Application of pattern recognition for damage classification in composite laminates

Structural Health Monitoring (SHM) systems are susceptible to rising false positive rates over time due to ageing materials, scheduled maintenance procedures and new structural repairs. The alternatives of, manually updating thresholds or retraining software are impractical, time-consuming and complicate certification. This paper discusses an adaptive SHM methodology to accommodate changes in structural response that are not attributable to damage. This methodology provides a path to implementing most standard damage detection algorithms, ranging in sophistication from percent change to pattern recognition, across an aircraft fleet in a static release format. The main departure from traditional SHM architectures resides in adaptive modules that can accommodate input changes, such as those due to manufacturing or installation variability, sensor health, bond quality and typical wear on a structure. The overall goal was to integrate these adaptive modules within standard algorithms and logic without impacting their underlying reasoning or validity. An application of this methodology is presented using data collected from graphite/epoxy laminates subjected to Lamb wave testing. In this example, a pattern-recognition algorithm is employed to provide information about the presence, type and severity of damage. The proposed methodology eliminates the need for re-training when slight variations are introduced to the experimental setup.