10:30
S9: Advanced monitoring strategies for composite structures at Amphi Bézier (AB)
Chair: Jesus Eiras Fernandez and Eric Monteiro
10:30
20 mins
|
Enhanced Health Assessment of Aeronautical Subcomponents: Data-driven Historical-Independent Health Indicators from Lamb Waves
Morteza Moradi, Juan Chiachío, Rinze Benedictus, Dimitrios Zarouchas
Abstract: To predict the remaining useful life (RUL) of a structure and assess its health condition, a comprehensive intermediary known as the health indicator (HI) plays a vital role, bridging the gap between sensory data from structural health monitoring (SHM) and prognostic models. An effective HI must meet specific criteria, including monotonicity, trendability, and prognosability. Nevertheless, developing a suitable HI for aerospace composite structures presents a formidable challenge due to the complex and stochastic nature of damage accumulation during operational conditions. This challenge is compounded when aiming for a HI that is independent of historical SHM data, as both HI and RUL predictions are influenced by past events, making them historically dependent. Extracting informative, historical-independent data requires a dependable SHM technique, and the Lamb wave (LW) technique emerges as a valuable method capable of achieving this. However, translating LW data into the corresponding HI value at each time step presents a significant hurdle. In this context, artificial intelligence, specifically deep learning, offers substantial mathematical potential to address this challenge. To tackle this issue, a convolutional neural network (CNN) is developed within a semi-supervised learning paradigm (SSCNN) to design the HI (see Figure 1). The model exclusively utilizes current LW data, eliminating the need for past or future data to generate the present HI value. Moreover, SSCNN is flexible enough to adapt to varying numbers, networks, and configurations of LW sensors. Evaluation criteria are subsequently computed based on the entire trajectory of HI values until the aeronautical subcomponents reach the end-of-life point. The proposed approach is studied and validated using two distinct datasets of composite specimens monitored using the LW technique: NASA, which involved tension-tension fatigue testing on composite dogbone specimens with varying layups [1], and ReMAP, which included compression-compression fatigue tests on T-single stiffener composite panels, involving impacts and disbond damage [2]. The LW data is initially processed using the Hilbert transform (as seen in Figure 2), with signal envelopes serving as inputs for the SSCNN. The results affirm the effectiveness of this approach in terms of the evaluation criteria. Additionally, HI quality was improved while deep learning randomness was decreased with the use of ensemble learning approaches. Prognostic criteria were fulfilled with high performance of up to 93% and 81%, respectively, for T-single stiffener composite panels under compression-fatigue and dogbone composite specimens under tension-fatigue loadings, demonstrating the validity of the technique (see Figure 3).
|
10:50
20 mins
|
Contactless inverse method for the characterization of adhesion within structural materials with guided elastic waves
Loïc Girardot, Mourad Bentahar, Silvio Montrésor, Nicolas Terrien
Abstract: Assembly methods in industry have improved considerably during the last decades. Indeed, mechanical fastening is increasingly abandoned in favor of structural bonding and/or welding methods. The need for improved nondestructive testing and evaluation (NDT&E) methods becomes therefore important particularly for thermoplastic composites welding. The latter generates subtle defects that can hardly be detected, which makes necessary the need to develop specific well-calibrated techniques. In addition to bonding testing, which is based on the existing contrast between high and weak bonded areas, it remains necessary to determine the adhesion quality, particularly in the case of multi-layered materials.
In order to better characterize bonding and adhesion, the Laboratoire d’acoustique de l’Université du Mans and the Institut de Recherche Technologique Jules Verne are developing a research project via a PhD thesis, which started on October 2022. The main objective of this work is to create a noninvasive NDT&E method to improve the characterization of such assembly, involving the evaluation of adhesion as well as the existing defects. One major issue in bond evaluation is the detection of subtle defect, such as Kissing Bond (KB) or weak adhesion. In that sense, note that nonlinear acoustic methods have already proved a good sensitivity to subtle defects [Del12, Yan10], but their application is not yet adapted to industrial processes due to their lack of repeatability and adaptability to the different structures. Due to this industrial context, the use of Air Coupled ultrasonic Transducers (ACT) is preferred. ACTs offer good repeatability and they can excite and detect both bulk and guided waves [Cas01]. Therefore, the present work is intended for the creation of an ACT-based NDT method, where the nonlinear acoustic signatures are used for the characterization of thermoplastic composite welded assemblies.
