Clustered Process Data Models for Scalable Sensor Fault Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current data cleansing systems for large-scale industrial processes, such as oil production facilities, face challenges in scalability and effectiveness due to reliance on variable correlations, which can lead to false alarms and inefficiencies in detecting and correcting sensor errors and anomalies.
Innovation Solution
An automated model building and maintenance system that performs clustering on data sources to form groups, builds data models based on training data, and updates these models using recent operational data, allowing for data cleansing and anomaly detection across multiple data sources, including isolated instruments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If data driven models including PCA or PLS are used to monitor process statistics to detect sensor failures, then fault detection capability is improved, but scalability to large-scale data collection systems deteriorates
Solution Approach 1:
The patent segments the large-scale data collection system into multiple clusters, each monitored by its own data driven model (PCA or PLS). This segmentation allows the system to maintain fault detection capability for each cluster while reducing the overall complexity and improving scalability, as each model only needs to process a subset of the total data rather than all data simultaneously.
2Measurement precision
If Kalman filter interpolation methods are used to detect outliers and reconstruct missing data streams, then data reconstruction accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies Kalman filter interpolation methods to reconstruct missing data streams and detect outliers before the main fault detection and analysis processes. By performing this data reconstruction in advance, the system improves measurement precision for subsequent operations while managing computational complexity through pre-processing rather than real-time computation during critical detection phases.
3Measurement precision
If fault detection systems rely on relationships between variables for fault detection, then detection accuracy for correlated sensors is improved, but effectiveness for isolated instruments deteriorates
Solution Approach 1:
The patent implements a universal fault detection approach that works for both correlated sensors and isolated instruments. For correlated sensors, it uses relationship-based detection methods, while for isolated instruments without significant correlations, it applies alternative detection methods that do not depend on variable relationships. This multi-functional approach ensures the system maintains effectiveness across different instrument types and correlation scenarios.
Data Source
AI summary
Methods and systems for building and maintaining model(s) of a physical process are disclosed. One method includes receiving training data associated with a plurality of different data sources, and performing a clustering process to form one or more clusters. For each of the one or more clusters, the method includes building a data model based on the training data associated with the data sources in the cluster, automatically performing a data cleansing process on operational data based on the data model, and automatically updating the data model based on updated training data that is received as operational data. For data sources excluded from the clusters, automatic building, data cleansing, and updating of models can also be applied.