Training Data Generation for Electromagnetic Wave Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for estimating electromagnetic wave intensity in electronic circuits, such as physical simulations and machine learning models, face challenges in accuracy due to the lack of consideration for return currents, which are crucial for predicting electromagnetic interference effectively.
Innovation Solution
A method for generating training data that includes information on return currents by specifying patterns of adjacent layers with slits, allowing the model to account for bypass routes and improve the accuracy of electromagnetic wave estimation using machine learning, specifically by generating circuit data that represents virtual wiring to accurately depict return paths.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If physical simulation is used to estimate electromagnetic wave intensity, then estimation accuracy is improved, but computational time and complexity increase significantly
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing return current path information during the training phase. The system pre-processes circuit board data to identify return current paths and stores this information in advance, so that during actual electromagnetic wave intensity estimation, the model can directly utilize this pre-computed information without performing time-consuming physical simulations, thus resolving the contradiction between accuracy and computational time.
2Productivity
If traditional machine learning models are used for electromagnetic wave estimation, then computational speed is improved, but estimation accuracy deteriorates due to lack of return current information
Solution Approach 1:
The patent introduces an intermediary component - a return current path extraction module that processes circuit board data to identify and extract return current paths. This extracted information serves as an intermediary input to the machine learning model, enabling the model to incorporate return current effects without sacrificing estimation speed. The intermediary module bridges the gap between traditional fast models and the need for accurate return current consideration.
3Measurement precision
If return current information is incorporated into training data, then electromagnetic interference prediction accuracy is improved, but data processing complexity increases
Solution Approach 1:
The patent applies the extraction principle by isolating and extracting only the essential return current path information from the complete circuit board data. Instead of processing all circuit board details, the system extracts specifically the return current path characteristics and uses only this extracted information for training the machine learning model. This selective extraction reduces data processing complexity while maintaining the accuracy benefits of incorporating return current effects.
Data Source
AI summary
An information processing apparatus specifies a first pattern indicating a first layer included in first circuit data. The information processing apparatus generates, based on first wiring included in a second pattern indicating a second layer that is adjacent to the first layer and a slit included in the first pattern, second circuit data by changing the first pattern to a third pattern including second wiring corresponding to the first wiring. The information processing apparatus generates, based on the second circuit data, training data for machine learning.