Phased Array Data Manipulation for Damage Tolerance Assessment of Composites using Finite Element Analysis

Richard FREEMANTLE 1, Stefanos GIANNIS 2, Vladimír MATĚJÁK 2

1 Wavelength NDT Ltd, The Paddock, Main Street, Elton, Derbyshire, DE4 2BU, UK

e-mail: richard@wavelength-ndt.com 2 Element Materials Technology, Wilbury Way, Hitchin, SG4 0TW, UK

Abstract In recent years, significant advances have been made in non-destructive testing of composites using techniques such as Phased Array Ultrasound and Computed X-ray tomography allowing three dimensional volumetric imaging of internal damage. In parallel to these applications, new advances in modelling techniques to assess damage tolerance are rapidly being implemented in finite element analysis codes and some assessment of the effect of delamination type damage on the residual load carrying capacity of composite structures can be made. The work presented in this paper aims to join the above technologies in such a way that non-destructive inspection data can be utilised to help generate a quantitative analysis of the structural integrity of the composite component being inspected. Thus, the main objective of this work is to develop the methodology and the associated software tools that will enable the transfer of non-destructive inspection information obtained from composite structures directly into the finite element code for subsequent failure analysis. To validate this methodology two use cases were considered, these being; (a) a flat panel with impact induced damage and (b) an impacted skin/stringer panel. Impact tests were performed and the resulting damage was fully characterised by means of Ultrasound inspection with further validation by Computed X-ray Tomography imaging. The ultrasonic inspection data formulated the basis for defining the topology of the damage which was then utilised in a commercial finite element code via bespoke software routines. A successful validation of the technique was achieved via comparison of the predicted failure modes with the experimental observations. Keywords: Carbon fibre composite, impact damage, ultrasonic NDT, FE modelling, damage tolerance 1. Introduction The use of phased array ultrasonic NDT inspection techniques for the inspection of composite structures has been firmly established in recent years. Due to the unprecedented imaging capabilities and data capture rates that the latest array controllers are delivering, new opportunities are have arisen for the analysis of anomalies and defects using volumetric full waveform data. Whilst this data is useful for a deeper understanding of the nature and cause of the indications that have been detected, questions often remain regarding what the effect of the anomaly or defect will be on the structure. Standard methodologies for addressing this issue range from the application of simple acceptance criteria for maximum allowable defect size, through to detailed classifications of defect type and size depending on the structural importance of the component. The origins of these acceptance criteria are usually based on mechanical test data for the composite material in question, combined with information on the expected loads that the structure may experience during service. This can be derived from full scale testing of the structure, or numerically through structural analysis codes which are widely used in all composite engineering fields. This approach informs the manufacturer and stress and design engineers on what the likely effect of the defect will be. However the generalised process of applying safety factors, to allow for anticipated variations in defect topology and location, worst case loading scenarios, and various measurement uncertainties can be very conservative. This in impacts on many aspects of composite engineering including designing for lightweight applications (structures end up being too heavy), assessing damage tolerance (tendency to replace components rather than monitor or repair) and manufacturing (problems with concession processes and scrap rates).

11th European Conference on Non-Destructive Testing (ECNDT 2014), October 6-10, 2014, Prague, Czech Republic

Figure 1 shows an example flow chart that may be used by composite engineers to develop processes and procedures for assessing the significance of defects and damage[1].

Figure 1: Flow chart for composite damage analysis processes.

In the flow chart the application of NDT inspection and analysis of the data for very large (no question of significant effect) or very small defects (no question that they are minor and of no

significance) is straightforward. However borderline indications in terms of size or location (circled in red in figure 1) would require significant analysis and consideration particularly if the significance of the damage is uncertain (circled in green in figure 1). In these cases detailed Finite Element (FE) modelling may be required combined with mechanical testing of specific examples of that type and size of damage or the component may need to be continually monitored in order to assess the risk of damage growth. Both these activities are time consuming and expensive. This paper looks at the application of volumetric phased array imaging data to develop direct 3D local models of damage that can be used to try and directly calculate the effect of defect using FE analysis using generalised mechanical property data for the composite in question combined with knowledge of the fracture mechanics of the material. The combination of specific modelling and volumetric NDT data combined with generalised material property data opens up the prospect of virtual testing allowing stress and design engineers to easily assess the significance of damage on a case by case basis without resorting to prohibitively expensive and complicated materials testing programmes. Results are presented for NDT measurements on impact damaged composite coupons which demonstrate how calibration of a virtual test model may be achieved by comparing predicted damage behaviour (based on NDT data and FE modelling) with the resultant actual damage observed (again measured by NDT). Further results looking at more complex assemblies are shown for stringer stiffened panels. 2. Methodology 2.1 Sample Geometries and NDT equipment Several geometries were considered during this project which contained simulated production defects or were subjected to impacts to represent in-service damage. Three monolithic carbon fibre geometries were analysed which included a quasi-isotropic carbon fibre plate, a stiffened carbon fibre plate and carbon fibre radius L-section.

