Correlation Tolerance Limit Setting Using Repetitive Cross-Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for correlation optimization and tolerance limit setting in reactor core states are inadequate in preventing intentional or unintentional distortion of material properties and fail to quantify the influence of such distortions, leading to increased costs and limitations in risk management.
Innovation Solution
A correlation tolerance limit setting system using repetitive cross-validation, which involves randomly classifying data into training and validation sets, optimizing coefficients for selected correlations, performing normality tests, and determining the departure from nucleate boiling ratio (DNBR) limits using parametric or nonparametric methods to prevent and quantify distortions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If correlation optimization is performed on separated datasets with limited cases, then the process is simpler and faster, but the risk of intentional or unintentional distortion of material properties cannot be prevented or quantified
Solution Approach 1:
The patent segments the validation process into multiple independent cross-validation cases, each with its own training and validation sets. This segmentation allows comprehensive risk assessment across different data partitions while maintaining systematic control over the correlation optimization process, preventing distortion through diverse validation perspectives.
Solution Approach 2:
The patent merges results from multiple cross-validation cases to establish comprehensive tolerance limits. By combining findings across different segmented validations, the system achieves both thorough risk prevention and quantification while maintaining operational efficiency through structured aggregation of validation results.
2Device complexity
If independent test datasets with same or similar design characteristics are operated separately, then individual tolerance limits can be set through simple statistical analysis, but the influence of material property distortion cannot be quantified and additional testing costs increase
Solution Approach 1:
The patent implements feedback mechanisms where validation results from each cross-validation case inform the overall tolerance limit setting. This feedback loop ensures that distortion influences are detected and quantified across multiple validations, preventing information loss while maintaining manageable analysis complexity through structured feedback aggregation.
Solution Approach 2:
The patent creates a universal cross-validation framework that can handle multiple test datasets with same or similar design characteristics. This multi-functional approach allows individual tolerance limit setting while simultaneously quantifying distortion influences across all datasets, eliminating the need for separate analyses and reducing overall complexity.
3Reliability
If repetitive cross-validation is performed with multiple cases, then risk prevention and distortion quantification are achieved, but the computational process becomes more complex and time-consuming
Solution Approach 1:
The patent performs preliminary actions by pre-segmenting data into multiple cross-validation cases and pre-defining validation protocols. This preliminary organization enables parallel processing of different validation cases, reducing overall validation time while maintaining comprehensive risk prevention and distortion quantification capabilities.
Solution Approach 2:
The patent employs periodic cross-validation actions where multiple validation cases are systematically executed in cycles. This periodic approach distributes computational workload over time, preventing excessive time concentration in single validation phases while ensuring thorough risk assessment through repeated systematic validations.
Data Source
AI summary
A correlation tolerance limit setting system using repetitive cross-validation includes: a variable extraction unit randomly classifying data of an initial DB set into training set data and validation set data at a specific rate and then extracting variables for determining a DNBR limit by optimizing coefficients of a selected correlation; a normality test unit testing normality for a variable extraction result; a DNBR limit unit determining whether data sets have a same population or not depending on normality result and determining DNBR limit from a distribution of 95/95 DNBR; and a controller.