The first step of this work is focused on the adhesion evaluation. An inverse method based on the dispersion curves of Lamb waves is performed. Lamb waves are used, where their sensitivity along the thickness of the propagating medium is exploited. Numerous works have used Lamb waves for the characterization of materials properties in various applications [Boc18, Hel00]. For the evaluation of adhesion, an approximation is usually made on the properties of an interface layer used as representative of the adhesion quality. Under the condition of a thin layer (when compared to the wavelength involved), a relation also exists between the interface stiffness and the velocities of bulk waves. These quantities are also dependent on the mechanical properties of the propagation medium. Therefore, a relationship between mechanical properties and equivalent interface stiffness can be defined.
Lamb waves properties are studied through dispersion curves, where the phase and/or group velocities (or wavenumber) is determined as a function of the excitation frequency. They can be obtained theoretically and experimentally. Various analytical models exist for their calculation, all including different approximation and hypothesis. A previously established model adapted to multilayered structures is used in this work [Low95], as an equivalent for the whole thickness was not practicable due to the specific study of the interface layer.
Double Fourier Transform method (2DFFT) [All91] is used to exploit the experimental data, which makes possible to see the dispersion curves of the propagating modes within the structures at the frequency of excitation. The inverse method then compares the theoretical and experimental dispersion curves, and, according to the least square criteria, tends to minimize the difference by varying the elastic modulus of the interface layer. A sensitivity study revealed that Young modulus (E) is the most influent parameter, where its variation only in the “thin” interface layer has shown important differences on the dispersion curves.
The established inverse method has provided good result after measurement of various samples with and without defects, each of them having different size and properties. Results have shown a good match with the created defects within the different samples, based on their size, location and type. In order to link the measured interface stiffness to the actual tensile strength stress or mechanical performances, mechanical destructive tests are to be done on these samples, where beyond the measured values, the stiffness ranking of the samples will be important.
Next steps of this work will be focused on the defect detection, location and definition using the nonlinear interaction between defects and the generated Lamb waves. Also, the method presented can still be improved by use for examples of different Lamb modes, or even Shear-Horizontal waves (SH) to improve sensitivity to certain types of defects.
[Del12] S. Delrue and K. Van Den Abeele. Three-dimensional finite element simulation of closed delaminations in composite materials. Ultrasonics. 52 (2):315-324, 2012.
[Yan10] D. Yan, B. W. Drinkwater and S. A. Neild. Experimental and theoretical characterization of kissing bonds in adhesive joint using non-linear ultrasonic measurement. Review of progress in quantitative nondestructive evaluation. 29: 1190-1197, 2010.
[Cas01] M. Castaings and B. Hosten. Lamb and SH waves generated and detected by air-coupled ultrasonic transducers in composite material plates. NDT&E International. 34 (4):249-258, 2001.
[Boc18] N. Bochud, J. Laurent, F. Bruno, D. Royer and C. Prada. Towards real-time assessment of anisotropic plates properties using elastic guided waves. The Journal of the Acoustical Society of America. 143 (2): 1138-1147, 2018.
[Hel00] K. Heller, L. J. Jacobs and J. Qu. Characterization of adhesive bond using Lamb waves. NDT&E International. 33 (8): 555-563, 2000.
[All91] D. Alleyne and P. Cawley. A two-dimensional Fourier transform method for the measurement of propagating multimode signals. The Journal of the Acoustical Society of America. 89 (3): 1159-1168, 1991.