Figure 2: Carbon fibre plate, stiffened carbon fibre plate, radius L-section.

Various NDT equipments were used to inspect the above sample geometries including standard pulse echo immersion C-scan, portable phased array inspection utilising wheel probes and flexible array technologies. The results presented in this paper will focus on compression after impact testing of flat panels and skin-stringer pull-off tests to help predict the damage tolerance of impacted carbon fibre structures. Inspections were made with both manual phased array (as would be used in-service) and immersion C-scan techniques (commonly used in production) to assess the use of different inspection approaches and data formats in damage tolerance prediction. 2.2 Flat panel impact trials To establish the proposed protocol for combining NDT data with FE models, initial studies focussed on the flat panel specimens, which were then subjected to impact damage to simulate typical in-service damage events. The carbon fibre plate and stiffened panel were hit by a 2.62 kg impactor mass that was dropped from various heights to produce energy levels ranging between 5 J and 55 J. Prior to impaction, all the samples were inspected with routine pulse echo C-scan to ensure there were no pre-existing defects in the structure. During impact testing the impact and rebound energies were measured via a pair of light emitting diodes fitted to the drop tower test rig. The impact forces at the sample were recorded using a 120 kN piezoelectric load cell. Photographs of the impact setup are shown in the figure 3.

Figure 3: Impact test rig.

2.3 C-scan inspection of impact specimens The impact samples were then inspected using 5 MHz pulse echo ultrasound and C-scan images with full waveform capture were recorded. Typical C-scan results are shown in figure 4. To extract suitable data that can be utilised in a FE model, a good understanding of the laminate specification and composite manufacturing process is required so that the damage locations in the laminate (for example at ply-ply interfaces) can be related to the ultrasound signals. In the case of the impact samples, these were constructed from a Non

Crimp Fabric (NCF) with a mixture of biaxial and triaxial ply groups. A detailed analysis of the impact zones using a micro-focus X-ray Computed Tomography (CT) showed that delaminations occurred at the interfaces between the biax and triax ply groups with little or no visible damage with the groups themselves.

Figure 4: Time-of-flight thickness and amplitude C-scans of a batch of 8 impacts specimens.

Following calibration of the ultrasound velocity in the specimens and using the laminate specification as a guide a comprehensive C-scan gating process was applied to the raw ultrasound data to extract the damage topology in the impact zones at the ply-group interfaces (numbered 1 to 16) as shown in figure 5. The software used to analyse this data is generic in nature and able to accept data files from a number of conventional and phased array C-scan systems.

Figure 5: Typical multiple gating regime to produce multiple C-scan images at each ply group interface.

As the damage develops deeper into the laminate later delaminations are masked by the preceding zone so that only the edges of the damage are imaged. Again micro-focus X-ray

CT showed that in most cases the damage detected at the edge of the impact zone propagated at the same ply group interface within the overall impact zone. On this basis the ply group interface damage zones were analysed and the extent of damage calculated manually as shown in figure 6.

Figure 6: Ply group analysis showing impact damage zones followed by identification of extent of damage

zones (manually defined at this stage).

2.4 Data translation into FE analysis codes In order to automate the processing of the multiple C-scan data shown in figure 6, a Python interfacing script was developed that could interpret the C-scan data files produced by the NDT software and translate it into a format that could be used by the AbaqusTM FE analysis codes. The scripting interface was designed to analyse the C-scan images and used ellipses to locate and register the position and extent of the impact zones at each ply group depth. This process is illustrated in figure 7.

Figure 7: Example of NDT data processing into Finite Element coding for impacted panels.

The proprietary algorithm of model creation and damage incorporation developed by the authors is summarised by the following steps:

1. Initially, the Abaqus™ FE model geometry is created based on geometrical features, lay-up, loading and boundary conditions.

