Industrial Fault Localization Without a Dynamic Process Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fault localization methods in industrial processes face challenges when a dynamic system model is not available, particularly in older plants with limited measurement data, requiring additional sensors and extensive data training, which complicates quick diagnosis.
Innovation Solution
A method and system that generate a linearized model of the industrial plant using structural and sensor data, without needing a dynamic model, additional sensors, or complex learning procedures, by creating a look-up table based on structural plant data to detect and localize faults.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If model-based methods are used for fault detection and isolation, then fault localization accuracy is improved, but the requirement of a dynamic model of the industrial plant increases system complexity
Solution Approach 1:
The patent creates a simplified copy of the system model - a static structural model that captures the essential topology and relationships without requiring complex dynamic equations. This structural model serves as a lightweight alternative to full dynamic models, maintaining fault localization capability while reducing model complexity requirements
Solution Approach 2:
The patent extracts only the essential structural information from the complex dynamic system model - specifically the topological relationships, component connections, and structural parameters. By taking out only the necessary structural elements and discarding complex dynamic equations, the system achieves fault localization without requiring full dynamic models
2Ease of manufacture
If data-driven methods are used for fault detection and isolation, then additional sensor installations are avoided, but the requirement for extensive data for training increases system complexity
Solution Approach 1:
The patent performs preliminary structural analysis during the system design phase to establish the structural model and identify potential fault patterns before actual operation. By pre-processing the structural information and creating the structural model in advance, the system eliminates the need for extensive online data collection and training, reducing operational complexity
Solution Approach 2:
The system uses the existing structural information and available sensor data to automatically build and update its own structural model without requiring external training datasets. The structural model self-adapts to the specific plant configuration, eliminating the need for extensive external data training while utilizing only existing sensors
3Measurement precision
If extensive data collection is performed for fault diagnosis in older plants, then fault detection accuracy is improved, but the time required for diagnosis increases
Solution Approach 1:
The patent segments the fault diagnosis process into two distinct phases: an offline structural model building phase and an online fault detection phase. By dividing the workload, complex structural analysis is performed once offline using available data, while online operation requires only simple parameter comparison against the pre-built model, dramatically reducing diagnosis time while maintaining accuracy
Solution Approach 2:
The patent performs comprehensive structural analysis and model building in advance during offline operation. By preparing the structural model, fault patterns, and reference parameters beforehand, the system eliminates the need for extensive real-time data collection during actual fault events, enabling rapid diagnosis when faults occur
Data Source
AI summary
A method and system for localizing faults in an industrial process is proposed. The industrial process includes a plurality of components. The method includes receiving structural plant data from an industrial plant. A structured model of the process is generated from the structural plant data. Sensor data measuring characteristics of the plurality of components is also received. Parameters of the structured model are identified from the received sensor data and stored. Faults are detected during operation of the industrial plant utilizing the identified parameters and detecting changes in the parameters by comparing current parameters to stored parameters. The fault information is then displayed via a display to an operator.