Evaluation of the bi Wave Method for Ultrasound Preload Determination in the Field with Machine Learning

Preload determination in bolts via ultrasound measurements is still a challenging task. Effects like scattering, interference and mode conversion produce signal distortion, which can cause invalid timeof-flight measurements and yield unreliable preload determination. There are different methods to detect invalid signals and eliminate them from the data analysis. However they have mostly been applied on controlled laboratory scale data sets. This paper evaluates the extension of these methods to more complex data collected on a wind turbine shaft.