Engineering structures are characterized with an increasing amount of data today. This includes their design as well as operational life and possibly even their recycling. The data process chain considered starts from loads applied and geometry, and further proceeds into stresses and strains, fatigue degradation and possible fracture to non-destructive testing and finally ends with structural health monitoring systems allowing the monitoring process to be automated. All of this data is provided in a variety of different formats and needs to be standardized and organized in a way such that it can be easily handled between the different process elements, continuously updated with data at any time and specifically visualized, such that the structure’s operator can always get the full picture of the structure based on the amount of data being available. For this purpose, a HDF5 data structure has been created to organize and manage the gathered data efficiently.

The process and procedure will be illustrated on a real structure as a demonstrator (Fig. 1) consisting of two I-beams made of steel, which are fixed at their ends and loaded under 3-point bending. The structure is intentionally simple such that the process of data accumulation and processing can be well understood. Furthermore, the existing hardware structure allows for monitoring of any data with any monitoring system even under various loading and degradation (i.e. cracking) conditions. The demonstrator is available to be reproduced by any party being interested to do so, specifically for future monitoring purposes. This may allow a broader discussion of the lateral integration of structural data, that is required to establish useful digital twins, allowing a structure’s life cycle condition to be assessed at any time. In excess, such a demonstrator can serve as a benchmarking platform for various monitoring techniques and help to demonstrate and possibly also improve the monitoring technologies’ performance. On the other hand, designers and operators of structures could define mechanisms (i.e. materials, loading conditions, notch geometries, plastic deformation conditions, crack lengths, etc.) that could be evaluated. Finally, the procedure presented for the demonstrator can be virtually updated to any more complex structure in the longer term. For the case being presented, are large set of data originating from numerical simulation and hardware monitoring will be presented and on how this has been merged and visualized as a first practical case.

A novel, physics-based data-driven formulation for quantifying the dissipative capacity of generic, complex multi-component systems called Analytical Root Mean Square Bandwidth (ARMS BW was developed [1]. This robust and accurate measure can be readily integrated in standard ground testing protocols, tying the modal properties to the dissipative capacity of a structure. Moreover, its computation is purely data-driven so it can be implemented in finite element (FE) simulations and experimental tests. For the FE model of a model aircraft - see Fig.(a), we visualized its ARMS BW distribution - see Fig.(b), and identified the locations of max and min dissipative capacity, (depending on the direction of an impact hammer test); this was verified by the velocity envelopes at these locations - see Fig. (c). This methodology provides physical insight into where attachments, e.g., stores and radomes, should be placed to maximize dissipative capacity and therefore reduce the dynamic response of the aircraft and/or component. As an example, if a new external component is to be integrated onto an existing platform, one can generate a spatial bandwidth plot of aircraft surfaces at candidate installation locations to select the optimal location to mitigate the transmitted vibration; this could increase the fatigue life of the component and therefore reduce aircraft downtime and logistics burdens. The same can be performed early-on in the development cycle of an aircraft for wing-store hardpoint placement using only finite element models and analysis.

[1] B. Chang, K. Moore, W. Silva, L.A. Bergman, and A.F. Vakakis, “Defining the Bandwidth of Multi-degree-of-freedom Classically Damped Linear Systems with Application to an Experimental Model Aircraft,” AIAA Journal (in review).

Reliable guided-wave monitoring of pipelines requires models that are both computationally efficient and capable of capturing the main physical propagation mechanisms in cylindrical waveguides. In thin-walled, large-diameter pipes, guided waves propagate locally in a plate-like manner while the cylindrical topology creates multiple deterministic surface-geodesic routes between actuator and sensor locations. Besides the direct route, spiral routes wrapping around the circumference can generate distinct received wave packets, and additional packets arise from reflections at pipe boundaries and local discontinuities. This paper presents a semi-analytical hybrid framework for pipeline guided waves that combines Wave Finite Element (WFE) dispersion extraction with a route-based long-range propagation engine. The model explicitly accounts for direct and spiral propagation routes as well as boundary-reflected contributions. Validation is performed in two steps. First, model predictions are compared to full 3D transient finite element simulations on a steel pipe, assessing arrival times, wave-packet structure, and route-dependent contributions. Second, experimental measurements on a steel pipe are used to identify and interpret the received wave packets. The results demonstrate that the proposed semi-analytical model captures the dominant wave packets observed in 3D FE and experiments, while requiring significantly lower computational effort than full transient simulation. The validated framework provides a foundation for subsequent model-assisted monitoring and data-driven localization studies in realistic pipeline environments.

In structural dynamics, linear system identification (SI) is highly developed, by which a structure of interest can be reduced to a set of single-degree-of-freedom linear oscillators. Unfortunately, the same cannot be said for nonlinear systems. One of the main problems with nonlinear SI (NLSI) has been that of structure detection or equation discovery. For nonlinear models, any functional term could be added, making its discovery a highly ill-posed problem.
While not one method stands out as all-encompassing, two prevalent groups of algorithms have emerged offering the prospect of a general framework; namely, Evolutionary Optimisation and Bayesian Inference. Within the group of algorithms comprising the former, a promising development has been in the use of genetic programming for optimising Tree-Adjoining Grammars (TAGs) tailored to NLSI. An attractive advantage of using TAGs is that they allow for the unification of many NLSI model structures within a single framework.
An extension of this method is presented here, whereby the optimisation problem is framed from a Bayesian perspective. By combining the TAG approach with Bayesian inference, the correct model can be found simultaneously with parameter estimation, providing their distributions and propagating uncertainty in a principled manner. This paper outlines the employed algorithm and presents the identification results from a simulated discrete-time polynomial input-output nonlinear model.

Online Health Management (HM) plays a pivotal role in optimizing the lifecycle of aircraft while ensuring safety, reliability, and structural integrity. During service, aircraft structures experience complex cyclic loading, making accurate load analysis essential for effective life monitoring and health management. Capturing realistic load histories throughout operation enables more precise estimation of remaining useful life, supports optimization of design limits for improved material utilization, and enhances the fidelity of simulation models by aligning them with real operating conditions.

Continuous load monitoring thus represents a key enabler for transitioning from conventional safe-life design philosophies toward advanced damage-tolerant and condition-based approaches. However, despite significant progress in sensing technologies and data processing methods, large-scale sensor deployment remains constrained by financial and logistical limitations, posing challenges to the realization of fully “smart” aircraft.

This study presents recent developments from the H2020 AVATAR project, which aims to establish a digital twin framework for optimizing aircraft lifecycle performance. A novel sensing skin has been developed to enable the integration of a sparse sensor network on composite wing structures. The work explores the use of machine learning (ML) to enhance load monitoring capabilities, demonstrating that ML techniques can maintain high predictive accuracy in reconstructing structural responses, such as forces, strains, and stresses, even with limited sensor data (from strain gauges and accelerometers). The proposed approach is validated on a composite wing under realistic operational loading, highlighting its robustness and potential for real-time structural health monitoring in demanding aerospace environments.