3.1 Publishable summary The VIBRATION project aims to provide an improved solution for the Structural Health Monitoring (SHM) of composite structures based on the vibration characteristics of the structure during operation. The main concept lays on the learning process of an electronic system fed by both testing results and Finite Element Analyses (FEA) of the specific structure that, by comparison with the in-flight response of the sensors, will provide information on the damage location and severity. The project has faced some problems related to the generation of usable finite elements models to be correlated to the testing results and has suffered some delays that have impacted on a number of deliverable reports that were not submitted. These delays will not have any impact on overall project objectives, as well as global schedule and budget remains unaltered. An important output from the mitigation activities carried out because of the delays in WP3 is a new approach to the statistical analyses of the damaged and undamaged structures test results that will allow identifying the damaged structures based on a new learning process developed throughout massive statistical analyses of the vibration tests outputs. (Big data analyses.) Although it is not included this approach in the DoW, the team will try to get the best from this new approach and present the results together with the targeted results. An industrial workshop that will take place at TWI's premises by middle of October 2015 will show the project results and promote the input from the industry to find additional applications. In the first period of the project WP1 was already completed. In the second period WP2 was also competed while WP3 was close to finalization by the end of April '15 and completed when this report is submitted. WP4 activities have started on time but focusing its activities on the mitigation of the problems arisen in WP3. WP5 and WP6 activities are in progress, as they had been moved forward, and the dissemination activities of the project included in WP7 keep updated the project website (http://www.fp7-vibration.eu/) and have produced the first two drafts for the Market Analysis (D7.2) A description of the work performed in individual WPs is given below: WP1: Finalisation of specifications As stated in the previous periodic report, the specifications for the small scale component (WP2), the vibration experiments (WP2), the FE modelling and simulation (WP3), the signal processing tools (WP4), the vibration SHM platform (WP4) and the demonstrator component (WP6) have been agreed. The demonstrator component will be a real size composite boom used in the manufacturing of UAVs by beneficiary IAI. (Figure 1)

Figure 1: Demonstrator position in the UAV

The small scale component that will be used in the development of the SHM methodology is a composite beam of rectangular cross-section. 45 units of the small scale component have been manufactured for WP2 by beneficiary ATR. (Figure 2)

Figure 2: Small scale composite part

The specifications for the SHM platform have been defined. The damage type that will be studied is damage after impact. Three distinct impact energies will be studied. The energy levels will be defined by beneficiary CTA. A summary of the specifications is given below:

1. The candidate characteristic quantities will be based on: • Frequency-domain non-parametric and parametric signal representations • Time-domain non-parametric and parametric signal representations • “Blind” selection methodologies

2. SHM platform operational conditions: • Random excitation • Operation under varying (temperature) conditions • Robustness under non-stationary excitation conditions

3. SHM platform damage diagnosis level: • Damage detection • Damage identification (localization) • Damage quantification (“low”, “medium”, “severe”)

o Low: Damage barely visible o Medium: Damage easily spotted o Severe: Damage that transverses the thickness of the composite part

4. SHM platform vibration sensors • IEPE accelerometer

The methodology for the FE modelling and the modelling of the damage had been decided, although later changed. The damage to be modelled will be considered as the elimination of different layers as they are not able to carry loads. The part will be divided into two regions and there will be two different damage severities. This information is sufficient to specify the damage site in the finite element model either by a partitioning strategy or element-location strategy.

The vibration experiments in the small scale composite parts have been defined. Table 1: Experiments at ambient temperature

Condition of the part

Experiments per part

Number of parts

Damage location Damage magnitude

Number of experiments

Healthy 5 Up to 45 No damage No damage Up to 225 Damaged 5 Up to 45 Up to 15 locations linearly spaced

throughout the part length Low damage

Medium damage High damage

Up to 225

Table 2: Experiments at low temperature

Condition of the part

Experiments per part

Number of parts

Temperature range

Damage location Damage magnitude

Number of experiments

Healthy 3 Up to 45 From -20°C to 20°C at 10°C

intervals

No damage No damage Up to 675

Damaged 3 Up to 45 From -20°C to 20°C at 10°C

intervals

Up to 15 locations linearly spaced

throughout the part length

Low damage Medium damage

High damage

Up to 675

WP2: Lab scale experiments The manufacturing of the 45 small scale composite parts (Figures 3 & 4) has been completed by beneficiary ATR that has procured dedicated tooling for the parts. One shot RTM was used for the production of four beams per curing / manufacturing cycle. The composite layup structure is [(+/-45º)_3, 0º_3]_s.

