Causal Response Monitoring for Rapid Abnormality Isolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional statistical and machine learning techniques face challenges in accurately and promptly tracing the cause of system abnormalities in complex social infrastructure systems, such as train maintenance, which hinders effective predictive maintenance.
Innovation Solution
A monitoring system that utilizes a causal structure model to estimate the response variable of a monitoring-target system based on measurement data, determining system abnormalities by comparing estimated and actual values, and identifying potential causes through a directional graph structure indicating causal relations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional statistical and machine learning techniques are used to estimate system abnormalities, then the system can process large amounts of data, but the accuracy and speed of cause tracing is insufficient
Solution Approach 1:
The patent segments the complex system into multiple subsystems and further into individual components, creating a hierarchical structure. This segmentation allows the causal structure model to focus on specific cause-effect relationships within each component rather than analyzing the entire system at once, thereby improving both accuracy and speed of abnormality detection
Solution Approach 2:
The patent introduces a causal structure model as an intermediary between raw sensor data and abnormality detection. This model acts as a mediator that encodes prior knowledge about cause-effect relationships, enabling faster and more accurate identification of abnormal causes by filtering and interpreting data through the lens of known causal mechanisms
2Reliability
If conventional monitoring methods are used, then the system structure is simple, but the ability to accurately identify root causes is insufficient
Solution Approach 1:
The patent performs preliminary action by pre-establishing the causal structure model before actual monitoring begins. This model encodes domain knowledge about cause-effect relationships in advance, allowing the system to quickly identify root causes during operation without requiring complex real-time analysis of all possible relationships
Solution Approach 2:
The patent changes the parameter representation from raw sensor values to causal relationships. By transforming the monitoring approach to focus on cause-effect parameter changes rather than individual sensor readings, the system achieves higher reliability in root cause identification while managing complexity through structured parameter transformation
Data Source
AI summary
A monitoring system that monitors a monitoring-target system is disclosed. The monitoring system includes one or more storage apparatuses that store a program, and one or more processors that operate according to the program. The one or more processors determine an estimated value of a monitoring-target response variable of the monitoring-target system on a basis of measurement data included in test data of the monitoring-target system and a causal structure model of the monitoring-target system. The one or more processors decide whether an abnormality has occurred in the monitoring-target system on a basis of a result of a comparison between a measurement value of the monitoring-target response variable included in the test data, and the estimated value.