Equipment Abnormality Diagnosis Using Time-Series Feature Change
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing abnormality diagnosis methods require specialized knowledge and skills, and are time-consuming, especially when identifying the cause and mode of equipment abnormalities, and struggle with accurately diagnosing rare abnormalities.
Innovation Solution
An abnormality diagnosis method that acquires multivariate time-series data, extracts features before and after an abnormality using feature extraction methods, calculates the amount of change in features, and uses a database to diagnose the abnormality cause and mode without requiring specialized knowledge, by generating a change amount vector and comparing it with registered vectors in a database.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If simulation based on physical model is used to identify abnormality cause and mode, then diagnostic accuracy is improved, but specialized knowledge and skills are required and diagnosis time increases
Solution Approach 1:
The system pre-calculates and stores diagnostic results for various abnormality conditions in a database before actual diagnosis occurs. When an abnormality is detected, the system compares current sensor data against pre-computed diagnostic patterns, enabling rapid diagnosis without requiring real-time physical model simulation or specialized knowledge.
Solution Approach 2:
Instead of performing complex physical model simulations during actual diagnosis, the system creates copies of diagnostic knowledge by pre-storing simulation results and diagnostic patterns in a database. This allows the system to retrieve and match diagnostic patterns rapidly without repeating the computationally intensive simulation process.
2Reliability
If physical model simulation is used to identify abnormality cause, then diagnostic capability is improved, but device complexity and operational difficulty increase
Solution Approach 1:
The system performs self-diagnosis by automatically comparing sensor data against pre-stored diagnostic patterns in the database. This eliminates the need for operators to manually perform complex physical model simulations or interpret sophisticated diagnostic tools, making the system easy to operate while maintaining high diagnostic reliability.
Solution Approach 2:
The system replaces complex mechanical or manual diagnostic processes (physical model simulation requiring specialized knowledge) with an automated information processing system that retrieves and compares diagnostic patterns from a database, significantly reducing system complexity and operational difficulty.
3Productivity
If database based on actual events is used to identify abnormality mode, then diagnostic speed is improved, but database enrichment is difficult for rare abnormalities
Solution Approach 1:
The system pre-computes and stores diagnostic patterns for various abnormality conditions including rare abnormalities in the database, even before they occur in actual operation. This allows rapid diagnosis of rare abnormalities when they occur, as the diagnostic patterns are already prepared and stored, eliminating the need to wait for multiple occurrences to enrich the database.
Data Source
AI summary
An abnormality diagnosis method for diagnosing an abnormality in equipment includes acquiring multivariate time-series data for a plurality of measurement items from the equipment, diagnosing an abnormality in operational state of the equipment based on the multivariate time-series data, and diagnosing a cause of the abnormality. The diagnosing a cause of the abnormality includes extracting a feature of a first section before the occurrence of the abnormality from the multivariate time-series data of the first section, extracting a feature of a second section after the occurrence of the abnormality from the multivariate time-series data of the second section, obtaining an amount of change in feature from a difference between the feature of the first section and the feature of the second section, and diagnosing a measurement item that is the cause of the abnormality based on the amounts of change in features of the plurality of measurement items.