Piotr Omenzetter and Simon Hoell's work within the Lloyd's Register Foundation Centre for Safety and Reliability Engineering at the University of Aberdeen is supported by Lloyd’s Register Foundation. The Foundation helps to protect life and property by supporting engineering-related education, public engagement and the application of research.Peer reviewedPostprin

The ever-increasing demand for efficiency and cost improvements in lightweight structures with guaranteed safety and reliability is leading to the application of a damage-tolerant design philosophy. Here, accurate knowledge of structural health is critical to avoid catastrophic failures. This knowledge can be obtained by using advanced structural health monitoring (SHM) systems. For thin-walled lightweight structures, methods utilizing guided waves generated by piezoelectric transducers are well suited. The interaction between the guided waves and potential damages can be described by so-called wave damage interaction coefficients (WDICs). These WDICs are unique for each damage and depend solely on its characteristics for a given structure. Therefore, the comparison of known WDICs with estimated ones allows drawing conclusions about the current structural state. In this paper, a novel damage identification method for plate-like structures based on a database of such WDICs is presented. Selected damages are simulated numerically with finite elements to generate WDIC patterns. However, these simulations are computationally highly demanding, thus only a very limited number of damage scenarios can be simulated. This study proposes an innovative technique to substantially enhance the resulting WDIC database by using deep neural networks (DNNs). These DNNs enable smart interpolations and allow not only predicting WDICs for previously unseen damages at low computational costs but also the discovery of knowledge about the complex relationship between damage features and WDIC patterns. A comparison to other machine learning algorithms clearly shows the superior performance of the utilized DNNs for interpolating complex WDIC patterns. The proposed damage identification method is verified using advanced time-domain simulations of a large aluminum plate. A statistical analysis of correct identification rates in a common three-sensor setting is employed for assessing the general performance. It is demonstrated that carefully identified DNNs enable to accurately replicate and interpolate complex WDIC patterns. Furthermore, it is shown that these predicted WDICs allow identifying damage characteristics with high confidence.

The central message of this article is that for robust and efficient damage detection the damage sensitive features 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 are proposed as vibration-based damage sensitive features. Scores calculated with principal component analysis of partial autocorrelation coefficients are the transformed damage sensitive features. Statistical distances between the damage sensitive feature 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 method and a genetic algorithm are used to optimize the detectability of damage by damage sensitive feature selection. The comparison between the two methods points out that fast forward 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 principal component analysis scores of the partial autocorrelation coefficients are superior compared to the initial features and allow identifying small damage confidently.

The present paper comprehensively investigates the complex interaction between ultrasonic guided waves and local structural discontinuities, such as damages, through highly sensitive features: so-called wave damage interaction coefficients (WDICs). These WDICs are unique for each structural discontinuity and depend solely on their characteristics for a given structure and condition. Thus, they can be particularly useful for advanced assessment of lightweight structures in the context of non-destructive evaluation and structural health monitoring. However, the practical application of WDICs entails significant difficulties due to their sensitivity and complex patterns. Therefore, this study attempts to guide researchers and practitioners in the estimation of WDICs from numerical simulations and physical experiments. Detailed investigations are made for an aluminum host plate modified by artificial structural discontinuities, i.e., surface-bonded steel sheets. The numerical simulations are performed to predict WDICs and study sensitivities using a sophisticated finite element model. The experimental setup uses piezoelectric transducers to excite guided waves in the host plate. A single scanning laser Doppler vibrometer measures the scattered guided waves caused by the surface-bonded steel sheets, and the resulting WDICs with possible influences are investigated. In both cases, the orientation and thickness of the attached steel sheets were varied to create 12 different damage scenarios. In general, the comparison between numerical and experimental WDICs show good agreement. This underpins the applicability of the general methodology for simulating and measuring WDICs over all scenarios. Furthermore, the WDIC scattering patterns reveal a clear dependency of the peaks in the back-scattered reflections for both the numerical and experimental amplitude coefficients on the damage orientation, basically following the law of reflection. However, some discrepancies between both studies were observed. Numerical sensitivity analysis identified the adhesive layer as one reason for such differences. Additionally, misalignment errors in the experimental measurements were also found to affect WDICs. Therefore, an improved baseline subtraction method is proposed, which clearly enhances the experimental WDICs. Finally, an experimental sensitivity study of WDICs for selected sensing radii revealed only a minor influence. All these investigations were made for the amplitude as well as the phase representation of WDICs. Thus, these findings may open the way to future research and development of techniques employing WDICs for advanced applications of non-destructive evaluation and structural health monitoring.