Dynamic Neural Network Controller for Variable Cycle Engine
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-variable control systems for variable cycle engines face challenges in managing the strong coupling between nonlinear control variables, leading to complex controller structures and reduced accuracy due to reliance on accurate engine models and inadequate neural network structure determination.
Innovation Solution
A dynamic neural network-based control method is introduced, utilizing a grey relation analysis method for structure adjustment during training to construct a dynamic neural network controller, which adjusts its structure dynamically to improve training efficiency and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional robust control, LQR control or intelligent optimization algorithms are used for multi-variable control of variable cycle engines, then the control system can be implemented, but the coupling between nonlinear control variables cannot be solved well and the controller structure becomes more complicated
Solution Approach 1:
The patent replaces traditional mechanical control systems (robust control, LQR control, intelligent optimization algorithms) with a neural network-based controller. The neural network learns the complex nonlinear relationships between multiple control variables through training, automatically handling the coupling effects without requiring complex controller architecture. This substitution transforms the control approach from model-based to data-driven, simplifying the controller structure while maintaining or improving control effectiveness.
Solution Approach 2:
The patent changes the fundamental parameter of control methodology from traditional control algorithms to neural network parameters. By training the neural network on engine operating data, the system adapts to the nonlinear coupling characteristics of variable cycle engines. The neural network's ability to dynamically adjust its internal parameters (weights and biases) allows it to handle the complex interactions between control variables more effectively than fixed-structure traditional controllers.
2Reliability
If a neural network-based controller is used to solve the coupling between multiple variables, then the coupling problem can be addressed, but the structure of hidden layers cannot be determined directly, causing slow training and overfitting if too large or insufficient accuracy if too small
Solution Approach 1:
The patent applies dynamics by making the neural network structure adaptive rather than fixed. The system dynamically determines the appropriate hidden layer structure during the training process based on the complexity of the control task and the available data. This dynamic approach allows the network to optimize its own architecture, avoiding the trade-off between network size and training efficiency. The structure can evolve during training to achieve the required accuracy without unnecessary complexity that would cause overfitting or slow training.
Solution Approach 2:
The patent implements feedback mechanisms during the neural network training process to monitor performance and adjust the network structure accordingly. By continuously evaluating the training progress and control accuracy, the system can determine when the optimal hidden layer structure has been achieved. This feedback loop prevents both overfitting (by stopping before the network becomes too complex) and underfitting (by ensuring sufficient training iterations and appropriate structure), thereby resolving the contradiction between training time and control accuracy.
Data Source
AI summary
An intelligent control method for a dynamic neural network-based variable cycle engine is provided. By adding a grey relation analysis method-based structure adjustment algorithm to the neural network training algorithm, the neural network structure is adjusted, a dynamic neural network controller is constructed, and thus the intelligent control of the variable cycle engine is realized. A dynamic neural network is trained through the grey relation analysis method-based network structure adjustment algorithm designed by the present invention, and an intelligent controller of the dynamic neural network-based variable cycle engine is constructed. Thus, the problem of coupling between nonlinear multiple variables caused by the increase of control variables of the variable cycle engine and the problem that the traditional control method relies too much on model accuracy are effectively solved.