Available online at ScienceDirect. Procedia Engineering 188 (2017 )

Transcription

1 More info about this article: Available online at ScienceDirect Procedia Engineering 188 (2017 ) th Asia Pacific Workshop on Structural Health Monitoring, 6th APWSHM Dimpling monitoring and assessment of satellite honeycomb sandwich structures by distributed fiber optic sensors Juho Siivola, Shu Minakuchi, Nobuo Takeda* TJCC (UTokyo-JAXA Center for Composites), Graduate School of Frontier Sciences, The University of Tokyo, Kashiwanoha, Kashiwa-shi, Chiba , Japan Abstract Dimpling in the face sheets of honeycomb sandwich structures due to mismatch in the thermal expansion coefficient of the constituent materials, with emphasis on its monitoring and assessment, was studied by utilizing embedded distributed fiber optic sensors. Strain distributions along optical fibers embedded in the face sheet of the sandwich structures were monitored during manufacturing. Based on finite element analysis results, the strain data was interpreted and dimpling monitoring capability was discussed. Simple analytical model to predict the dimple profile and thus assess the dimpling condition from the strain data was also presented and evaluated Published The Authors. by Elsevier Published Ltd. This by Elsevier is an open Ltd. access article under the CC BY-NC-ND license ( Peer-review under responsibility ofthe organizing committee of the 6th APWSHM. Peer-review under responsibility of the organizing committee of the 6th APWSHM Keywords:Honeycomb sandwich structures; dimpling; strain monitoring; fiber optic sensor 1. Introduction Weight reduction of honeycomb core sandwich structures for satellite applications requires the use of thinface sheets and large cell size core. This, however, can lead to notable dimple formation in the face sheets during manufacturing or operation at low-temperature conditions due to mismatch in thermal expansion coefficients of the constituent materials, such as carbon fiber reinforced plastic (CFRP) and aluminum honeycomb core. While dimpling can reduce the mechanical properties of the structures [1], it can also promote failure in adhesion of components attached on the face sheets [2]. A way to detect and assess dimpling of the face sheets is therefore needed to verify the quality and health of the structures. * Corresponding author. Tel.: ; fax: address:takeda@smart.k.u-tokyo.ac.jp Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license ( Peer-review under responsibility of the organizing committee of the 6th APWSHM doi: /j.proeng

2 Juho Siivola et al. / Procedia Engineering 188 ( 2017 ) In previous studies [2,3], fiber Bragg grating (FBG) sensors have been applied on the face sheets to detect dimpling. The measurements are however limited to only one point within a single cell, and thus cannot give detailed information about the dimpling condition and the state of the whole sandwich structure. Rayleigh backscattering based fiber optic distributed strain measurements[4] however seem promising as they can provide the accuracy and high spatial resolution needed to detect the small deformations induced by face sheet dimpling. Face sheet dimpling of honeycomb sandwich structures, including a method to monitor and assess the dimpling, was studied by utilizing embedded distributed fiber optic sensors. Strain distributions along optical fibers embedded in the face sheet of the sandwich structures were monitored during manufacturing. Based on finite element analysis results, effect of adhesive configuration onthe dimpling monitoring capability was discussed. Simple analytical beam model was also presented to predict the dimple profile and thus assess the dimpling condition from the strain data. The model was evaluated against finite element analysis predictions and measurement results. 2. Experimental setup Sandwich panel specimens as shown in Fig. 1 consisting of CFRP face sheets and aluminum honeycomb core were manufactured for the monitoring experiments. 100 mm 100 mm face sheets consisting of three T700S/2592 pre-preg plies in [0/90/0] configuration were cured beforehand in an autoclave. Fiber optic sensors (cladding diameter 100 μm) were embedded in the 0-degree plies of the face sheet. The optical fibers are thus embedded in the top side (outer surface)and in the bottom side (adhesion surface) of the face sheet. The face sheet and the core were adhered in oven using REDUX 312UL adhesive film. Varyingdimple depths were obtained by using one to five layers of adhesive film resulting in different fillet sizes. 3/8 cell diameter aluminum honeycomb core was used. Before adhesion the core was aligned with the face sheet so that the embedded fibers would go through the center of the cells. The specimens were cured for 3 hours at 135ºC while MPa vacuum was applied. Before cooling, the vacuum was released and the specimen were let to cool freely to room temperature. At the end of the heating cycle the adhesive has cured thus adhering together the constituent parts. During cooling, dimples are formed in the face sheets due to the different thermal shrinkage of the core, adhesive and face sheet materials. Strains along the embedded optical fiber sensors were therefore measured during the cooling stage using a Rayleigh backscattering based distributed strain monitoring system (LUNA OBR 4600) capable of millimeter-order spatial resolution. After cooling, the face sheet profile at the locations of the embedded optical fibers was measured using a laser displacement sensor. Fig. 1. Honeycomb sandwich panel specimen with embedded optical fiber sensors

