Pattern recognition of Acoustic Emission (AE) data is generally performed based on prior knowledge about the acoustic signatures of material damage mechanisms. Traditionally, these signatures are implicitly assumed to be unaffected by damage accumulation during the fatigue life of the material. This study investigates the influence of cumulated damage under cyclic loading on the acoustic signatures of local fracture mechanisms in composites. Artificial AE sources are created using an ultrasonic transmitter and a Pencil Lead Break in order to reproduce various real-like AEs, such that the waveforms, measured at the sensor level, are characterized as similar to real acoustic signatures of local fractures in carbon/epoxy composites. It is shown that these waveforms are distorted with damage accumulation in both time and frequency domains, leading to important changes in the AE-features used in data classification. Consequently, this can engender unreliable statistical representation of the AE sources in such non-stationary media.

声发射（AE）数据的模式识别通常是基于有关材料损坏机制的声信号的先验知识来执行的。传统上，这些特征被隐式地假定为不受材料疲劳寿命期间损坏累积的影响。这项研究调查了循环载荷下累积损伤对复合材料局部断裂机制的声学特征的影响。人工AE源是使用超声波发射器和自动铅笔笔断裂创建的，以便再现各种真实的AE，从而在传感器级别测量的波形的特征类似于碳/环氧树脂局部裂缝的真实声学特征。复合材料。结果表明，这些波形在时域和频域均会因损伤累积而失真，导致数据分类中使用的AE功能发生重大变化。因此，这可能导致在这种非平稳媒体中对AE源进行不可靠的统计表示。