Event Analyzer Bayesian Network Causal Inference
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing event analyzers in plants can only perform simple statistical processing, making it difficult to determine the causal relationships among events, especially considering the occurrence-time differences, which hinders the prediction of subsequent events and efficient plant operation.
Innovation Solution
An event analyzer that constructs a Bayesian network using an event matrix to analyze the causal relationships among device events, taking into account the occurrence-time differences, by improving the K2-algorithm to account for time-lags and using a distributed database for flexible scalability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If simple statistical processing is performed on events, then the analysis process is simple and fast, but the ability to determine causal relationships among events is insufficient
Solution Approach 1:
The patent introduces Bayesian network theory as an intermediary framework to analyze event relationships. The event analysis module uses Bayesian probability calculations to infer causal relationships from event sequences, serving as a mediator between simple statistical processing and complex causal analysis, thereby resolving the contradiction between analysis speed and causal relationship detection capability
Solution Approach 2:
The patent changes the analysis parameter from simple event frequency statistics to Bayesian probability values that represent causal relationships. By calculating posterior probabilities of event sequences and introducing time-lag parameters, the system transforms the nature of analysis parameters to enable causal relationship determination while maintaining computational efficiency
2Measurement precision
If causal relationship analysis considering time-lags is performed, then the prediction accuracy of subsequent events is improved, but the device complexity increases
Solution Approach 1:
The patent segments the complex causal analysis process into distinct functional modules: event sequence determination module, time-lag determination module, and event analysis module. Each module handles a specific aspect of the analysis, breaking down the complex task of time-lagged causal relationship analysis into manageable segments that can be processed efficiently
Solution Approach 2:
The system implements feedback mechanisms where event sequences and time-lag information are fed into the Bayesian network model, which then outputs probability assessments that can be used to predict future events. This feedback loop allows the system to continuously refine its predictions based on observed event patterns and time-lag relationships
3Device complexity
If users manually determine relationships among events, then no complex analysis system is needed, but the efficiency of grasping causal relationships is low
Solution Approach 1:
The patent implements self-service functionality where the event analysis system automatically determines causal relationships among events without requiring manual user intervention. The system autonomously processes event data, calculates Bayesian probabilities, and generates causal relationship assessments, thereby eliminating the need for users to manually analyze event relationships while significantly improving analysis efficiency
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An event analyzer includes an event log collection module configured to collect an event log, an event log storage module configured to convert the event log into an event matrix and to store the event matrix, an event occurrence-order determination module configured to determine occurrenee-order of device events based on the event matrix, a Bayesian network generation module configured to specify device events serving as parent node candidates, to calculate evaluation values corresponding to each specified parent node candidate at a virtual device event obtained by shifting the device event serving as a processing object each unit time, and to select an device event serving as a parent node together with a shift amount, and an analysis module configured to output an device event estimated as a cause of an device event to be analyzed, or an device event estimated to occur later based on the shift amount.