Anomalous Data Detection for Gas Turbine Baseline Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Incorrectly established baseline operating conditions due to anomalies in initialization data can lead to unreliable comparisons, hindering the assessment of a device's health and performance.
Innovation Solution
A diagnostic system for gas turbine engines that receives initialization data, detects and removes anomalies such as outlying data points and bimodality, and establishes a cleaned data set to accurately determine baseline operating conditions, using methods like slope calculation and T-like statistic analysis to identify and correct for trends and bimodalities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If initialization data is used directly to establish baseline operating conditions, then the baseline can be established quickly, but the baseline accuracy deteriorates due to anomalies in the initialization data
Solution Approach 1:
The patent applies preliminary action by detecting and removing anomalies from initialization data before establishing the baseline operating conditions. The system performs anomaly detection on temperature, pressure, and flow rate data during the initialization phase, removes identified anomalies, and then uses the cleaned data to establish accurate baselines. This preliminary cleaning process ensures that the baseline accuracy is not compromised by anomalous readings while maintaining a reasonable initialization timeline.
2Measurement precision
If anomaly detection and removal processes are applied to initialization data, then the baseline accuracy is improved, but the system complexity increases
Solution Approach 1:
The patent applies local quality by implementing anomaly detection specifically for critical parameters (temperature, pressure, flow rate) rather than analyzing all possible data points. The system focuses computational resources on detecting anomalies in these key operational parameters using targeted algorithms, which reduces the overall data processing complexity while still significantly improving baseline accuracy for the most important operational metrics.
Data Source
Figure 1~2
Figure 3
AI summary
An example data assessment method for a diagnostic system includes receiving a set of initialization data (110), analyzing the set of initialization data (120) to identify a data outlier in the set of initialization data, and determining whether the set of initialization data is bimodal. The method further includes establishing a set of cleaned data (130) based on the analysis and establishing a baseline operating condition (140) for the device using the set of cleaned data.