Fault Detection Using Performance Indicators for Correlated Sensors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing anomaly detection systems in industrial environments face challenges due to the large number of possible states generated by multiple correlated sensors, making it computationally expensive and time-consuming to detect faults effectively.
Innovation Solution
The system employs principles like mass balance and energy balance to establish relationships between real-time device measurements and optimal system operation, allowing for the calculation of performance indicators to detect anomalies and identify faulty devices.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional anomaly detection methods observe each device within a system, then detection accuracy is improved, but computational cost and time consumption increase significantly
Solution Approach 1:
The patent extracts and focuses only on the critical performance indicator that directly reflects system health, rather than monitoring all device parameters. By identifying and tracking this single key indicator, the system achieves effective anomaly detection without the computational burden of analyzing every sensor reading across the entire system.
Solution Approach 2:
The patent segments the complex system monitoring task into two parts: (1) using relationships to predict expected behavior of the performance indicator, and (2) comparing actual measurements against these predictions. This segmentation allows efficient anomaly detection by focusing computational resources only on the critical indicator rather than the entire system state.
2Measurement precision
If conventional anomaly detection methods observe each device within a system, then detection accuracy is improved, but system complexity increases
Solution Approach 1:
The patent extracts and focuses only on the critical performance indicator that directly reflects system health, rather than monitoring all device parameters. By identifying and tracking this single key indicator, the system achieves effective anomaly detection without the computational burden of analyzing every sensor reading across the entire system.
Solution Approach 2:
The patent segments the complex system monitoring task into two parts: (1) using relationships to predict expected behavior of the performance indicator, and (2) comparing actual measurements against these predictions. This segmentation allows efficient anomaly detection by focusing computational resources only on the critical indicator rather than the entire system state.
3Measurement precision
If machine learning models are trained on all possible sensor states, then detection accuracy is improved, but training cost and data requirements increase exponentially
Solution Approach 1:
The patent extracts and focuses only on the critical performance indicator that directly reflects system health, rather than monitoring all device parameters. By identifying and tracking this single key indicator, the system achieves effective anomaly detection without the computational burden of analyzing every sensor reading across the entire system.
Solution Approach 2:
The patent segments the complex system monitoring task into two parts: (1) using relationships to predict expected behavior of the performance indicator, and (2) comparing actual measurements against these predictions. This segmentation allows efficient anomaly detection by focusing computational resources only on the critical indicator rather than the entire system state.
Data Source
AI summary
Methods, systems, and computer program products for automated fault monitoring and detection in a system. The fault detection method comprising receiving real-time device measurement data from a plurality of input and/or output devices in the system; identifying at least one relationship that correlates the real-time device measurement data to a set of parameters that define operation of the system; calculating a performance indicator for the system based on the real-time device measurement data using the identified at least one relationship; and if the performance indicator is outside a predetermined range, then an anomaly is detected in the system. The fault detection method further comprises: identifying a faulty device among the plurality of input and/or output devices; and reporting the identified faulty device.