[Low95] M.J.S. Lowe. Matrix techniques for modeling ultrasonic waves in multilayered media. IEEE Transactions on Ultrasonics, Ferroelectrics and Frequency Control, 42(4):525–542, 1995.
|
11:10
20 mins
|
A Modelling platform for Structural Health Monitoring
Fabrice Foucher, Bastien Clausse, Roman Fernandez
Abstract:
ABSTRACT
Among Structural Health Monitoring methods, guided wave SHM is a technique relying on the permanent integration of a transducers’ network aiming at detecting potential damages affecting the structural integrity of the monitored structure. To be part of a competitive predictive maintenance strategy, the choice of relevant GW-SHM sensors (excitation mode, number, location in the specimen) is of paramount importance to optimize critical flaws detection while limiting their number and thus the whole cost. Since building monitored and flawed prototypes is even more costly than simple mock-ups, simulation can play a key role as a virtual prototyping tool before physical implementation.
To increase the industrial deployment of GW-SHM systems, a solid performance demonstration stage is also required. Indeed, due to the fully automated measurement process, the large number of influential parameters involved, the numerous sources of false alarms and the potential failures of the embedded sensors themselves, reliability studies must be conducted. Experimental campaigns become rapidly prohibitive if you want to take all uncertainties into account. This is why simulation can help since it can provide a large amount of data and information regarding some of the most influential parameters, which will greatly reduce the need for costly mock-ups, limiting their number to a relevant sampling.
The CIVA platform is a well-known multi technique simulation and analysis software in NDT which is developed by CEA LIST, but also results from the contribution of numerous industrial and academic partners within Europe. CIVA can simulate UT, GWT, ET, RT, CT, and now Thermography in NDT. CIVA has also recently expanded its modelling tools to Structural Health Monitoring based on guided waves. Using a combination of spectral Finite Elements methods and a parametric macro-mesh technique, CIVA SHM can address many industrial SHM configurations (metallic or composite materials, planar, cylindrical, or curved geometries) with very competitive performances [1].
As for any simulation software, it is important to rely on experimental validation references to verify the relevance and accuracy of the models and use it with confidence in an industrial project. Several validation studies have been conducted around CIVA SHM for isotropic metallic plates, cylinders, or for composite panels. In this presentation, some validation cases will be compared to experimental data provided by the « Open Guided Waves » initiative [2]. This will deal with the instrumentation of a composite panel made with 16 carbon-epoxy plies for a 2mm total thickness. This type of structure exhibits anisotropic properties with respect to the ultrasonic guided waves propagation.
Figure 1: Monitored composite panel proposed in the validation benchmark.
ICCBMA, 11-13 September 2024
Paris, France
12 piezo-electric patches of 10mm diameter are located on the panel. Several excitation frequencies are considered, all sensors are successively excited as transmitters (with a radial load) while the 11 others are receivers (also called round-robin).
Figure 2: Detailed A-Scans received by the 12 sensors considering one of these sensors as the transmitter (simulation in red, experiment in blue).
After presenting some experimental validation cases, several application examples of CIVA SHM in the context of a damage monitoring in a composite plate will be illustrated. Various parameters will be studied (sensors to specimen edges distance, number and location of sensors, transducer size and frequency, different defect type and dimensions). Imaging capabilities provided by the CIVA software (See Figure 3) will be illustrated. Finally, other applications in terms of specimen geometries and materials will be introduced.
Figure 3: Imaging of the signature for a Hole Ø20mm in a Plate of 1000mm*1000mm.
References
[1] Imperiale, Alexandre, and Edouard Demaldent. "A macro‐element strategy based upon spectral finite elements and mortar elements for transient wave propagation modeling. Application to ultrasonic testing of laminate composite materials." International Journal for Numerical Methods in Engineering 119.10 (2019): 964-990.
[2] Moll, Jochen, et al. "Open guided waves: online platform for ultrasonic guided wave measurements." Structural Health Monitoring 18.5-6 (2019): 1903-1914. http://openguidedwaves.de
|
11:30
20 mins
|
Upscaling numerically the laser shock-based disassembly of an adhesively bonded composite/metallic foreign object damage panel
Konstantinos Tserpes, Panagiotis Kormpos, Laurent Berthe
Abstract: The extensive use of multi-material structures in aviation industry, especially utilizing adhesive bonding, complicates the end of life management of aircraft structures [1,2]. In order to fully exploit the recent developments in the recycling processes of composite materials, efficient disassembly techniques are required for composite aerostructures. In recent works, it has been shown by tests [3] and numerical models [4] that the laser shock technique is a very promising disassembly method with many environmental and technical benefits. Since the proof-of-concept has been made at coupon level, it is very important to investigate the efficiency of the technique at larger levels. In the present work, a numerical upscaling of the method is performed by simulating the full disassembly of a part of an adhesively bonded composite/metallic foreign object damage (FOD) panel, which replicates a composite engine fan blade. The FOD panel is made from 3D woven CFRP material and incorporates an adhesively bonded titanium layer to its leading edge.
