Classifying condition of ultra-high-molecular-weight polyethylene ropes with wide-angle X-ray scattering

Abstract

Ropes of ultra-high-molecular-weight polyethylene (UHMWPE) are replacing steel wires in many applications and nondestructive testing to monitor their condition is of scientific and commercial interest. In this work, wide-angle X-ray scattering (WAXS) combined with linear discriminant analysis (LDA) is proposed as classification method to distinguish between healthy and damaged UHMWPE ropes. Healthy (as produced, after pre-stretching) and damaged (in-field use) ropes (⌀=22mm) have been analyzed using synchrotron radiation. Firstly, it is demonstrated that scans of healthy and damaged ropes can be distinguished with 100% cross-validated test classification accuracy using LDA; this is shown both with the input data consisting of pre-processed 1D WAXS data and with physical parameters retrieved by fitting the WAXS data. Secondly, it is demonstrated that the classification performance is similar using the two forms of input data and that the noise could be increased by a factor of three while maintaining 100% test classification accuracy across all the three cross-validation folds.

Keywords

• Polyethylene
• Wear
• Wide angle X-ray scattering (WAXS)
• Machine learning