No abstract

No abstract

No abstract

Considering jointly damage sensitive features (DSFs) of signals recorded by multiple sensors, applying advanced transformations to these DSFs and assessing systematically their contribution to damage detectability and localisation can significantly enhance the performance of structural health monitoring systems. This philosophy is explored here for partial autocorrelation coefficients (PACCs) of acceleration responses. They are interrogated with the help of the linear discriminant analysis based on the Fukunaga-Koontz transformation using datasets of the healthy and selected reference damage states. Then, a simple but efficient fast forward selection procedure is applied to rank the DSF components with respect to statistical distance measures specialised for either damage detection or localisation. For the damage detection task, the optimal feature subsets are identified based on the statistical hypothesis testing. For damage localisation, a hierarchical neuro-fuzzy tool is developed that uses the DSF ranking to establish its own optimal architecture. The proposed approaches are evaluated experimentally on data from non-destructively simulated damage in a laboratory scale wind turbine blade. The results support our claim of being able to enhance damage detectability and localisation performance by transforming and optimally selecting DSFs. It is demonstrated that the optimally selected PACCs from multiple sensors or their Fukunaga-Koontz transformed versions can not only improve the detectability of damage via statistical hypothesis testing but also increase the accuracy of damage localisation when used as inputs into a hierarchical neuro-fuzzy network. Furthermore, the computational effort of employing these advanced soft computing models for damage localisation can be significantly reduced by using transformed DSFs.

The central message of this paper is that for robust and efficient damage detection the damage sensitive features (DSFs) should be selected optimally in a systematic way such that only these features that contribute the most to damage detectability be retained. Furthermore, suitable transformations of the original features may also enhance damage detectability. We explore these principles using data from a wind turbine blade. Several damage extent scenarios are introduced non-destructively. Partial autocorrelation coefficients (PACCs) are proposed as vibrationbased DSFs. Scores calculated with principal component analysis (PCA) of PACCs are the transformed DSFs. Statistical distances between the DSF subsets estimated from the healthy and a reference damage state are calculated with respect to a statistical threshold as a measure of optimality. The fast forward (FF) method and a genetic algorithm are used to optimize the detectability of damage by DSF selection. The comparison between the two methods points out that FF offers a comparable performance at a lower computational cost. The classifiers based on the optimal features are tested on data from several previously unseen healthy and damaged cases and across a range of statistical detection thresholds. It is demonstrated that the selected PCA scores of the PACCs are superior compared to the initial features and allow identifying small damage confidently.