Page 149 - Neural Network Modeling and Identification of Dynamical Systems

4.3 APPLICATION OF ANN MODELS TO ADAPTIVE CONTROL PROBLEMS UNDER UNCERTAINTY CONDITIONS 139
some cases, the accuracy degenerates, which in to a changing situation by modifying some of its
turn leads to poor adaptation performance of elements. We assume that usually these modi-
the synthesized neurocontroller. Ways to over- fications are applied to the control laws imple-
come these difficulties are discussed in Chap- mented by the control system and to the con-
ters 5 and 6. trolled object model. Modifications of these sys-
tems may affect the corresponding parameter
values as well as the structure of the control laws
4.3 APPLICATION OF ANN and/or system models.
MODELS TO ADAPTIVE The analysis of the adaptive control algo-
rithms concerning aircraft motion is carried out
CONTROL PROBLEMS FOR
in the following subsections regarding such ba-
NONLINEAR DYNAMICAL sic types of adaptive systems as the model refer-
SYSTEMS OPERATING UNDER ence adaptive control (MRAC) and model pre-
UNCERTAINTY CONDITIONS dictive control (MPC).
Another option within the framework of the
4.3.1 The Demand for Adaptive Systems discussed approach to the control of nonlin-
ear dynamical systems is considered in [27,28].
One of the most important classes of dy- In this case, we solve the problem of obtain-
namical systems is aircraft of various types. ing the specified characteristics of the control-
As already noted in Chapter 1, it is crucial lability of the aircraft through automation. The
that we provide for control of the motion of question that is solved by ANN tools is high-
modern and advanced aircraft under conditions precision control over the entire range of flight
of significant and varied uncertainties in the modes. Synthesis and testing of neural network–
values of their parameters and characteristics, based control algorithms are performed using
flight regimes, and environmental influences. the full nonlinear mathematical model of a ma-
Besides, during flight, various emergencies can neuverable aircraft through three control chan-
arise, in particular, equipment failures and struc- nels for the current flight mode. In the structure
tural damage, the consequences of which in of the system, the internal contour of the control
most cases can be overcome by an appropri- of angular velocities is distinguished, which is
ate reconfiguration of the aircraft control sys- formed by the Inverse Dynamics method, which
tem. is based on the feedback linearization [29]. At
The presence of significant and diverse un- the same time, using the feedback transforma-
certainties is one of the most severe factors that tion, the controlled object is reduced to an equiv-
complicate the solution of all the three classi- alent linear form, after which the control is se-
cal problems (analysis, synthesis, identification) lected so that the object moves along a predeter-
for dynamical systems and, in particular, for air- mined desired trajectory. The external contour of
craft. The problem is that the current situation control of the angle of attack in this system con-
can change dramatically, significantly, and un- tains a PI controller.
predictably due to the uncertainties. The con- The problem of improving the accuracy of the
trolled system under such conditions must be control system should be considered provided
able to adapt quickly to changes in the situa- that the aerodynamic and other characteristics of
tion to ensure the normal operation of the sys- the aircraft are nonlinear and are characterized
tem. by a high level of uncertainty. Including, as it
As already noted in Chapter 1,weconsider was noted earlier, one of the types of uncertain-
the system to be adaptive if it can quickly adapt ties can be interpreted as the failure of aircraft