Cross-Facility Process Standards With Objective Outlier Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for removing outlier data in data-driven model development are subjective and prone to bias, affecting the fairness and representativeness of analysis results, particularly in the computation of greenhouse gas emissions standards.
Innovation Solution
A data-driven approach using a computer-implemented method to objectively remove outliers by iteratively generating predicted values, error sets, and error thresholds, adjusting model coefficients to minimize errors, and terminating based on performance criteria.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If subjective outlier removal methods are used, then the process is simpler to implement, but the accuracy and fairness of data analysis deteriorates due to introduced bias
Solution Approach 1:
The patent replaces subjective manual outlier removal with an automated statistical system using iterative algorithms (e.g., Huber M-estimation, Least Absolute Deviations) that objectively identify and remove outliers based on mathematical criteria rather than human judgment, thereby eliminating bias while maintaining implementation simplicity through computer automation
Solution Approach 2:
The patent changes the approach from fixed threshold outlier removal to dynamic parameter adjustment where statistical parameters (such as robust standard deviation, breakdown points) are iteratively refined to optimally distinguish outliers from valid data, improving accuracy without requiring complex manual intervention
2Reliability
If iterative outlier removal with model re-fitting is performed, then the reliability of data-driven models improves, but the complexity of the analysis process increases
Solution Approach 1:
The patent implements dynamic iterative processes where the model is repeatedly re-fitted after each round of outlier removal, allowing the system to adaptively converge to optimal results. The complexity is managed through automated iteration control and convergence criteria that prevent infinite loops while ensuring thorough analysis
Solution Approach 2:
The patent incorporates feedback mechanisms where model performance metrics are continuously evaluated during iteration, and outlier removal decisions are adjusted based on how well the model explains the remaining data. This feedback loop ensures reliability improvement while maintaining process manageability through automated performance monitoring
3Measurement precision
If strict data removal guidelines are applied, then data quality improves, but the quantity of available data for analysis decreases
Solution Approach 1:
The patent uses adjustable statistical parameters (such as robust standard deviation multiples, breakdown point thresholds) that can be tuned to balance data quality improvement against data quantity retention. These parameters allow flexible control over how strictly outliers are removed, enabling optimization based on specific analysis needs
Solution Approach 2:
The patent applies partial outlier removal where not all potentially problematic data points are eliminated, but rather a controlled portion is removed based on statistical significance. This partial action approach improves data quality sufficiently without excessively reducing the available data quantity for meaningful analysis
Data Source
AI summary
A system and method is described herein for performing at least one industrial process at each facility of a plurality of facilities based on an industrial process standard generated by reducing functional, and trend line outlier bias in data of one or more process parameters as measured by one or more sensors. Outliers are removed from the data set through an objective method. Bias is determined based on absolute, relative error, or both. Error values are computed from the data, model coefficients, or trend line estimates. Outlier data records are removed when the error values are greater than or equal to one or more criteria.