Multivariable fuzzy neural network based on the one-layer network and Takaga-Sugeno fuzzy model was proposed in this paper where the parameters of the fuzzy rules and the inference process were all realized by neural network on-line, and the network was trained with the method of gradient descent. The presented method which possesses the ability of online learning and improving was applied to aero-engine accelerate process controller design, whose parameters vary significantly over the operation condition, the parameters of the controller were deduced real-time based on the change of the aero-engine condition with the adoption of single-layer network. An ECU-in-the-loop real-time simulation platform based on the rapid prototyping real-time simulation approach was constructed and the hardware-in-the-loop simulation was done with the aero-engine nonlinear component model, the results showed the well tracking and decouple as well as robust performance of the controller, meanwhile the effectivity of fuzzy neural network with the ability of self-learning and self improving has been proved.
Application of fuzzy neural network to aero-engine controller design
2009
4 Seiten, 10 Quellen
Conference paper
English
PI type multivariable robust controller design of aero-engine
British Library Online Contents | 1997
|A NEURAL NETWORK ADAPTIVE CONTROLLER CONSIDERING EXPERT SYSTEM FOR AERO-ENGINE (AIAA 2017-4010)
British Library Conference Proceedings | 2017
|Engineering Index Backfile | 1916