Technical System Validation Using Component ML Deviation Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Complex technical systems with intricately linked components are difficult to validate or verify due to their stochastic and complex nature, making it hard to predict their behavior and ensure desired outcomes, especially in safety-critical systems where extensive real-world data collection is impractical.
Innovation Solution
A method involving obtaining models for system components, training machine learning models with validation measurements, propagating test inputs through these models to determine deviations, and assessing the probability of fulfilling desired criteria by offsetting deviations, allowing for component-level validation without extensive real-world data collection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If extensive real-world data collection is performed to validate complex technical systems, then validation accuracy is improved, but time consumption and resource requirements increase significantly
Solution Approach 1:
The patent applies preliminary action by training machine learning models in advance on component-level validation measurements before actual system validation. These pre-trained models capture the stochastic behavior patterns of individual components, enabling accurate system-level validation predictions without requiring extensive real-world system data collection during the validation phase itself.
Solution Approach 2:
The patent segments the complex technical system into individual components, obtaining validation measurements for each component separately. Machine learning models are trained on these segmented component data independently, and then the models are combined to validate the entire system. This segmentation allows validation to proceed with much less data than would be required for the complete system.
2Productivity
If component-level validation is performed without extensive real-world data, then validation speed is improved, but reliability of validation results may deteriorate
Solution Approach 1:
The patent introduces machine learning models as intermediaries between component-level validation measurements and system-level validation conclusions. These models learn the stochastic behavior patterns from component data and use this knowledge to predict system behavior, bridging the gap between limited component data and reliable system validation without requiring extensive real-world system data.
Solution Approach 2:
The patent changes the parameter representation by transforming raw component validation measurements into learned parameters through machine learning model training. The models extract meaningful behavioral patterns and parameters from component data, which then serve as reliable inputs for system-level validation, maintaining validation reliability even with limited data.
3Quantity of substance
If machine learning models are trained on component validation measurements, then the need for real-world system data is reduced, but model training complexity increases
Solution Approach 1:
The patent applies partial action by training machine learning models only on the specific component validation measurements that are available, rather than requiring complete system data. The models are trained on partial data sets from individual components, which is sufficient to capture the essential stochastic behavior patterns needed for system-level validation.
Data Source
AI summary
A method for verifying and/or validating whether a technical system fulfills a desired criterion. The technical system emits output signals based on input signals supplied to the technical system. The method includes: obtaining models for a plurality of components of the technical system; obtaining a plurality of validation measurements; for each component, training a machine learning model to predict outputs of the respective component; obtaining first test outputs from a last model based on test input; determining, second test outputs from the machine learning model corresponding to the last model and based on the test inputs of the models; determine a deviation which characterizes a difference between first test outputs determined from the last model and second test outputs determined by the machine learning model corresponding to the last model; verifying and/or validating whether the technical system fulfills the criterion.