Digital Twin State Estimation With Physics-Constrained Neural Operators
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in designing a state observer that can integrate with black-box simulation environments, such as digital twins, and effectively use them for controlling mechanical systems like vapor compression systems.
Innovation Solution
A system and method that estimate the internal states of a digital twin using a neural network trained to satisfy constraints derived from the physics of the mechanical system, employing an autoencoder to estimate unconstrained states and a neural operator to enforce constraints.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a neural network is trained using loss functions to enforce internal states to satisfy physics constraints, then constraint satisfaction is improved, but the training only guarantees satisfaction with some probability governed by statistical nature
Solution Approach 1:
The patent introduces a neural operator as an intermediary component between the autoencoder and the constraint requirements. This neural operator learns to map unconstrained internal states to constrained states that satisfy physics constraints, providing deterministic constraint satisfaction while maintaining the statistical advantages of neural network training
Solution Approach 2:
The patent segments the state estimation task into two distinct components: an autoencoder that estimates unconstrained internal states and a neural operator that enforces physics constraints. This segmentation allows each component to specialize in its function while working together to achieve both statistical efficiency and deterministic constraint satisfaction
2Reliability
If an analytical module is designed to enforce satisfaction of constraints, then constraint satisfaction is improved, but it may go against statistical guarantees and computational advantages of neural networks
Solution Approach 1:
The patent replaces traditional analytical constraint enforcement methods with a neural operator that learns constraint satisfaction through training. This substitution maintains computational efficiency and statistical advantages of neural networks while achieving deterministic constraint satisfaction, avoiding the computational burden of analytical methods
3Ease of operation
If digital twin modules are made black-box for protection of proprietary information and reduction of code complexity, then ease of operation is improved, but the digital twin may not admit a simple model structure for classical model-based state observer design
Solution Approach 1:
The patent creates a simplified copy or surrogate model of the black-box digital twin that captures essential dynamics for state estimation purposes. This copying approach allows classical model-based state observer design to proceed with a simpler model structure while the original black-box digital twin remains intact for proprietary protection and complex simulations
Data Source
AI summary
To perform a task based on an internal state of a digital twin simulating an operation of a mechanical system, where at least some state variables of the internal states of the digital twin are subject to constraints derived from the physics of a structure of the mechanical system, a processor executes a neural network including an autoencoder trained to process a current internal state of the digital twin and a current control input to the mechanical system to produce a current output of the mechanical system caused by the current control input and an unconstrained next internal state of the digital twin transitioned from the current internal state based on the current control input, and a neural operator trained to modify the unconstrained next internal state of the digital twin to produce a constrained next internal state of the digital twin satisfying the constraints.