DCS Operation Sequence Clustering Using Intent-Based Event Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing event analysis systems struggle to cluster operation procedures of DCS operators at a high level of abstraction, making it difficult to standardize manual operations due to variations in external conditions and operator skills, and emphasizing small differences in operation events rather than overall flow.

Innovation Solution

An event analysis system that extracts operation intentions and methods from event logs, defines basic and unit operations, and clusters operation sequences based on similarity using weighted edit distance to consider plant operation features, allowing for the abstraction of operation sequences and visualization of cluster analysis results.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If event sequences are clustered based on the number of events that need to be removed or added to change one sequence to another, then the similarity between event sequences can be measured, but small differences in the number of operation events are too emphasized, making it difficult to cluster operation procedures at a high level of abstraction

Engineering Contradiction:
Improvesimilarity measurement precisionVSAvoidabstraction level of operation sequences
Core Design Contradiction:
Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The patent segments the event sequence into multiple levels of abstraction: basic unit operations (individual events), unit operations (grouped events with common characteristics), and operation sequences (ordered series of unit operations). This hierarchical segmentation allows clustering to focus on the overall flow and structure rather than minor variations in individual events, resolving the contradiction between measurement precision and abstraction level.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the parameters used for similarity measurement from raw event counts to abstracted operation characteristics. By transforming the measurement parameters to operate at higher levels of abstraction (unit operations and operation sequences rather than individual events), the system can measure similarity without being overly sensitive to small differences in event numbers, thus enabling high-level clustering.

Inventive Principle:
Principle #35Parameter changes

2Loss of information

If manual analysis is performed on operation event sequences from each work period, then detailed understanding of specific operations can be achieved, but the manual analysis work becomes difficult when the work is carried out many times

Engineering Contradiction:
Improveoperation detail understandingVSAvoidmanual analysis time
Core Design Contradiction:
Loss of informationVSLoss of time

Solution Approach 1:

The patent performs preliminary automated preprocessing of operation event sequences by extracting basic unit operations, grouping them into unit operations, and organizing them into operation sequences. This preliminary action prepares the data in advance for clustering, so that when analysis is needed, the system can quickly process pre-organized sequences rather than manually analyzing raw event data each time, reducing time loss while preserving operation details.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system performs self-service automated analysis of operation sequences, automatically extracting patterns and clustering operations without requiring manual intervention for each work period. The automated extraction and clustering processes serve themselves, eliminating the need for repeated manual analysis while maintaining detailed operation understanding through the structured representation of unit operations and sequences.

Inventive Principle:
Principle #25Self-service

3Quantity of substance

If event logs are collected and stored in a database, then operation histories can be recorded for analysis, but the data processing and analysis become complex when dealing with large volumes of event data from multiple work periods

Engineering Contradiction:
Improveevent log data volumeVSAvoiddata processing complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent segments large volumes of event log data into manageable units through hierarchical organization: individual events are grouped into basic unit operations, which are further grouped into unit operations, and finally organized into operation sequences. This segmentation reduces processing complexity by breaking down the large data volume into structured, hierarchical components that can be processed more efficiently.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the data representation parameters from raw event logs to abstracted operation sequences. By transforming the data parameters to a higher level of abstraction, the system reduces the complexity of handling large volumes of detailed event data while preserving the essential operational information needed for analysis and clustering.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3165982B1An event analysis apparatus, an event analysis method, and an event analysis program
Publication Date: 2021.07.14 YOKOGAWA ELECTRIC CORP
  • EP3165982B1 patent drawingFigure 1
  • EP3165982B1 patent drawingFigure 2
  • EP3165982B1 patent drawingFigure 3

AI summary

An event analysis device includes an event collector configured to collect event log data which represents operation history of a DCS operator; an event analyzer configured to analyze event log data collected by the event collector to extract analysis result data representing operation method based on the operation of the DCS operator or analysis result data representing operation intention based on the operation of the DCS operator; an operation sequence extractor configured to extract analysis result data respectively in every period in specified multiple periods and extract operation sequence sorting analysis result data based on operation time; and an operation sequence clusterer configured to cluster the operation sequence based on the degree of the similarity between the operation sequences which are different from each other.