Applying network-free renormalization and clustering algorithms to reveal the crack evolution laws of laterally loaded composite T-joints

Abstract

Delamination or crack are standard failure patterns of composite T-joints. Most existing studies focus on pull-off loaded composite T-joint cracking but little focus on adverse conditions such as lateral loading. The mainstream composite T-joint research predicts the macroscale failure based on the composite’s microscale/mesoscale crack starting and evolution. However, the cracking process is within the microscale, mesoscale, and macroscale, making detecting its starting based on phenomena complicated. This work attempts to directly model the macroscale deformation distribution from a thermodynamic perspective to reveal the cracking/failure evolution of a laterally loaded composite T-joint. Firstly, we conducted eight laterally loaded composite T-joint tests with different molding processes. Then, network-free renormalization was applied to construct Matrices (Modes) and Hamiltonians (Characteristic parameters) that can characterize their cracking evolution process. Further, a clustering algorithm was applied to reveal the cracking starting points (phase transition loads) embedded in the macroscale deformation distribution. We can verify its rationality by comparing the composite T-joints localized and systematic phase transition loads based on Wilson’s phase transition theory. In summary, this work applies network-free renormalization and clustering algorithms from a thermodynamic perspective to reveal the cracking starting point embedded in the macroscale deformation distribution of the laterally loaded composite T-joints.