| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 389.69 KB | Adobe PDF |
Advisor(s)
Abstract(s)
In a novel branch of soft computing developed in the past few years the desired and the expected response of the system is mapped to each other. In the case of mechanical systems the compared values are the second time-derivatives of the joint coordinates for the estimation of which certain finite element approximations are used in a digital control. This may result in a kind of noise and estimation sensivity. In the present paper these integer order derivatives are replaced by discrete numerical estimations of fractional order derivatives near the order of two to make the control more stable and accurate. For this purpose Caputo's form is considered the numerical approximation of which can be extended over the limits of the original definition. In this view differentiation seems to be an operation with some time-invariant Green function. Simulation results obtaind for the adaptive control of an inaccurately modeled electromechanical system containing an unmodeled and undriven internal degree of freedom illustrate that the quality of the control can be improved if the order of derivation in the signals used for comparison are increased form 2 to 2.25.
Description
Keywords
Soft computing Integer order derivatives Fractional order derivatives