3 188 Juho Siivola et al. / Procedia Engineering 188 ( 2017 ) Fig. 2. Measured strains and face sheet profile along the optical fibers (a) on the top side and (b) on the bottom side of the face sheet. 3. Measurement results Results for the specimen with 5 plies of adhesive film are now presented and discussed in this paper. Average face sheet thickness, adhesive thickness, and fillet size of the specimen were measured to be mm, mm, and mm, respectively. Strains along the optical fibers in the face sheet obtained by comparing states before and after cooling are shown in Fig. 2 along with the measured face sheet profile at the fiber location. The shape of the strain distribution curves indicate that the strains are bending strains induced by dimpling of the face sheet. The strains also correlate well with the measured face sheet profiles with high strain peaks seen at the locations where the face sheet is supported by the core cell walls. Due to possible variations in the adhesive layer thickness and adhesive fillet size, the dimple depths and thus strains vary from cell to cell. The strain curves nowalready give a direct indication of dimpling. A simple estimation of the dimple size can thus be given by correlating the strain range at each cell with the average measured dimple depth. For example, in Fig. 2, around 800 μɛ strain range corresponds with 30 to 40 μm dimple depth. Generally, dimples of 20 μm or more in depth could be clearly identified from the strain data obtained in this study. The strain data is thus a good indicator of the presence of dimpling in the face sheet. 4. Finite element analysis 4.1. Finite element model To further understand the formation of dimpling, the dimpling behavior was also studied by finite element analysis (FEA). Taking symmetry into account, a finite element model based on one eighth of a specimen with 5 plies of adhesive was constructed and analyzed using ABAQUS C temperature change was applied to the model to simulate the cooling phase after adhesion of the core and face sheet. Strains in the middle of the topmost ply of the face sheet and deformation of the surface of the face sheet were then observed in the cell in the middle of the specimen. The geometric parameters of the adhesive layer and adhesive fillet were varied and their effect on the dimpling behavior was observed.

4 Juho Siivola et al. / Procedia Engineering 188 ( 2017 ) FEA results The effect of the presence of the adhesive layer and adhesive fillet was studied first. Fig. 3 (a) shows the predicted dimple shapes and strains for cases with only adhesive layer, only fillets, or both adhesive layer and fillet present. Without adhesive fillets hardly any dimpling occurs. Slight dimpling can only be seen, as the adhesive layer connecting the face sheet to the core cell walls pulls them closer to each other due to thermal shrinkage. Presence of adhesive fillets, on the other hand, produces notable deformation and large strains near the core edges due to local bending of the face sheet induced by shrinkage of the fillets. Varying the thickness of the adhesive and size and shape of the fillets also revealed that thick adhesive layer reduces the dimple depth due to the face sheet/adhesive layer combination trying to bend to the opposite direction during thermal shrinkage. This can be seen in Fig. 3 (a), where the dimple depth decreases while the strains increase slightly from the fillet only configuration with the introduction of adhesive layer. The angle of the fillet can also have notable effect on dimpling. Fig.3 (b) shows how larger dimple angle, i.e. gentler slope, produces deeper dimple and higher strains in the face sheet. When the adhesive shrinks, a gently sloping fillet reaches further to the center of the cell, and can thus pull the face sheet towards the cell walls at larger area, leading to deeper dimple depth. Based on the observations, the fillets and their shape have notable effect on the dimple depth. The dimple depth and the magnitude of the strains are also related. However, the adhesive layer can reduce the dimple depth while increasing strains, which can make solely strain databased assessment of dimpling difficult. Fig. 3. Finite element analysis results showing the effect of (a) the adhesive layer and fillet and (b) the adhesive fillet angle on the strains and deflection of the face sheet.

5 190 Juho Siivola et al. / Procedia Engineering 188 ( 2017 ) Dimpling estimation from strain data 5.1. Beam model Classical beam theory was used to provide some estimate of the dimple profile from the strain data. Following the theory, strains in the beam consist of in-plane strains and bending strains (1) where z is the distance from the neutral plane to the observed plane. Deflection w of the face sheet is then (2) Boundary conditions for simply supported (SS) case when the diameter of a cell is aare, (3) An additional condition needed for clamped (CC) boundary conditions is (4) Choosing asuitable function for the strain, the deflection of the face sheet can then be predicted. Looking at the measured and FEA strain data, a fourth order polynomial function is chosen and the strain is thus defined as (5) where A and B are fitting parameters. Solving the deflection using Eq. 1. gives (6) ɛ 0, C and D are solved by using the boundary conditions in Eq. 3. and 4., (7) For simple supported boundary conditions, ɛ 0 is solved from (8) and therefore (9)

