Turbine Ice Accretion Detection With Image-Based De-Icing Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for detecting ice accumulation on machines like turbines are inefficient, often relying on theoretical models that are not device-specific, leading to unnecessary de-icing events and decreased performance, and lack real-time accuracy in monitoring and corrective action.
Innovation Solution
A deviation monitoring system (DMS) that uses optical cameras and machine learning models to analyze images and sensor data from turbines, autonomously detecting and classifying ice accumulation, and activating de-icing systems only when necessary, independent of device models or operating geometries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If theoretical curves are used for ice detection, then the system can operate independently of device model, but the detection accuracy decreases and unnecessary de-icing events occur
Solution Approach 1:
The system creates a digital twin or virtual model of the specific turbine component geometry and uses ray tracing to simulate ice accumulation patterns on that exact geometry. This copying approach allows the system to achieve both device-specific accuracy and model independence by generating custom simulations for any turbine design without requiring physical prototypes.
Solution Approach 2:
The system changes the parameters of the simulation by adjusting environmental conditions (temperature, humidity, wind speed, precipitation rate) and ice formation parameters to match real-time operating conditions. This allows the theoretical model to adapt to specific device geometries and operating conditions, resolving the contradiction between model independence and detection accuracy.
2Reliability
If conservative icing boundaries are used, then equipment damage is prevented, but user power output and performance decrease
Solution Approach 1:
The system implements continuous feedback by monitoring real-time environmental conditions, comparing simulated ice accumulation against actual sensor data, and dynamically adjusting de-icing activation thresholds. This feedback loop allows the system to maintain equipment protection while optimizing power output by avoiding unnecessary de-icing events when ice accumulation is below critical levels.
Solution Approach 2:
The system transitions from static conservative boundaries to dynamic, condition-based thresholds that adapt to real-time operating conditions. By continuously updating the icing boundaries based on current environmental parameters and device state, the system maintains reliability while maximizing productivity by allowing operation closer to the true icing limit.
3Reliability
If de-icing systems are activated frequently, then ice damage is prevented, but energy consumption increases and performance decreases
Solution Approach 1:
The system performs preliminary simulations to predict ice accumulation trends and activates de-icing only when simulations indicate approaching critical thresholds. This preliminary action approach allows the system to prepare for potential icing events while avoiding premature de-icing activation, thereby reducing energy consumption while maintaining equipment protection.
4Speed
If current image analysis methods are used, then real-time detection is achieved, but the system lacks accuracy in classifying material accumulation
Solution Approach 1:
The system introduces an intermediary layer of physics-based simulation between image capture and classification. The ray tracing simulation acts as a mediator that translates raw images into meaningful ice accumulation measurements by comparing them against simulated ice patterns under various environmental conditions, thereby improving classification accuracy while maintaining real-time performance.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system for monitoring at least one component is provided. The system includes at least one processor in communication with at least one memory device. The at least one processor is programmed to store a plurality of baseline information associated with the at least one component to be monitored and receive a plurality of current images of the at least one component to be monitored. The at least one processor is also programmed to detect a deviation from baseline based upon a comparison of the plurality of current images and the plurality of baseline information. The at least one processor is further programmed to classify the deviation and implement a corrective action based on the classification of deviation.