Root Cause Identification Using Cross-Correlation Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current diagnostic methods in process plants are limited in identifying the root cause of variability without prior knowledge or process models, and they require costly and time-consuming steps, making it difficult to determine the source of process variations across complex systems.
Innovation Solution
A method that uses cross-correlation analysis of time-series data to identify the most likely root cause of variation by evaluating correlations and lead/lag times between a primary reference variable and other variables, eliminating the need for a-priori process knowledge or models, and presenting results in a user-friendly graphical format.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If cross-correlation analysis is used to identify root cause without prior process knowledge or models, then the need for extensive mathematical knowledge and model development is reduced, but the complexity of analyzing multiple time-series variables increases
Solution Approach 1:
The patent replaces complex mathematical modeling and expert knowledge (mechanical/intellectual system) with automated cross-correlation analysis (computational system). The method uses algorithmic processing of time-series data to identify root causes without requiring manual model development or extensive mathematical knowledge from users.
Solution Approach 2:
The system performs self-diagnosis by automatically analyzing process data and identifying root causes without requiring external expert intervention. The automated cross-correlation analysis enables the system to serve itself in diagnosing process variations, eliminating the need for manual model development and expert analysis.
2Measurement precision
If traditional diagnostic methods with process models are used, then the accuracy of root cause identification is improved, but the time and cost required for model development increases
Solution Approach 1:
The patent performs preliminary cross-correlation analysis on historical data to establish relationships between variables before actual root cause diagnosis is needed. This pre-computed correlation information is stored and can be quickly applied when root cause analysis is required, eliminating the need for time-consuming model development at the time of diagnosis.
Solution Approach 2:
The system creates simplified correlation models from historical data that replicate the behavior of complex process models. These correlation-based models capture the essential relationships between variables without requiring detailed process knowledge, providing accurate root cause identification with minimal development time.
3Reliability
If comprehensive process data is analyzed to determine root cause, then the reliability of diagnosis is improved, but the quantity of data to be processed increases
Solution Approach 1:
The patent extracts only the most relevant correlations from the comprehensive process data by identifying the strongest cross-correlations between the reference variable and other process variables. This extraction of key relationships maintains diagnostic reliability while reducing the effective data volume that needs to be analyzed in real-time.
Solution Approach 2:
The analysis focuses on local relationships between specific variable pairs rather than processing all possible combinations of process data. By concentrating computational effort on the most significant correlations (those with highest cross-correlation coefficients), the system maintains high diagnostic reliability while processing a manageable quantity of data.
Data Source
AI summary
An exemplary embodiment includes a diagnostic which can identify the source, or “root cause” of variability of process and process control parameters. A plurality of correlations is provided, each representing a possible cause of variation. One of the correlations is identified as the most likely root cause of variation. The remaining possible root causes are also listed, in sequence, from most likely to least likely. The method applies to both normal and abnormal operating conditions.