December 7, 2024

CoE Site Index  |  
WESEP: Wind Energy Science, Engineering, and Policy

Reliability and Health Monitoring

Thrust Goal and Connection to Project Objectives

The primary goals of this thrust are to develop new sensing approaches to monitor the health of the wind turbine structures and components, to utilize the data to support improved designs and manufacturing processes, and to develop life cycle management strategies based on reliability analyses. This thrust will address RO2 and RO3 by enhancing our understanding of the factors that affect turbine reliability, which is critical to improving the cost-effective operation of land-based turbines and is essential for their movement to offshore installations, thus aiding to expand penetration limits.

Background and Need for Research

As the population of utility scale wind turbines increases, usage and failure data become more prevalent, pointing to gearbox reliability as a critical issue. Blade reliability is also of concern as evidenced by reports presented in a series of DOE workshops. Traditional reliability field data have consisted of failure times for units that failed and running times for units that had not failed. These data are used for tasks such as predicting warranty and maintenance costs and planning for future capital expenditures. Due to changes in technology, the next generation of reliability field data will provide information that is considerably richer. The most important differences between carefully controlled laboratory accelerated test experiments and field reliability results are due to uncontrolled field variation (unit-to-unit and temporal) in variables such as use rate, load, vibration, temperature, humidity, UV intensity, and UV spectrum. Use rates, load and environmental conditions are important sources of variability in product lifetimes. Historically, use rate, load, and other environmental data have not been available to reliability analysts. Today it is possible to install sensors and smart chips in a product to measure and record use rate and/or environmental data and provide this information in real time. This is already happening in wind turbine generators. Research is needed to properly use such data.

Lifetime models that use rate and/or environmental data have potential to explain much more variability in field data than has been possible before. The information can also be used to predict individual unit. In addition to the time series use-rate/load/environmental data, turbines can be outfitted with sensors that provide information, at the same rate, on degradation or indicators of imminent failure (e.g., a broken tooth on a gear will change a unit’s vibration signal). Depending on the application, such information is also called “system health” and “materials state” information. Such system data from wind turbines can be returned in real time to a central location for real time process monitoring and especially for prognostic purposes. An appropriate signal in these data might provoke rapid action to avoid a serious system failure (e.g., by shutting down a turbine with an initial failure before serious damage is done to the system). Also, should some issue relating to system health arise at a later date, it will be possible to sort through historical data to identify detectable signals useful in providing early warning.

Dissertation Project Examples

Sensor optimization: Current generation systems include a limited number of sensors, primarily on gearbox components. Additional sensors that better characterize the gearbox as well as other components (blade, tower, nacelle) are needed to correlate loads, degradation and performance data with wind conditions. This involves innovations to create new classes of sensors and the development of a physics-based understanding of the relationship of the sensor output to the load, degradation and performance conditions of interest.

Reliability analysis: Studies within this rubric will evaluate strategies for reliability analysis of health monitoring data from the blade, drive train/gearbox, and the tower/foundation. Correlations between health monitoring data with meteorological and energy production data will be sought with the goal of improving the mechanistic understanding of the relationship of data to damage evolution.

Life cycle management strategies: While wind power generation has a long history, recent growth in the industry implies the development of new systems and early warranty operation. As systems age, attention turns to life extension strategies and detection of wear-out issues and associated damage. Inspection and life management strategies will be developed for key components such as blades, gearboxes and towers.

Prediction of component residual life based on use history: For purposes of medium and long-term budgeting, companies need to be able to predict the remaining life of system components. By modeling the relationship between a component’s use history (e.g., load, stresses, amount of use), it is possible to build predictive models that are much more accurate than those that have been based solely on traditional field reliability data that lacks such information. Depending on the component, response data may come in the form of failure times or some measure of degradation (either physical or performance degradation). The resulting models will be able to generate prediction intervals for individual components (e.g., turbine blades or gear-boxes) as well as for the aggregate number of components failing in future periods of time.

Hierarchical Bayesian model for component lifetimes: A company with wind turbines from different “fleets” (different manufacturers, designs, or locations) will have reliability data characterizing each fleet. Because of differences among the fleets (e.g., model number or geographical location), it is not possible to simply pool these data to make reliability inferences. Stratifying the data will generally be an inefficient way to extract information. A carefully constructed Bayesian hierarchal model would allow the “borrowing of strength” across a large group of fleets, allowing available data to be used in the most efficient manner. The Bayesian approach to combining information from different sources also provides a convenient way to integrate non-exact physics and engineering information into the reliability model.

Key Faculty

  • Bruce Babcock, Professor – Economics
  • Lisa Brasche, Associate Director – Center for Nondestructive Evaluation
  • William Meeker, Distinguished Professor – Statistics
  • Frank Peters, Associate Professor – Industrial and Manufacturing Systems Engineering