Signals received by piezoelectric transducers (PZT) network can be influenced by many factors. Apart from environmental conditions, whose variability should be compensated, significant difference in a signal can be also caused by relative geometry changes of a designed sensors node, e.g. the damage localization and its orientation with respect to sensors location in the node. In the adopted approach a set of damage indices (DIs), carrying marginal signal information content and correlated with the total energy received by a given sensor are proposed. These are sensitive to the two main modes of guided wave interaction with a fatigue crack, i.e. its transmission and reflection from a damage. Detailed description of DIs detection capabilities are delivered in the paper. Two dimensional reduction techniques: Principal Component Analysis and Fishers Linear Discriminant are compared. The results of the data collected from specimen fatigue test are used to compare several classification models based on the emerged effective damage indices.

Słowa kluczowe: pękanie zmęczeniowe, trwałość zmęczeniowa, funkcja gęstości, równa-nie Fokkera-Plancka.

This paper presents technique for qualitative assessment of fatigue crack growth monitoring, utilizing guided elastic waves generated by the sparse PZT piezoelectric transducers network in the pitch – catch configuration. Two Damage Indices (DIs) correlated with the total energy received by a given sensor are used to detect fatigue cracks and monitor their growth. The indices proposed carry marginal signal information content in order to decrease their sensitivity with respect to other undesired non-controllable factors which may distort the received signal. The reason for that is to limit the false calls ratio which besides the damage detection capability of a system, plays a crucial role in applications. However, even such simplified damage indices can alter over a long term, leading to the misclassification problem. Considering a single sensing path, it is very difficult to distinguish whether the resultant change of DIs is caused by a damage or due to decoherence of these DIs. Therefore, assessment approaches based on threshold levels fixed separately for DIs obtained on each of the sensing paths, would eventually lead to a false call. An alternative approach is to compare changes of DIs for all sensing paths. Developing damage distorts the signal only for the sensing paths in its proximity. In order to decrease the misclassification risk, a method of compensating such DIs drift is proposed. The main features and damage detection capabilities of this method will be demonstrated by conducting a laboratory fatigue test of an aircraft panel. The proposed approach has been verified on a real structure during fatigue test of a helicopter tail boom.

No abstract

No abstract