Heuristic Model for Structural Deformation Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for identifying structural deformation, such as those in aircraft antenna systems, lack accuracy and are complex, time-consuming, and expensive, making it difficult to compensate for deformation in real-time and maintain performance within selected tolerances.
Innovation Solution
A method and apparatus using a heuristic model trained with deformation and strain data from multiple training cases to estimate deformation with high accuracy, allowing for real-time adjustment of control parameters to compensate for structural deformation and maintain performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If optical systems, imaging systems, fiber optic systems, CMM systems, cameras are used to identify deformation, then measurement capability is provided, but measurement precision and device complexity are insufficient
Solution Approach 1:
The patent replaces complex optical and mechanical measurement systems with a computational approach using a trained neural network model that processes strain gauge data. This substitution eliminates the need for sophisticated optical infrastructure while achieving superior measurement precision through machine learning algorithms that analyze strain patterns to predict deformation with high accuracy.
Solution Approach 2:
The patent introduces simple strain gauge sensors as intermediary elements that convert physical deformation into electrical signals. These sensors serve as mediators between the structural deformation and the computational model, providing raw data that the neural network processes to generate precise deformation measurements without requiring direct optical or mechanical measurement of the structure.
2Reliability
If complex measurement systems are deployed, then deformation can be detected, but time consumption and cost increase
Solution Approach 1:
The patent performs preliminary training of the neural network model using historical deformation and strain data before actual deployment. This pre-training phase establishes the model's predictive capabilities in advance, enabling rapid real-time deformation assessment during operation without requiring time-consuming measurements or complex system setup during critical periods.
Solution Approach 2:
The system uses readily available strain gauge data that is already being collected by the platform's existing sensor infrastructure. By leveraging this self-generated data and processing it through the trained model, the system eliminates the need for external measurement equipment and reduces time consumption while maintaining reliable deformation identification for performance maintenance.
3Measurement precision
If currently available systems are used for deformation identification, then some measurement is provided, but accuracy and cost-effectiveness are insufficient
Solution Approach 1:
The patent employs inexpensive strain gauge sensors instead of costly optical or CMM systems. These simple, low-cost sensors provide the necessary raw data that the neural network transforms into high-precision deformation measurements. The approach trades the high initial cost of sophisticated measurement equipment for affordable sensors combined with computational intelligence, achieving superior accuracy at lower implementation cost.
Solution Approach 2:
The patent transforms the approach to deformation measurement by changing from direct physical measurement parameters (optical coordinates, mechanical dimensions) to electrical strain parameters that are processed computationally. This parameter transformation enables the use of simple sensors with complex data processing, achieving high measurement precision through algorithmic analysis rather than through expensive measurement hardware.
Data Source
AI summary
A method and apparatus for identifying deformation of a structure. Training deformation data is identified for each training case in a plurality of training cases. Training strain data is identified for each training case in the plurality of training cases. The training deformation data and the training strain data are configured for use by a heuristic model to increase an accuracy of output data generated by the heuristic model. A group of parameters for the heuristic model is adjusted using the training deformation data and the training strain data for the each training case in the plurality of training cases such that the heuristic model is trained to generate estimated deformation data for the structure based on input strain data. The estimated deformation data has a desired level of accuracy.