Optical Dispersion Measurement Using Neural Network Parameter Fitting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional optical measurement schemes for precision components like gratings require extensive computations and are time-consuming, with database size limitations restricting the measurement of geometric parameters with varying ranges, especially in hundreds of nanometers.
Innovation Solution
A method utilizing a neural network model to generate simulated dispersion curve data based on coordinate data and preset geometric parameters, with gradient optimization to update parameters and recalculate distances, enabling rapid and accurate measurement by reducing computational load and database reliance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional optical measurement schemes use simulation algorithms with analog computations for all parameters in the spatial range of the geometric model, then measurement accuracy can be maintained, but the computational load increases significantly and measurement time becomes very long
Solution Approach 1:
The patent pre-establishes a database containing optical responses for various geometric models and their parameters before actual measurement. This preliminary computation and storage of reference data eliminates the need for real-time analog computations during measurement, significantly reducing measurement time while maintaining accuracy through comparison with pre-computed reference data
Solution Approach 2:
The patent creates a database that copies and stores simulated optical responses for different geometric models and parameter ranges. During measurement, the system compares actual optical responses against these pre-stored copies rather than performing new computations, thereby reducing computational load and measurement time while preserving measurement accuracy
2Adaptability or versatility
If the geometric model parameters vary in a range of hundreds of nanometers to cover wide measurement ranges, then measurement versatility improves, but the database size increases exponentially making the solution intractable
Solution Approach 1:
The patent divides the wide parameter range into multiple discrete levels or segments (e.g., coarse, medium, fine levels). The database stores optical responses at these segmented parameter levels rather than continuously across the entire range. During measurement, the system performs hierarchical searches starting from coarse levels and progressively refines to finer levels, reducing database size while maintaining measurement versatility through multi-level parameter coverage
Solution Approach 2:
The patent introduces a hierarchical level dimension to organize the database structure. Instead of storing all possible parameter combinations in a flat structure, the data is organized across multiple levels of parameter discretization. This dimensional reorganization allows the system to handle wide parameter ranges efficiently by navigating through levels rather than searching through an exponentially growing flat database
3Measurement precision
If conventional schemes perform analog computations for all parameters in the spatial range of the geometric model, then complete parameter coverage is achieved, but the computational complexity becomes very high
Solution Approach 1:
The patent pre-computes and stores optical responses for various geometric models and parameter combinations in a database before actual measurement. This shifts the computational complexity from the measurement phase to the database preparation phase, making the actual measurement process computationally simple while maintaining complete parameter coverage through the pre-stored reference data
Solution Approach 2:
The patent creates a comprehensive copy of the parameter space and its corresponding optical responses in the database. During measurement, the system performs simple pattern matching against these pre-computed copies rather than executing complex analog computations, thereby reducing computational complexity while achieving complete parameter measurement coverage
Data Source
AI summary
A method comprises: generating input data for inputting to a neural network model based on coordinate data of dispersion curve data and preset values of geometric parameters with respect to a reference model; generating simulated dispersion curve data associated with the preset values of the geometric parameters based on the neural network model trained via a plurality of samples; obtaining measured dispersion curve data in relation to an object calculating a distance of the measured dispersion curve data from the simulated dispersion curve data to determine whether the distance meets a predetermined condition; and in response to determining the distance does not meet the predetermined condition, determining a gradient for updating the preset values of the geometric parameters with respect to the reference model based on the distance, to regenerate simulated dispersion curve data via the neural network model based on the updated preset values to recalculate the distance.