Full-field imaging learning machine for structural dynamics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for measuring and characterizing structural dynamics are inadequate, as they rely on discrete-point contact-type sensing which provides low spatial resolution and is affected by environmental conditions, and non-contact methods like laser-based vibrometers are time-consuming and impractical for processing high-resolution data, especially for nonlinear structures.
Innovation Solution
A computer-implemented method using a deep complexity coding artificial neural network and autoencoder to analyze spatio-temporal image data from cameras, decomposing it into modal components and determining dynamic properties of structures with high spatial resolution, enabling full-field pixel-level analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If discrete-point contact-type sensing methods (accelerometers, strain gauges) are used, then measurement capability is provided, but spatial resolution is low and environmental conditions affect accuracy
Solution Approach 1:
The patent replaces mechanical contact-type sensing systems (accelerometers, strain gauges) with optical imaging systems. Instead of using physical sensors attached to the structure, the invention uses cameras to capture visual data and processes this data through image processing algorithms to extract dynamic properties, thereby eliminating the limitations of discrete sensing points and their susceptibility to environmental conditions.
Solution Approach 2:
The patent creates a virtual copy of the physical structure through image data. By capturing images of the structure and processing these visual copies through algorithms, the system can analyze dynamic properties without physically contacting the structure. This allows full-field analysis across the entire structure surface, providing high spatial resolution without the complexity of numerous physical sensors.
2Reliability
If laser-based vibrometers are used, then non-contact measurement is achieved, but processing time increases significantly
Solution Approach 1:
The patent merges multiple imaging data streams into a unified processing system. By combining information from multiple camera captures and integrating them through parallel processing algorithms, the system achieves comprehensive full-field analysis while reducing overall processing time. The merged data is processed simultaneously across multiple computational pathways to extract dynamic properties efficiently.
Solution Approach 2:
The patent performs preliminary data processing and feature extraction during the image capture phase. By pre-processing the imaging data to identify and isolate relevant dynamic features before final analysis, the system reduces the computational burden during subsequent processing stages, thereby decreasing total processing time while maintaining non-contact measurement capability.
3Measurement precision
If video-based methods with high spatial resolution are used, then full-field analysis is achieved, but processing becomes impractical due to data volume
Solution Approach 1:
The patent extracts and isolates the most critical dynamic information from the large volume of imaging data. By using algorithms to identify and separate significant dynamic features from redundant data, the system processes only the essential information needed for dynamic property analysis. This extraction approach maintains high pixel-level spatial resolution while dramatically reducing the effective data volume that requires processing.
Solution Approach 2:
The patent applies partial processing to the most informative regions of the data first. Instead of uniformly processing all data at full resolution, the system prioritizes processing of areas with significant dynamic activity or pre-identified features, achieving adequate analysis results with reduced computational resources and faster processing speeds.
Data Source
AI summary
A method of determining dynamic properties of a structure (linear or nonlinear) includes receiving spatio-temporal inputs, generating mode shapes and modal components corresponding to the spatio-temporal inputs using a trained deep complexity coding artificial neural network, and subsequently generating the dynamic properties by analyzing each modal component using a trained learning machine. A computing system for non-contact determination of dynamic properties of a structure includes a camera, a processor, and a memory including computer-executable instructions. When the instructions are executed, the system is caused to receive spatio-temporal image data, decompose the spatio-temporal image data into constituent manifold components using an autoencoder, and analyze the constituent manifold components using a trained learning machine to determine the dynamic properties.