Wind Turbine Rotor Blade Anomaly Detection via Adaptive Neural Network

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional methods for monitoring wind turbine rotor blades face challenges due to incomplete or non-existent data on damage patterns, as the structure of rotor blades is constantly evolving and changing, making effective damage detection difficult.

Innovation Solution

A method and device utilizing an adaptive algorithm, such as a neural network, that learns the normal state of a wind turbine through sensor measurements and detects anomalies without pre-existing damage patterns, allowing for real-time identification of changes in the rotor blade's condition, such as changes in natural frequency, without the need for predefined damage knowledge.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional methods use detailed databases with damage patterns for monitoring rotor blades, then damage detection capability is improved, but the method becomes inapplicable due to incomplete or non-existent damage pattern data

Engineering Contradiction:
Improvedamage detection capabilityVSAvoiddamage pattern data availability
Core Design Contradiction:
ReliabilityVSLoss of information

Solution Approach 1:

Instead of comparing measurements against known damage patterns (conventional approach), the patent inverts the approach by training the neural network on normal, undamaged states and detecting deviations from this baseline. This allows damage detection without requiring damage pattern databases, resolving the contradiction between detection capability and data availability.

Inventive Principle:
Principle #13The other way round (Inversion)

Solution Approach 2:

The patent performs preliminary training of the neural network on normal operating states before actual damage detection is needed. This preliminary action establishes a baseline model of healthy rotor blade behavior, enabling subsequent anomaly detection without requiring pre-existing damage pattern information.

Inventive Principle:
Principle #10Preliminary action

2Adaptability or versatility

If an adaptive algorithm is trained on normal state data to detect anomalies, then the need for damage pattern databases is eliminated, but the system requires extensive normal state training data and computational resources

Engineering Contradiction:
Improvedamage detection without damage patternsVSAvoidtraining data requirements and computational resources
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The neural network performs self-training by learning normal rotor blade behavior patterns from operational data without requiring external damage pattern databases. The system serves itself by automatically adapting to the specific characteristics of each rotor blade through unsupervised learning on normal states.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent transforms the monitoring approach by changing from pattern-matching parameters to statistical deviation parameters. Instead of searching for specific damage signatures, the system monitors deviations in vibration frequencies, amplitudes, and other operational parameters from the learned normal state, reducing dependency on extensive training databases.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3513066A1Method and device for monitoring a status of at least one wind turbine and computer program product
Publication Date: 2019.07.24 VC VIII POLYTECH HLDG APS

AI summary

The invention relates to a method (200) for monitoring a status of at least one wind turbine. The method (200) comprises: detecting first measurement signals via one or more sensors (210), wherein the first measurement signals provide one or more parameters relating to at least one rotor blade of the at least one wind turbine in a normal status; training a trainable algorithm based on the first measurement signals of the normal status (220); detecting second measurement signals via the one or more sensors (230); and recognising an undetermined anomaly via the trainable algorithm trained in the normal status, if a current status of the wind turbine, determined based on the second measurement signals, deviates from the normal status (240).