Our heavyweight helicopter equal in the world does not have
In Rostov started production of the most load-lifting rotary-wing car The Russian holding «Helicopt[...]
ANNs IN CONTROL
ANNs are emerging paradigms for solving complex problems in science and engineering [7-10]. ANNs have the following features:
1. They mimic the behavior of the human brain.
2. They have massively parallel architecture.
3. They can be represented by adaptive circuits with an input channel, weights (parameters/coefficients), one or two hidden layers, and an output channel with some nonlinearities.
4. Weights can be tuned to obtain optimum performance of the neural network in modeling a dynamic system or nonlinear curve fitting.
5. They require a training algorithm to determine the weights.
6. They can have a feedback-type arrangement within the neuronal structure.
7. Trained network can be used to predict the behavior of the dynamic system.
8. They can be easily coded and validated using standard software procedures.
9. Optimally structured neural network architecture can be hardwired and embedded into a chip for practical applications—this will be the generalization of the erstwhile analog circuits-cum-computers.
The neural network-based system can be truly termed as a new generation powerful parallel computer.
The aerodynamic model (used in the design of flight-control laws) could be highly nonlinear and dependent on many physical variables. The difference between the mathematical model and the real system may cause performance degradation. To overcome this drawback ANN can be used and the weights adjusted to compensate for the effect of the modeling errors. ANNs can be used for control augmentation in several ways: (1) conventional control can be aided by ANN controllers for online learning to represent the local inverse dynamics of an aircraft, (2) they can be used to attempt to compensate for uncertainty without explicitly identifying changes in the aircraft model, (3) the ANN nonlinearity can be made adaptive and used in the desired dynamics block of the flight controller, (4) the learning ability can be incorporated into the gain-scheduling process, and (5) they can be used in sensor/ actuator failure detection and management. Some of the benefits would be (1) the controller becomes more robust and more insensitive to the aircraft/plant parameter variations and (2) the online learning ability would be useful in handling certain unexpected behavior, in a limited way, leading to reconfigurable/restructurable control.
