Process Instrument Noise Correlation for Multi-Fault Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing fault detection methods in process plants rely on statistical techniques that can ignore slow drifting measurements, average gross errors with good data, and fail to detect multiple faults, leading to inaccurate precision and missed fault detections due to their probabilistic nature.
Innovation Solution
A system utilizing Artificial Intelligence (AI) based data analysis techniques, such as Time Series Analysis, to correlate noise from measuring instruments with reference noise, identify deviations, and detect faults like sensor malfunctions, drifts, and leakage by correlating noise and parameter deviations with predefined thresholds.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If statistical techniques are used for fault detection, then the system can process data with historical context, but slow drifting measurements are ignored and averaged with good data, reducing detection accuracy
Solution Approach 1:
The patent extracts and isolates noise components from sensor signals using signal processing techniques. By separating the noise portion from the useful signal, the system can analyze noise characteristics independently to detect faults without the noise being averaged with good measurements. This extraction principle allows the system to identify subtle fault patterns that would otherwise be masked by statistical averaging.
Solution Approach 2:
Instead of analyzing the main signal for faults, the patent inverts the approach by analyzing the noise component itself for fault detection. The system correlates noise signals across different sensors to identify faults, reversing the conventional wisdom that noise should be filtered out. This inversion enables detection of faults that statistical techniques on main signals would miss.
2Reliability
If gross error detection techniques are used, then outliers can be identified, but multiple faults in measuring instruments and process equipment might not be detected due to probabilistic nature
Solution Approach 1:
The patent merges noise signals from multiple sensors through correlation analysis. By combining information from noise signals across different measurement points, the system can detect multiple simultaneous faults that would be difficult to identify using individual sensor analysis. The correlation of noise patterns provides a unified approach to detect various fault types including sensor malfunctions, drifts, and process equipment issues.
Solution Approach 2:
The patent uses noise correlation as an intermediary mechanism to detect faults. Rather than directly analyzing sensor readings for faults, the system uses the correlation properties of noise signals as a mediator to indirectly detect fault conditions. This intermediary approach allows the system to identify multiple faults by analyzing how noise patterns relate across different sensors, providing versatility in fault detection.
3Ease of operation
If averaged measurements containing gross errors are used in reconciliation, then data processing is simplified, but fault detections are missed and precision of reconciled data is affected
Solution Approach 1:
The patent performs preliminary noise analysis and correlation before conducting data reconciliation. By pre-processing the signals to extract and correlate noise components, the system identifies potential faults beforehand. This preliminary action allows the reconciliation process to use cleaned, validated data, improving the precision of reconciled data while maintaining processing simplicity through automated preprocessing steps.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
The present invention relates to a method and a system for production accounting in process industries using Artificial Intelligence (Al). More particularly the present invention relates to fault detection in a plurality of measuring instruments (102) and process equipment in a process plant. A plurality of measured signals from the measuring instruments (102) is received by the process control system and noise is extracted from the plurality of measured signals. The extracted noise is correlated with a noise extracted from a plurality of reference signals using an Al based data analysis technique. Further, the process control system identifies deviations in the one or more parameters. The process control system detects the faults the plurality of measuring instruments (102) or the process equipment using the correlated noises and the identified deviations of the one or more parameters.