Figure 3: Small scale composite part layup

Figure 4: Small scale composite part manufactured

The LY564/XB3486 resin system was used for the manufacturing. A unidirectional (UD) tape and a Twill style fabric, both of high strength / intermediate modulus carbon fibre, were used as reinforcement. The first parts were delivered to the beneficiary UOP for vibration characterisation of the healthy state. It was later found later that most of them were already damaged in an undetermined (not systematic failure) because of low bonding (low resin content) in the middle of the laminate. Destructive testing by means of water-jet cutting was performed to assess the importance of the

failure. The final decision was to continue with the testing as the short time (less than 30 second) and low amplitude wouldn't induct nay failure progression in the laminates / beams. Beneficiary CTA received intact beam and proceeded to produce impact damage in a controlled environment. NDI testing at CTA and TWI followed the impact in order to ensure that the damage extension was according to the literature and that the initial delaminations hadn't produced unexpected damage extension. Infrared Thermal Tomography (IRT) and Ultrasound (US) testing was performed in the healthy parts (as manufactured) and in the damaged parts (damaged induced by CTA using a certified pendulum hammer).

Figure 5: Impact location

In the healthy parts, IRT revealed superficial scratches and small differences in the thermal properties of the exposed surfaces that do not represent mechanical faults. Representative findings included the areas where the fabric liner was placed and the areas of the tool inlet and outlet points. US testing in the healthy parts revealed no defects in the areas where the small differences were detected during the IRT tests. It has been concluded that the findings of IRT in the healthy parts represented thermal discontinuities that bear no mechanical faults.

Figure 6: IRT of beam No. 4 Figure 6: US inspection of beam No. 4

US measurement on the damaged parts revealed the area that has been affected by the impact. These areas have been denoted by a white marker in the composite. The affected areas were then measured by decomposition into simple geometries (rectangular, triangles and ellipses). Linear relationships of the area affected by the impact versus the impact energy have been constructed. The vibration experiments and characterisation work was performed in task 2.3. It is necessary to highlight the fact that there were variabilities in the small scale composite beams mass and dimensions that would inadvertently affect the vibration response of the specimens. The experimental setup can be seen in the attached file. The experimental database layout has been created. The size of the database is quite substantial (~250 GB), which prevents the uploading to the project website.

Damage

Figure 7: Structure of the data base that contains the extracted characteristic quantities

An analysis based on non-stationary time dependent excitations for the extraction of the characteristic quantities (task 2.4) has been performed using non-stationary excitation and the TARMA model, which enhanced the probability of detecting a true defect while the probability of detecting a false event is greatly reduced. All the experiments are now complete, including as an output - Experimental set up design (small scale beams) - Data acquisition system - Data acquisition and pre-processing - Construction of the database

Figure 8: Testing device and specimen set up

44 healthy and 30 damaged beams were used with strong variability in their dynamics. Each beam represents the boom of a commercial UAV. Clamping simulates connection to fuselage (IAI suggestion) and aluminium tail mass simulates part of the aircraft tail.

Figure 9: PSD estimates for one healthy beam and its damaged counterpart

3,953 experiments/signals were retrieved.

Figure 9: AR(47) model parameters for one healthy beam and its damaged counterpart

The variability in dynamics of healthy beams is overlapping with the variability in the damaged beams.

Figure 10: FRF magnitude estimates for all healthy beams are contained in the blue zone while the red zone contains the FRF

magnitude estimates of the damaged parts (44 healthy beams; 33 damaged beams; response Points Y1-Y3;

1,500 more experiments were run with this second this setup. Under each temperature, the beams exhibit variation; healthy and damaged beams temperature measurements still overlap but a different statistical analysis, to be performed in WP4, promises better results. WP3: FE and damage modelling The initial very detailed plan, as was analysis of the beam based on solid elements to be able to simulate real impact damage, which meant also the modelling of the impacting element, and do analysis based on that has been avoided because it needed days to run each model. The problems were that, in order to model impact energy using high velocity impact with low mass and small dimensions, there were penetration of particle in laminate, which it wasn't realistic. Looking for smaller dimensions on damage area, like those produced by stones, invisible initially, was not affordable and not real.

