This paper presents practical aspects of implementing a dead zone inverse on a hydraulic wrist. A dead zone occurs over a range of small input values for which a system does not respond. It was desirable to use the most straightforward method available to achieve improved system performance while requiring the least amount of modification to the controller. Thus a fixed parameter dead zone inverse (DZI) was added to an existing proportional-integral (PI) controller. First, the parameters of the dead zone were characterized from open loop testing. These parameters are the break points, or input values between which the system does not respond at all, and the slope of the system’s response just outside the break points. The DZI augments the PI signal input to the plant, effectively adding or subtracting a constant equal to the size of the dead zone break points and scaling the input by its slope. Simulations predicted perfect system tracking, but implementation on the hardware revealed several practical issues. First, the dead zone slope parameters vary throughout the robot’s workspace. Overestimation can lead to non-ideal system performance, but the more extreme problem is underestimation, which effectively increases control loop gain and can lead to system instability. However, performance is not affected significantly unless these parameters are off by an order of magnitude. Overall the system performance is relatively robust to modeling errors in the slope parameters. The second issue is that noise can be magnified by the dead zone inverse and cause chattering. This problem was very noticeable in the wrist when the estimated dead zone break points were used in the DZI. This problem can be eliminated by reducing the dead zone break points or reintroducing a small artificial dead zone back into the control loop to envelope the expected noise level. The requirements for successful implementation of the DZI were found to be a basic characterization of the dead zone and an understanding of practical system issues that can be accentuated by its use. The effectiveness of the technique was tested through simulations and experiments on the wrist.

No abstract

Systems with flexible dynamics often vibrate due to external disturbances, as well as from changes in the reference command. Feedback control is an obvious choice to deal with these vibrations, but in many cases, it is insufficient or difficult to implement. A technique that does not rely on high performance feedback control is presented here. It utilizes a combination of vibration absorbers and input shapers. Vibration absorbers have been used extensively to reduce vibration from sinusoidal disturbances, but they can also be implemented to reduce the response from transient functions. Input shaping has proven beneficial for reducing vibration that is caused by changes in the reference command. However, input shaping does not deal with vibration excited by external disturbances. In this paper, vibration absorbers and input shapers are designed sequentially and concurrently to reduce vibration from both the reference command and from external disturbances. The usefulness of this approach is demonstrated through computer simulations and experimental results.

No abstract

Mechanical systems with flexible dynamics often suffer from vibration induced by changes in the reference command or from external disturbances. The technique of adding a vibration absorber has proven useful at eliminating vibrations from external disturbances and rotational imbalances. Traditionally, vibration absorbers have been designed for systems subject to sinusoidal or random excitations. Here the applicability of vibration absorbers to systems with steplike changes in the reference command or similar disturbances is studied. This type of motion is more common in robotic applications. Here absorbers are designed using two methods; the first technique uses a weighting on peak overshoot and settling time to allow tradeoffs between the two performance criteria. The second simpler method utilizes an eigenvalue technique to predict the time constant. Both of these techniques provide the possibility of significant improvement in settling time. The performance of this absorber design strategies is compared with previously proposed vibration absorbers and experimental results verify its utility.

Micro milling requires both high speed and high accuracy in order to economically produce parts with features on the scale of 1 µm. Micro mills are smaller and more flexible than traditional large-scale machines. Therefore, vibration of the machine structure is a significant problem. Given that micro milling requires high positioning precision, even small vibrations in the controller dynamics are problematic. The small-scale operation has considerably lower tool/workpiece interaction forces than traditional-scale milling. These low cutting forces have minimal effect on the machine structural response. Therefore, the dominant dynamic factor in exciting vibration can be the machine tool motion, rather than the workpiece/tool interaction. Given this realization, properly shaping the motions of the micro mill is a promising approach to reduce vibration. This paper presents a nonlinear commandshaping technique to reduce the vibrations of a micro mill that can be implemented with a standard CNC controller. The robustness of this technique to modeling errors and disturbances is investigated. Theoretical proofs and experimental demonstrations of the command-shaping technique are presented. The improved performance from the command shaping enables higher throughput and improved accuracy of the micro mill.

No abstract

No abstract