Dynamic Uncertain Causality Graph Variable Ranking for Abnormality Diagnosis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for diagnosing system abnormalities in industrial, social, and biological systems are inefficient in identifying the root cause quickly and cost-effectively due to the complexity of causal relationships and the need to detect multiple variables, leading to suboptimal detection of state-unknown X-type variables in Dynamic Uncertain Causality Graphs (DUCG).
Innovation Solution
A method to rank state-unknown X-type variables optimally using a CPU, allowing for sequential or parallel detection of their states to update evidence and improve diagnosis accuracy and cost-effectiveness, by calculating importance scores based on connectivity, concern, and conditional probabilities within the simplified DUCG framework.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all state-unknown X-type variables are detected to improve diagnosis accuracy, then the completeness of evidence increases, but the detection cost and time increase significantly
Solution Approach 1:
The system automatically calculates importance scores for each state-unknown variable based on the DUCG structure and current evidence, generating its own detection priority list without external intervention. This self-service mechanism optimizes the detection sequence to achieve accurate diagnosis with minimal detection operations.
Solution Approach 2:
The patent transforms the static DUCG into a dynamic model where variable importance scores are recalculated as evidence updates. This parameter change approach allows the detection priority to adapt dynamically, ensuring that the most informative variables are detected first while minimizing overall detection cost and time.
2Productivity
If more X-type variables are detected to reduce detection cost, then the efficiency improves, but the complexity of variable selection and prioritization increases
Solution Approach 1:
The patent replaces manual or heuristic variable selection methods with an automated computational system that calculates importance scores based on DUCG structure and evidence. This substitution eliminates the complexity of manual variable prioritization while maximizing detection efficiency through algorithmic optimization.
3Speed
If detection is performed on high-importance variables first to improve diagnosis speed, then the time to identify root cause decreases, but the risk of missing critical low-ranked variables increases
Solution Approach 1:
The system implements feedback by continuously updating the DUCG with detected evidence and recalculating importance scores for remaining undetected variables. This feedback loop ensures that the detection sequence remains optimal throughout the diagnosis process, maintaining both speed and reliability by dynamically adjusting priorities based on accumulated evidence.
Data Source
AI summary
Provided is a method for ordering X-type variables having states to be measured in a dynamic uncertain causality graph (DUCG). The method comprises: step 1: determining, on the basis of a DUCG simplified using E(y) as a condition, a measurable X-type variable having a state to be measured, an index set thereof being S x(y), where y is a time series; step 2: if there is only one element in S x(y), ending ordering; step 3: calculating ranking importance I i(y) for X i(i∈S x(y)); step 4: ranking X i(i∈S x(y)) according to the ranking importance I i(y), and performing state measurement on an X-type variable of i∈S x(y) by referring to the ranking; and step 5: adding 1 to y, and repeating steps 1-5 until there is no X-type variable to be measured. The technical solution of the invention can quickly diagnose a cause of an object system abnormality at minimum cost, and effectively returns the object system to normal.