Model-Based Architecture for Control System Design with Application of SIMO Neural Network

Abstract

The design problem is considered built on the model-based architecture approach as applied to the design of control systems for nonlinear control objects. The application of a neural network is discussed, a feature of which is the structure presented in the single input multiple output class. The possibility of using this class of a neural network as a brick in the framework of an architectural approach for constructing control systems for dynamic multi-mode control objects is considered. An example of developing a fault-tolerant control system for a nonlinear control object is discussed. As a model of control system fault, it is proposed to use a decision tree, with the help of which the approximation of the fault image presented in the space of modes and behavior of the control error is performed