Acoustic emission technology-based multifractal and unsupervised clustering on crack damage monitoring for low-carbon steel

Structural crack damage caused by high-stress concentration and external alternating load is characterized by imperceptible and low damage load, which makes online monitoring difficult. To address above challenges, an online monitoring method for crack damage recognition is proposed based on acoustic emission technology by combining multifractal and unsupervised clustering. Compared with base metal, the mechanical properties of welding area have changed significantly due to the influence of welding on local plasticity and strength of the material. The generalized Hurst value of welded tensile parts increases in each q order. The unsupervised clustering model was used for scanning online pattern recognition of unknown data to realize in-service crack pattern recognition. In addition, clustering coefficient is used to prove the superiority of DBSCAN method. This work can provide guidance for online monitoring of crack growth by acoustic emission technology during the service of nuclear power ship.