On the long-term stability of normal condition for damage detection in a composite panel

Novelty detection provides an effective strategy for carrying out the lowest level of damage identification. The idea is to create a model of normal condition based on data measured when the system is known to be undamaged and then examine subsequent data to see if there are statistically significant deviations from normality. Problems arise with this approach if the normal condition data varies with the environmental conditions. If the object is simply to detect significant change, variation between different normal conditions with differing environments must be factored out. The current paper examines the problem of environmental change for the situation where a composite panel is monitored for damage under changing conditions of humidity. A previous investigation was concerned with variation in temperature and, for the purpose of comparison, a set of temperature cycling tests was included at the end of the humidity experiment. The features used for detection are measurements of Lamb wave profiles and the detection method is outlier analysis. The exercise is carried out experimentally and the controlled conditions are provided by an environmental chamber.