Wind Turbine Anomaly Detection Using CNN Scatter Plot Diagnostics
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for anomaly detection in industrial assets, such as wind turbines, face limitations in accurately identifying root causes of abnormal sensor measurements, relying on manual diagnostic processes that are uncertain and limited to simple outlier patterns.
Innovation Solution
A deep learning model leveraging a diagnostic expert domain knowledgebase and convolutional neural networks is developed to automatically detect anomalies and identify root causes by analyzing time series sensor measurements, using scatter plots to recognize patterns and classify root causes, with a feedback loop for continuous model improvement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual diagnostic processes are used to identify root causes of abnormal sensor measurements, then the system can detect anomalies, but the accuracy and reliability of root cause identification deteriorates due to limitations in distinguishing complex outlier patterns
Solution Approach 1:
The patent replaces manual diagnostic processes with an automated machine learning system that uses unsupervised learning algorithms to detect and classify outlier patterns. The system automatically identifies root causes by analyzing sensor measurements without human intervention, thereby improving reliability while handling complex patterns that manual methods cannot distinguish.
Solution Approach 2:
The patent introduces an intermediary machine learning model that acts as a bridge between raw sensor measurements and root cause identification. This intermediary system processes complex outlier patterns and translates them into actionable diagnostic information, resolving the contradiction between manual detection limitations and the need for accurate root cause identification.
2Productivity
If automated deep learning models are implemented for anomaly detection and root cause identification, then productivity and accuracy improve, but device complexity increases
Solution Approach 1:
The patent segments the diagnostic system into distinct functional modules: data collection from sensors, preprocessing of sensor measurements, anomaly detection using unsupervised learning, root cause classification, and feedback mechanisms. This segmentation allows the complex deep learning model to be managed as separate, manageable components, reducing overall system complexity while maintaining high productivity.
Solution Approach 2:
The patent implements feedback loops where the system continuously monitors its own performance and uses detected anomalies to refine its diagnostic capabilities. This feedback mechanism allows the system to improve accuracy over time without requiring proportional increases in model complexity, as the learning process adapts to patterns in the data.
Data Source
Figure 1~2
Figure 3~4
Figure 5A~5B
AI summary
A system and method including receiving historical time series sensor data associated with operation of an industrial asset; generating visual representation images of scatter plots based on the historical time series sensor data based on a reference to a digitized knowledge domain associated with the industrial asset; assigning a root cause label to each image; generating a convolutional neural network (CNN) model trained and tested using subsets of the labeled images; and processing, by the CNN model, a real-time image to detect at least one anomaly in the real-time image and one or more root causes associated with the at least one anomaly.