Wind Turbine Component Monitoring via Kinematic Vibration Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine learning models for fault detection in wind turbines require extensive and expensive training data, which are typically generated by inducing faults in components, and are limited to a specific turbine type, making them ineffective for monitoring different turbine types.
Innovation Solution
A method that maps vibration signals from a monitored wind turbine to another type using kinematic parameters, enabling a pre-trained machine learning model to detect faults in components by converting signals from one turbine to another, utilizing supervised or unsupervised learning and neural networks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are trained using fault data from test stands or actual faults, then the model can detect faults accurately, but the training data generation becomes expensive and time-consuming
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models on simulated fault data before deployment. Virtual training data is generated in advance through simulation environments, allowing the model to be pre-trained without requiring actual fault occurrences or expensive test stand operations. This resolves the contradiction by preparing the detection capability beforehand using synthetic data, eliminating the need for time-consuming actual fault data collection while maintaining detection accuracy.
2Measurement precision
If machine learning models are trained on one turbine type, then they achieve good performance for that specific type, but they cannot be effectively used for other turbine types
Solution Approach 1:
The patent implements universality by developing a transfer learning framework that enables a single machine learning model to adapt across multiple turbine types. The model is pre-trained on simulated data that captures universal fault patterns, then can be transferred to different turbine types through domain adaptation techniques. This resolves the contradiction by making the model universally applicable to various turbine types while maintaining fault detection accuracy through simulation-based pre-training that generalizes across platforms.
Solution Approach 2:
The patent applies parameter changes by adjusting model parameters and training conditions when transferring between different turbine types. The simulation environment generates training data with varying parameters representing different turbine configurations, allowing the model to learn parameter-invariant fault detection patterns. This enables the model to adapt to different turbine types by modifying its parameters based on the target turbine characteristics while maintaining detection performance.
3Reliability
If extensive training data including actual fault data is collected, then the machine learning model achieves high detection performance, but the cost and complexity of data collection increases significantly
Solution Approach 1:
The patent applies copying by creating virtual copies of fault data through simulation rather than collecting actual fault data from physical systems. The simulation environment generates synthetic training datasets that replicate real fault conditions without requiring actual faults to occur or expensive test stands. This resolves the contradiction by copying fault scenarios virtually, maintaining detection reliability while eliminating the complexity and cost of physical data collection infrastructure.
Data Source
AI summary
Provided is a method for computer-implemented monitoring of a component of a wind turbine, having access to a trained machine learning model which has been trained for one or more components of the same type of wind turbines. The trained machine learning model is configured to provide an output referring to a predetermined fault occurring at a component of a wind turbine by processing vibration signals in a predetermined domain which are measured in the vicinity of the component during the operation of the wind turbine. Vibration signals are mapped to corresponding vibration signals valid for the component based on one or more given kinematic parameters of the component and one or more given kinematic parameters of another component. The machine learning model is applied to the vibration signals valid for the component, resulting in an output referring to the predetermined fault occurring at the another component.