6 Juho Siivola et al. / Procedia Engineering 188 ( 2017 ) Application on FEA data The beam model was first applied to the finite element analysis data to verify its applicability. The strain equation of Eq. 5. was fitted to the data by the maximum and minimum strains as shown by the max/min fitting curve in Fig.4. Face sheet profiles predicted by the model are shown in Fig.5 (a) and (b). While the prediction by the simply supported model is far from the measurement data, the clamped model gives a decent estimation of the dimple profile, as seen in Fig. 5 (a). Accuracy of the prediction however reduces when the effect of the fillets increases, as can be seen in Fig.5 (b). This is due to the shrinking fillets pulling the face sheet near the cell edges, but also due to the increased effect of transverse in-plane stiffness of the face sheet when the deflection becomes larger. Furthermore, as was noted in section 4.2, the fillets can lead to increased dimple depth while the relative strains are reduced. This can be partly compensated by overestimating the strains at the fillet locations, as shown by the overestimated fit curve in Fig. 4. Improvement in the predicted dimple profile by using the overestimated fit, for example for the case with no adhesive layer, is shown in Fig.6. Fig. 4. Strain fitting functions for predicted strain data. Fig. 5. Dimple profiles by FEA and beam model using FEA strain data showing the effect of (a) boundary conditions and (b) adhesive and fillet on dimple profile prediction.

7 192 Juho Siivola et al. / Procedia Engineering 188 ( 2017 ) Application on measurement data Fig. 6. Effect of strain fitting methods on predicted dimple profile Due to the limited resolution of the strain measurements, the strain curve is not accurately captured at all locations, especially at the high strain peak locations where the core and face sheet connect. A periodic function based on Eq. 5 was therefore fitted to the strain data to compensate for the loss of information, especially around the strain peaks. Fig. 7 now shows the strain functions fitted to the strain data using the max/min fit and overestimated fit principles from section 5.2 with the corresponding face sheet deflection profiles predicted using Eq. 6. The clamped boundary conditions give dimple depth predictions that the simply supported boundary conditions. While the dimple depths predicted using the max/min fit strain curve are around half of the measured values, improved prediction can be obtained by using the overestimated fit function. Overall, good understanding of the adhesive layer and fillet geometry, and of the optical fiber location within the structure is required for accurate dimple profile prediction. While the beam model can give rough estimate of the dimpling condition, a more detailed model would be needed that takes the aforementioned effects into account. Fig. 7. Measured strain and face sheet profile data with the fitted strain functions and predicted dimple profiles.

8 Juho Siivola et al. / Procedia Engineering 188 ( 2017 ) Conclusion Face sheet dimpling in honeycomb sandwich structures was monitored by strain distribution measurements of the face sheet using an embedded optical fiber sensor. Based on finite element analysis and a beam model, usability of the strain data to assess the dimpling condition was discussed.the measured strain data was seen to give a clear indication of dimples formed in the face sheets of honeycomb sandwich structures during manufacturing. The relationship between the measured strains and dimple size are however affected by the adhesive layer and fillet configuration and thus no estimation of the dimple size can be made from the strain data only.a simple beam model was presented to provide a rough estimation of the dimple depths. However, a more advanced method that takes the effect of the fillets into account is still needed to accurately estimate the dimple size from the measured strain data regardless of the adhesive layer and fillet configuration. Acknowledgements This work was supported by JSPS Grant-in-Aid for Scientific Research (S) Number References [1] K. Komatsu, T. Kajikawa, N. Miyashita, T. Kamiya, S. Utsunomiya, Bending Rigidity Reduction by the Dimples of Sandwich Plate, Proc. 56th Space Science and Technology Conference, Beppu, Japan, 2012, 1O13. [2] K. Sekine, H. Takeya, H. Seko, M. Takahashi, S. Utsunomiya, Study on Damage Detection for Solar Array Panel using Optical Fiber Sensors, Proc. 53th JSASS/JSME/JAXA Structures Conference, Akita, Japan, 2011, 2B13. [3] N. Miyashita, T. Kajikawa, K. Komatsu, A Study of Dimple Shape on CFRP Sandwich Panels, Proc. 54th JSASS/JSME/JAXA Structures Conference, Kumamoto, Japan, 2012, 1A13. [4] S. Kreger, D. Gifford, M. Froggatt, B. Soller, M. Wolfe, High Resolution Distributed Strain or Temperature Measurements in Single- and Multi-mode Fiber Using Swept-wavelength Interferometry, Proc. 18th International Conference on Optical Fiber Sensors, Cancún, Mexico, 2006, ThE42.