2. Each NDT C-scan slice is read and high amplitude reflections from the ply-group interfaces are thresholded and their point coordinates recorded.

3. These points are imported into Abaqus™/CAE sketch.

4. Extreme x-y locations are found and ellipses are inscribed. Ellipses’ extremes are inherited for each ply interface from top to the bottom, meaning that, only larger ellipses can be created for subsequent layers for typical cone shaped impact damage topologies.

5. The relevant sub-laminate part surfaces are partitioned by the created ellipses and the remaining characteristics of the FE model are defined (i.e. contact interaction, crack growth criterion, boundary conditions, output request, material properties, etc.).

3. Results 3.1 Flat panel (compression after impact) damage tolerance prediction Initial calibration of the FE model was achieved by comparing the FE model predictions for the Stress Strain behaviour to physical measurement of the flat plate specimens as shown in figure 8.

Figure 8: Simulated stress strain curve (solid line) plotted with experimental data (dashed lines).

Based on this initial calibration, and the results of non destructive and destructive testing appropriate damage models were developed for the FE codes. These included in plane (i.e. Hashin damage model) and fracture (i.e. Virtual Crack Closure Technique Criterion) [2] which were implemented for each ply and each interface respectively. The panels were then subjected to compression after impact testing based on ASTM D7137 [3] and the mechanical test results were compared to the FE model predictions based on NDT inspection of the initial

impact damage. These initial results show good agreement between the model and experiment in terms of the Stress/Strain profile and the point at which the panel fails. Analysis of the comparison data showed that ultimate load variation predicted by the model was to within 6% of the experimental results as shown in figure 9. Figure 10 shows how the numerical simulation robustly captures the damage features of failed specimen by comparing typcial experimental results and the simulation results of the outer ply compressive failure modes.

Figure 9: Predicted and measured stress-strain profiles for compression after impact samples.

Figure 10: Comparison between experimentally observed outer ply compressive failure and numerical predictions.

3.2 Skin stringer (pull-off test after impact) damage tolerance prediction Post-impact NDT ultrasonic C-scan data were generated and converted in FE compatible data files for each interface, as previously described. By applying the algorithm generated in Python scripting language the delaminated and still attached parts of the stringer to the skin were analysed to identify the crack front as shown in figure 11.

Figure 11: Interpolation of delamination edge by spline

It was found that the proposed algorithm is not able to accurately substitute fairly complicated delamination shapes without compromising in mesh quality. As the results indicate, ultimate load is highly dependent on delamination shape as shown in figure 12.

Figure 12: Skin/stringer – model with delamination response

In the case of specimen #12B, the simulation was able to predict ultimate load, but the displacement during later stages of loading did not match the observed experimental results. One reason for this deviation is that during testing some degree of delamination branching was observed after the initial growth, resulting in further increase in load. These phenomena were not taken into account in the model. Ultimate load from simulation of specimen #9A is approximately 50% of experimental results.

4. Conclusions The results presented in this paper show the potential for using volumetric C-scan data, such as that produced by portable phased array systems, to accurately predict the damage tolerance of impacted composite panels. A process for manipulation of the C-scan data ply-by-ply to allow it to be utilised in FE modelling has been developed and comparisons of damage prediction based on the C-scan data have compared well to the mechanical tests in the case of flat panels. The results for more complicated assemblies such as a bonded skin-stringer configuration show more variation and further work is required to increase the complexity of the modelling in order to improve the predictive capabilities of this technique.

Acknowledgements The authors would like to acknowledge the support from the UK Technology Strategy Board to conduct this work under the project NDT2DT: New ICT Approach to Automate Non-destructive Testing and Inspections with Evaluation of Damage Tolerance in Composite Structures. References

1. Large Yachts: Examination of Carbon Fibre Masts and Spars, Maritime and Coastguard Agency, July 2011 (http://www.dft.gov.uk/mca/carbon_masts_and_spars_guidance.pdf).

2. K. R, “Virtual Crack Closure Technique: History, Approach and Applications,” Applied Mechanical Reviews, vol. 57, pp. 109-143, 2004.

3. ASTM Standard D7137, “Standard Test Method for Compressive Residual Strength Properties of Damaged Polymer Matrix Composite Plates,” ASTM International, West Conshohocken, PA, 2012.