Figure11: New beam model

It was decided to replace the solid brick models by a simpler model based on shell elements and modelling the damage in the beam by eliminating different "non-working" plies, the dimension of the damage is easier to see in ultrasonic analysis. This came from CTA analysis. Infrared is less clear. Highly damaged area at the back of the laminate, no penetration is produced. The scale of detail of damage determines which plies of the laminate are damaged – most at the centre of the impact. Results are more reliable in simpler process than the first process.

Figure12: Natural frequency variation for damaged beams in relation to intact beams

Initial analyses based on the change in natural frequencies of the beam have shown that the approach is correct. Later analyses producing FRF functions from the FEA results demonstrate that changes in graph changes are correlated to the existence of damage. The positioning of the damage has not yet been proved as a feasible output of the FEA.

Figure13: Example of FRF variation from intact to damaged beams. Abaqus results.

WP4: Advanced signal processing tool. There are results of stationary and non-stationary excitations based on a version of the new method developed for SHM and presented that is, based on characteristic quantities which are extracted from dependent residuals and multiple healthy models. The healthy beams used were randomly selected, and therefore had mixed variability. A function of the second and third sensors was used and that there were similar results for other combinations of sensors. It is also confirmed that this test was for damage detection, not damage location detection. The beam that proved problematic during the stationary excitations was detected well during non-stationary testing, proving the method to be robust, and also that it has the capability to perform damage detection successfully. It remains to be seen how this will work with the FEA results. Using both stationary and non-stationary excitations allows the problem to be tackled from different angles; the signal can be manipulated in different ways to understand the state of the beam’s structure. A combination of all approaches can be used as part of the final methodology, as each test runs very quickly, meaning the results can be combined, giving extra validation. WP5: Prototype SHM platform development The prototype system has been developed and is used to: • Collect vibration measurements and other operational parameters; • Training phase will provide reference models used in online scheme; • To compare with current vibration measurements and other parameters for performing final damage detection. The system chosen is a CompactDAQ platform, due to its high accuracy A/D converter, its flexible platform and robust communications. This training system uses input vibration responses and parameters to return data, and then compares these results to SHM methodology reference data.

The functionality of CompactDAQ training system includes: • Time signals; • Embedded signal analysis (PSD, Frequency response function (FRF), coherence function estimate and up to 4 graphs simultaneously on large screen); • Produces comparative graphs. The software for the online damage detection has not yet been written but will start over the next few months, based on the final SHM methodology. WP6: Full scale demonstrator IAI are in the process of manufacturing four booms; three are in the assembly stage and the last is in the NDT verification stage. Impact test have been performed using beams with similar dimensions to the real booms, to give an idea of the energy level needed to produce appropriate levels of damage. The impacts were all made on the same wet layup beam, using a variety of weights dropped from one height. An aluminium sliding sheet was used to avoid unintentional damage to the layup beam. The energies estimated did not cause the damage expected because of the global elasticity of the boom, as well as glass layer resistance properties. The results were obtained using both ultrasonic and tapping tests. The next stage will be repeating the tests on the second side of the imitation beam, using higher levels of energy (14J/mm and above), and also to test coupons made from the Carbon/Ep materials used for the Demonstrator. WP7: Exploitation and dissemination Project website: The website is up and running, and is controlled, maintained and updated by CERTH as and when necessary. Online database: Likewise, this is available for all project related information and documents to be uploaded. Mid-term workshop: The workshop is due in November 2015, to be hosted at TWI alongside the next partners’ meeting. A draft invitation and brochure has been prepared by CERTH. The date needs to be finalised, invitations sent to research community and followed up by proceedings and feedback from attendees. Industrial seminar: The seminar will be held near the project end, at TWI, with presentations on the in-process quality system and its applications. The industrial community is to be invited and the event will be publicised on websites and in newsletters, followed up by feedback from attendees and conference proceedings. Conferences and exhibitions: Planned CANSMART (UoP) and SAMPE Europe (TWI) both in 2015 second half.