Aero-Thermo Engine Power Control With Real-Time Thrust Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fluid-based engineering control systems face computational inefficiencies and lack accuracy in real-time control, particularly in monitoring and controlling power output, due to limitations in prior engine parameter on-board synthesis (EPOS) models.
Innovation Solution
A control system that includes an actuator and a model processor generating an estimated thrust value using an open loop model based on dynamic states and mathematical abstractions of physical laws, with a control law directing the actuator to adjust the control device and account for errors such as wear and fuel quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If piecewise linear state variable representations are used for control systems, then device complexity is reduced, but measurement precision and control accuracy deteriorate
Solution Approach 1:
The patent transforms the control approach by changing from piecewise linear parameter representations to a neural network-based continuous non-linear parameter mapping. The neural network learns optimal parameter transformations during training, enabling accurate estimation of fluid-based engineering parameters without requiring complex segmented models.
Solution Approach 2:
The patent replaces traditional mechanical/mathematical control system structures (piecewise linear models, iterative solvers) with an intelligent software-based neural network system. This substitution eliminates the need for complex segmented mathematical models while maintaining or improving accuracy through learned patterns from training data.
2Measurement precision
If stationary simulations are deployed in real-time environment, then measurement precision improves, but productivity and reliability deteriorate due to large model size, iterative solvers, and high maintenance cost
Solution Approach 1:
The patent performs the computationally intensive work in advance by training the neural network offline using stationary simulation data and iterative solvers. Once trained, the network contains pre-learned parameter mappings that can be rapidly evaluated in real-time without requiring iterative computation during actual control operations.
Solution Approach 2:
The patent creates a simplified copy of the complex stationary simulation model by training a neural network to replicate its input-output behavior. The neural network serves as a lightweight surrogate model that mimics the accurate but computationally expensive stationary simulation, enabling real-time operation with reduced model size and no iterative solvers.
3Measurement precision
If direct measurements of system parameters are implemented, then measurement precision improves, but device complexity and cost increase due to additional sensors and measurement equipment
Solution Approach 1:
The patent introduces a neural network as an intermediary computational layer that infers difficult-to-measure parameters (such as thrust or power output) from easily measurable quantities (pressures, temperatures, flow rates). This intermediary model acts as a virtual sensor, providing accurate parameter estimates without requiring physical direct measurement devices.
Data Source
AI summary
Systems and methods for controlling a fluid-based system are disclosed. The systems and methods may include generating a model output using a model processor, processing a model input vector and setting a model operating mode, and setting dynamic states of the model processor, the dynamic states input to an open loop model based on the model operating mode. Synthesized parameters are generated as a function of the dynamic states and the model input vector based on a series of utilities, where at least one of the utilities is a configurable utility including one or more sub-utilities. An estimated state of the model is determined based on at least one of a prior state and the synthesized parameters. An actuator associated with the control device is directed as a function of a model output, where the model output includes an estimated thrust value for the control device.