Development of reliability program for risk assessment of composite structures treating defects as uncertainty variables
Riddle, William Wilson, III.
MetadataShow full item record
It has been reported that a leading cause of repairs and failures in wind turbine blades is attributable to manufacturing defects. The size, weight, shape and economic considerations of wind turbine blades have dictated the use of low cost composite materials. Composite structure manufacturing quality is a critical issue for reliability. While significant research has been performed to better understand what is needed to improve blade reliability, a comprehensive study to characterize and understand the manufacturing flaws commonly found in blades has not been performed. The work presented herein is focused on performing mechanical testing of flawed composite specimen and developing probabilistic models to assess the reliability of a wind blade with defects. The analysis postulates that one should assess defects as a design parameter in a parametric probabilistic analysis. A consistent framework has been established and validated for quantitative categorization and analysis of flaws. Results from this effort have shown that the probability of failure of a wind turbine blade with defects, can be adequately described through the use of Monte Carlo simulation. The two approaches detailed in this analysis have shown that, by treating defects as random variables, one can reduce the design conservatism of a wind blade in fatigue. Reduction in the safe operating lifetime of a blade with defects, compared to one without has shown that the inclusion of defects is critical for proper reliability assessment. If one assumes that defects account for some of the uncertainty in the blade design and these defects are analyzed with application specific data, then safety factors can be reduced. It has been shown that characterization of defects common to wind turbine blades and reduction of design uncertainty is possible. However, it relies on accurate and statistically significant data.