機密文件

機密等級 : 想看就看,懶的理你


Problem Formulation

Consider the PID control system in Fig.1. with the perturbation denotes torque constant. The plant (induction motor) with the perturbation , the tachometer and the PID controller are of the following form:

where is a conversion constant, is inertia of the motor and load , is viscous constant, is constant and the plant perturbation and are assumed to be stable but uncertain. Suppose and are bounded according to

,

where the function and positive constant are stable and known.

The robust stability result in [] reveals that if a controller is chosen so that the nominal closed loop system (free of ) in Fig.1 is asymptotically stable, and the following inequality holds,

then the closed loop system in Fig.1 is also asymptotically stable under plant perturbation in ( ) , where the norm in ( ) is defined as

i.e., the maximum peak of the spectral density of .

However, often robust stability alone is not enough in control system. Optimal tracking performance is also appealing in many practical control engineering applications. Therefore, the mixed control problem is formulated as follows:

for the nominal closed loop system in Fig.1 , subject to the robust stability constraint ( ), where is the tracking error and is actuator signal. That is, under the constraint ( ), the summation of error energy and the control energy in ( ) must be as small as possible. from the above analysis, the PID mixed control design problem involves how to specify a PID controller to achieve the optimal tracking ( ) subject to the robust stability constraint ( ).

In the case with uncertain disturbance and measurement noise (see Fig.1), if the desired attenuation level and are specified, then

where and are weighting functions.

PID Mixed Optimal Control Design

By Parseval's theorem [ ], we have

where both and for i=1,2 are Hurwitz polynomials of s with appropriate degree.

The minimization problem in the above equation can be solved via the aid of the residue theorem. Let and ;if

The value of can be found easily from published tables[ }: