Sensitivity studies on sensor selection for crack growth investigation

This work focuses on an unsupervised, data driven, modified k-nearest neighborhood technique to detect and monitor fatigue crack growth in lug joint samples using a surface mounted piezoelectric sensor network. A lug joint is an important structural hotspot in which damage initiates and progresses under fatigue loading. Early detection of fatigue cracks in a lug joint can help in taking preventive measures, thus avoiding any possible structural failure. The lug joint samples used in this study are prepared from an Al 6061 T6 plate with 0.25 inch thickness and are instrumented with a surface mounted piezoelectric actuator/sensor network. Experiments are conducted on lug samples with a single notch and multiple notches that are symmetrically placed. For early initiation of cracks, samples are notched at the shoulders. Under the influence of fatigue loading, the crack growth rate is different even when the notches are symmetrically placed. It is found that although cracks propagate from both the notches, the sample fails from one of the shoulders once the critical crack length is reached. For the given sensor architecture, which is symmetric, the objective of this study is to detect, isolate and monitor fatigue crack growth in each zone. The methodology presented helps in identifying sensors that are most sensitive to the presence of single and multiple cracks. Thus, the computational expense for damage localization studies can be reduced by not making use of redundant sensors.