Numerical simulation has been performed using the LS-Dyna explicit FE code. The 3D woven substrate was modeled using a progressive damage model, while for the titanium layer the Johnson-Cook plasticity model in combination with the Grüneisen equation of state were used. The adhesive between the two materials was modeled using a cohesive zone model utilizing a bilinear traction separation law. To reduce the computational effort, a strip of the FOD’s leading edge was modeled.
Due to the complicated geometry of the FOD panel, a preliminary study on the effects of the inclination was performed. It was found that due to the small inclination of the FOD panel, there are negligible variations on the back face velocity and the shock wave propagation. Prior to the simulation of the full disassembly process, two additional simulation sets (spot on top surface and on bottom surface) were performed for the definition of the double pulse delay times and the evaluation of the effect of spot location. The parameter identification phase involved 29 shots. The first 14 shots targeted the top interface using a fixed delay time between the double pulses and shots 15 to 29 targeted the bottom interface and the delay time between the double pulses follows a second order polynomial equation across the length of the strip. The multi-step process simulation is utilized using a python script.
After 29 shots the FOD strip is almost completely debonded except for the leading edge, which was not targeted. The debonding of the top titanium/CFRP interface relies solely on laser characteristics and the thickness of the titanium layer. However, after the full debonding of the top layer, when targeting the bottom interface of the component, the delay between laser pulses varies and can be described by a second-order polynomial equation. By employing this study's findings to the automated multi-shot simulation procedure, the complete disassembly of the component can be achieved. The damage developed in the composite substrate is confined to matrix cracking parallel and perpendicular to fibers. The simulations did not predict any fiber failure.
The research leading to these results is part of the MORPHO project and has received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement No. 101006854.
References
[1] Pantelakis Sp, Tserpes KI. Adhesive bonding of composite aircraft structures: Challenges and recent developments. Sci China Phys Mech Astron 2014;57:2–11. https://doi.org/10.1007/s11433-013-5274-3.
[2] Zhao X, Verhagen WJC, Curran R. Disposal and Recycle Economic Assessment for Aircraft and Engine End of Life Solution Evaluation. Applied Sciences 2020;10:522. https://doi.org/10.3390/app10020522.
[3] Kormpos P, Unaldi S, Berthe L, Tserpes K. A Laser Shock-Based Disassembly Process for Adhesively Bonded Ti/CFRP Parts. Processes 2023;11:506. https://doi.org/10.3390/pr11020506.
[4] Kormpos P, Tserpes K. Αn efficient numerical model for the simulation of debonding of adhesively bonded titanium/CFRP samples induced by repeated symmetric laser shocks. The Journal of Adhesion 2023:1–25. https://doi.org/10.1080/00218464.2023.2255532.
|
11:50
20 mins
|
Laser Shock for NDT/SHM of composites materials and bonded joints: recent advances
Nicolas Cuvillier, Mathieu Ducousso, Marine Scius-Bertrand, Tomas Bergara, Laurent Videau, Laurent Berthe
Abstract: Structural bonding is a key industrial process for composites uses, in production as well as in Maintenance, Repair and Overhaul (MRO). In production, some parts such as for instance the leading edge of the Leap fan blade car bonded and in MRO some composites can be repair by using bonded patches. In all the cases, a Non Destructive Testing (NDT) able to prove that the mechanical performances of the bonding are equal to that expected is required. However, today, there is no industrial solution for such requirement.
Similarly, when controlling composites materials, bulk delaminations are a possible research defects, especially for MRO inspections. For this, whatever the physical- based method used to performed inspection (ultrasound, thermography or X-ray are the main ones), inspection setup is first calibrated by using a reference part. Such reference part contain in general a simulation of delaminations (for instance, a flat bottom hole or a teflon insert) as, today, we do not know how to produce real delamination in a composite part. Such limitation is also a main issue for SHM developments, as proposed methods, cannot been validated from real experiments.
