Neural Network Model for Physical Quantity Level Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for solving partial differential equations (PDEs) with spatial constraints, such as those involving physical obstacles, require significant computational resources and are not adaptable to new environments, limiting their reuse and efficiency.
Innovation Solution
A learning method that uses a masking and correction function sequence, where a neural network determines a model for a physical quantity's spatio-temporal evolution by representing physical obstacles as spatial constraints, allowing the model to approximate solutions under different spatial conditions without additional learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If numerical methods are used to solve PDEs with spatial constraints, then solution accuracy is improved, but computational resources and memory requirements increase significantly
Solution Approach 1:
The patent creates a digital copy of the physical environment (spatial constraints) and uses it to train a neural network model. Once trained, the model can predict physical quantity levels in new environments without requiring complex numerical simulations, thus achieving high accuracy while reducing computational resource requirements.
Solution Approach 2:
The patent replaces traditional numerical simulation methods (finite difference, finite element, or spectral methods) with a machine learning approach using neural networks. This substitution eliminates the need for complex discretization and numerical computation while maintaining solution accuracy.
2Measurement precision
If models are trained for specific spatial constraints, then prediction accuracy for those constraints is improved, but adaptability to new environments decreases
Solution Approach 1:
The patent trains a universal neural network model that can handle multiple types of spatial constraints and environmental configurations. The model learns general patterns from diverse training data representing different spatial configurations, enabling it to accurately predict physical quantity levels in new, unseen environments without requiring retraining or model adjustment.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
A computer-implemented method for determining a level of a physical quantity with spatio-temporal evolution in the presence of physical obstacles in any area, comprising: - in a learning phase (100), determination (110), by means of machine learning receiving as input a first set of physical obstacles (11) and a first set of data (12), of a model (13) for said physical quantity in the predefined area; - in an exploitation phase (200), determination (210) of a second level (24) of the physical quantity in any area, from the model (13) for said physical quantity receiving as input a second set of physical obstacles (21), distinct from the first set of physical obstacles (11), and a second set of data (22).