Hybrid Neural Network Arrhenius Model Material Aging Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining the aging behavior of materials, such as the relative temperature index (RTI), are limited to predicting the 50% loss of mechanical properties and require extensive measurement data across various parameter ranges, which is time-consuming and inefficient, especially for safety-critical applications.
Innovation Solution
A hybrid method combining neural networks with the Arrhenius model processes parameters to predict aging behavior for a wide range of values, including higher or lower percentages of mechanical resilience without extensive real measurements, using a hybrid neural network and Arrhenius model to simplify and refine the determination process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a pure neural network is used to predict aging behavior, then the model can learn complex patterns from data, but it cannot extrapolate to parameter ranges for which it was not trained
Solution Approach 1:
The patent combines a neural network with an Arrhenius model into a hybrid system. The neural network processes input parameters (temperature, time, material properties) and outputs predictions that are then refined by the Arrhenius model, which provides physical-based extrapolation capability. This merging allows the system to maintain high prediction accuracy while extending to untrained parameter ranges.
Solution Approach 2:
The Arrhenius model acts as an intermediary between the neural network and the final prediction. The neural network's output serves as input to the Arrhenius model, which then adjusts the prediction based on temperature-accelerated aging relationships. This intermediary enables physically consistent extrapolation while leveraging the pattern recognition capabilities of the neural network.
2Measurement precision
If extensive measurement data across various parameter ranges is collected, then the prediction accuracy improves, but the time and resources required for testing increase significantly
Solution Approach 1:
The patent performs preliminary training of the neural network using a limited set of measurement data at various temperatures and times. Once trained, the hybrid model can predict aging behavior for untested parameter ranges without requiring additional physical testing. This preliminary action with minimal data enables subsequent rapid predictions.
Solution Approach 2:
The Arrhenius model enables parameter transformation by using temperature-accelerated aging relationships to extrapolate from short-term high-temperature tests to long-term low-temperature behavior. This allows prediction of aging over years based on tests conducted over days or weeks at elevated temperatures.
3Device complexity
If the prediction is limited to 50% loss of mechanical properties, then the methodology is simpler, but it cannot provide information for safety-critical applications requiring higher or lower thresholds
Solution Approach 1:
The hybrid neural network-Arrhenius model is designed to predict relative tensile strength (RTS) for any threshold value, not just 50% loss. The model can output predictions for 10%, 20%, 30%, or any other RTS threshold, making it universally applicable to different safety criteria and application requirements while maintaining a single unified methodology.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables accurate and efficient prediction of material aging behavior for various parameter ranges, including higher or lower mechanical property losses, reducing the need for extensive measurement data and speeding up the calculation process, while maintaining high accuracy.
Implementation Method 1
processing the result in an Arrhenius model to obtain a predicted value
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
The method involves feeding two sets of physical parameters indicating storage temperature and storage time of a test specimen, into the respective neural networks (400,402) for performing a simulation. The simulation results including different coefficients and/or terms and/or variables are obtained and supplied to an Arrhenius model (404). A prediction value which is not fed to the neural networks, is obtained from the simulation results. Independent claims are included for the following: (1) a computer program product for determining the aging behavior of a material; and (2) a system for determining the aging behavior of a material.