A collaborative project have been proposed to find solutions to these limitations. Project is named MONARQUE. The partners of the MONARQUE project are Airbus, Safran, Thales, CEA, IO, Armelio, Rescoll and PIMM-ENSAM. In this project, partners proposed to used laser-driven shock waves to break the three mentioned locks, respectively 1) quantitative NDT of structural strength of bonding [1]; 2) delamination creation on a custom basis; 3) validation of SHM procedure for delamination detection.
The shock waves were generated using ns-scale intense laser pulses in the GW/cm² intensity range. Such illumination creates a plasma expansion from the sample surface and, thanks to reaction principles, an intense stress, in the GPa order, propagates into the bulk of the material. After back face reflection, phase change occurs, and the compressive shock wave becomes a release shock wave. In the MONARQUE project, several investigations have been performed to control the laser-induced pressure's loading [2]. Moreover, by using two laser pulses, two contra-propagating shock waves propagates in the materials and the resulting two release shock waves cross at a point driven by the time-delay between the two lasers pulses : if there is no time-delay, the two shock wave cross at the middle of the sample, while if one laser is delayed with respect with the other, they cross in an other point.
Such laser-loading loading have been used to the different applications previously mentioned. Applied to bonding, main works have been focused on the leading edge of the Leap fan blade (a titanium part bond on a composite one), to propose a method to control and disbond such bonding, for MRO applications. Applied to NDT and SHM, we demonstrated that we are able to real delaminations, at an interlayer position driven by the time-delay between the two laser pulses.
One laser shock platform, a major deliverable of the project, have also been installed and is now open for industrial applications in RESCOLL’s building (Pessac, Nouvelle-Aquitaine, France).
We propose an overview of the Monarque project in this communication.
Ultrasonic imaging of delaminations in a composite part generated by laser-driven shock waves. Delaminations in green are not at the same interply than delaminations in blue.
[1] M. Ducousso, S. Bardy, Y. Rouchausse, T. Bergara, F. Jenson, L. Berthe, L. Videau, N. Cuvillier, Appl. Phys. Lett., 112, 111904 (2018)
[2] M. Scius-Bertrand, L. Videau, A. Rondepierre, E. Lescoute, Y. Rouchausse, J. Kaufman, D. Rostohar, J. Brajer, and L. Berthe. Laser induced plasma characterization in direct and water confined regimes:new advances in experimental studies and numerical modelling. J. Phys. D: Appl. Phys. 54, 055204 (2021)
|
12:10
20 mins
|
Machine Learning Approach for LPRE Bearings Remaining Useful Life Estimation Based on Hidden Markov Models and Fatigue Modelling
Federica Galli, Philippe Weber, Ghaleb Hoblos, Vincent Sircoulomb, Giuseppe Fiore, Charlotte Rostain
Abstract: Ball bearings are one of the most critical components of rotating machines. They ensure shaft support and friction reduction; thus, their malfunctioning directly affects the machine’s performance. As a consequence, it is necessary to monitor the health conditions of such a component to avoid major degradations which could permanently damage the entire machine.
In this context, HMS (Health Monitoring Systems) and PHM (Prognosis and Health Monitoring) methodologies propose a wide range of algorithms for bearing diagnosis and prognosis. The present article proposes an end-to-end PHM approach for ball bearing RUL (Remaining Useful Life) estimation. The proposed methodology is composed of three main steps: HI (Health Indicator) construction, bearing diagnosis and RUL estimation. The HI is obtained by processing non-stationary vibration data with the MODWPT (Maximum Overlap Discrete Wavelet Packet Transform). After that, a degradation profile is defined and coupled with crack initiation and crack propagation fatigue models.
Lastly, a MB-HMM (Hidden Markov Model) is trained to capture the bearing degradation dynamics. This latter model is used to estimate the current degradation state as well as the RUL. The obtained results show good RUL prediction capabilities. In particular, the fatigue models allowed a reduction of the ML (Machine Learning) model size, improving the algorithms training phase.
|
|