.

The first law studied in this work analyzes a moving window implementation which monitors the output errors and previous commands to create an update to the change in the command for the current time step using the inverse of the steady state frequency response matrix. Asymptotic convergence conditions for zero tracking error are derived.

When the number of samples in one period is not an integer number, the moving window method is not feasible without interpolation. Therefore a second method based on the projection algorithm from adaptive control is developed and analyzed.

In linear constant coefficient systems, one generally needs to incorporate a frequency cutoff filter to robustify to high frequency model error. The additional intricacies of designing a cutoff filter for periodic systems is considered, aiming to handle the fact that for periodic coefficient systems, addressing error components below the intended cutoff can excite harmonics above the cutoff.

The control laws developed in this work are applicable to any nonlinear system which may be linearized about a periodic trajectory.

Development of these control laws is motivated by improving the performance of a cam follower system. Additional improvements in cam follower behavior can be done through parameter optimization. This includes optimizing a nonlinear follower spring such that it provides just sufficient force to maintain contact while reducing the load on the cam.

## Subjects

## Files

- Yau_columbia_0054D_13755.pdf binary/octet-stream 1.74 MB Download File

## More About This Work

- Academic Units
- Mechanical Engineering
- Thesis Advisors
- Longman, Richard W.
- Degree
- Ph.D., Columbia University
- Published Here
- February 6, 2017