Medium scale testing and simulation of aircraft crash Ari Vepsä Senior Scientist VTT Technical Research Centre of Finland Espoo, Finland [email protected]
Kim Calonius Research Scientist VTT Technical Research Centre of Finland Espoo, Finland [email protected]
Simo Hostikka Principal Scientist VTT Technical Research Centre of Finland Espoo, Finland Simo.Hostikka @vtt.fi
Markku Tuomala Professor, Tampere University of Technology, Tampere, Finland [email protected]
Summary The topic of an aircraft crash against a nuclear power plant building has received increasing attention during the last decade. VTT has carried out series of jointly funded and designed impact tests and also developed numerical methods and models with which this event can be analysed. In addition, simplified computation methods that are especially useful in parametric studies have been developed at Tampere University of technology (TUT). During the crash, fuel in the fuel tanks is likely to burst out of the ruptured fuel tanks. Fuel burning in pools close to the building might cause significant heating of the structures. Also, the flames and smoke may be transported into the air intakes of the building. These two phenomena are mainly studied via numerical fire simulations with experimentally determined boundary conditions describing the fuel spray behaviour. This paper describes the work carried out in these sub-topics of an aircraft crash. Keywords: nuclear power plant, large aircraft crash, impact testing, structural analysis, fire simulation
1. Introduction Crash of a large commercial aircraft against a building made out of reinforced concrete give rise to several threats for the structures of the building and persons and equipment inside it. Loading is comprised of multiple factors resulting different types of response of the structure. Based on the loading type and harm they induce, parts of an aircraft can be roughly divided as follows: fuselage of the aircraft, motors and other semi-hard parts, fuel bursting out of the desintegrated tanks, and wings. The fuselage of an aircraft is considered to be much more deformable, or softer, than the concrete structure that it impacts against. It causes loading which is considered to be mainly mass flow. Response of the structure to this type of loading is mainly bending, leading in extreme cases to large displacements and rupture of reinforcement longitudinal reinforcement and consequently loss of load bearing capacity of the structure. The hard parts in turn are considered to be much less deformable than the concrete structure that it impacts against. These parts try to perforate the wall causing also scabbing off concrete at the surface of the wall opposite to the impact surface.
The rapid combustion of the aircraft fuel bursting out of the tanks during the impact can create three types of threats to the safe operation of the power plant. First, the thermal impact of the flame ball can cause thermal failures of any unprotected devices and cabling in the vicinity of the plant. The distance at which such a risk is relevant depends on the amount of fuel released and the conditions of the impact. Secondly, significant heating of the structures is possible if the released fuel does not evaporate and burn in the air but forms a burning pool close to the building. One of the motivations for the research has been to develop and use the numerical tools to investigate this question. Finally, the flames and smoke produced may be transported into the air intakes of the building, challenging the safe operation of people and machinery. Wings of the aircraft introduce possible knife-effect to the crash. It has to be verified that the wings do not cut through the already damaged wall. The impact also induces vibration which can propagate from the impact wall to the interior of the building via the structures causing possible malfunction of the equipment. These two phenomena have so far received only minor attention in the testing campaigns and they are therefore not discussed in this paper. In Finland, the nuclear safety authority requires that the new NPP should withstand the crash of a large passenger airplane. Conformity with the requirements is of course shown with verified numerical calculations or empirical formulas. Verification of the models and formulas are carried out against data from experimental tests in which the phenomena included in the impact is simulated in a smaller scale. However, data from this type of tests is not in general publicly available. This scarcity of experimental data and need for development of numerical as well as semi-empirical methods has been the driving force behind the study presented in this paper.
2. Testing In order to obtain relevant and reliable experimental data for verification and development of semiexperimental and numerical methods and models, VTT has carried out series or impact tests. These tests were funded and designed jointly together with domestic as well as foreign partners and carried out within projects called IMPACT I-III. In addition, 5 tests have been carried out at VTT to be used as reference data in an OECD/NEA benchmark experiment IRIS2010 . All these medium scale tests have been carried out using VTT’s impact test-bed. 2.1 Impact test-bed A sketch of the testing apparatus is shown in Fig. 1. The basic idea is to use air pressure in acceleration of the missile to its target speed. Pressure is gradually increased in a pressure Fig.1: Schematic drawing of the test apparatus. accumulator tube (Fig. 1, left) until it reaches a predefined test-specific value. The pressure accumulator tube is separated from an acceleration tube (Fig. 1, middle) by a flange with a set of plastic membranes taped on both sides of it. The missile rests on rails which are located on top of the acceleration tube. When the predetermined value of pressure is achieved, the plastic membranes gets punctured and released air pushes the piston, which is located inside the acceleration tube. The missile is then pushed forward by a fin of the piston. While the missile continues its flight and ultimately hits the target, the piston is stopped by a piston catcher before it hits the target. The slab to be tested is mounted on a steel frame which rests on wooden planks that are used to adjust the vertical position of the slab. The frame is supported in horizontal direction by four supports called back pipes. The apparatus has been designed for missiles with mass up to 100 kg, a diameter between 150 - 400 mm, impact velocities up to 200 m/s; depending on the mass of the missile, and square slabs with side-length of 2.1 m and thickness up to 250 mm. Supporting of the slab is simple supporting either in one or two directions with the span width being 2000 mm. If the slab is replaced with a steel plate, behind which a set of load cells is attached, the apparatus can also be used for estimation of the impact force when using soft missiles.
2.2 Force plate tests An example of impact forces measured during some of the tests are shown in Fig. 2. In all the four tests, the missile is made out of stainless steel pipe with a diameter of 150 mm and slab thickness of 1.5 mm. Three of the missiles were either fully or partially (2/3) filled water with on empty missile. The total mass of the missile and the mass of water are written in the figure. The impact velocity in these tests varied between 96.3 and 99.3 m/s. One of the purposes of these tests was to check the effect of the liquid inside the missile into the Fig. 2: Impact force as a function of time from several force- time function. tests with different amount of water inside the missile. 2.3 Bending behaviour tests In testing, impacts resulting bending type of behaviour of the impacted structure are realized using a hollow, aluminium or stainless steel pipe with a closed end to present the fuselage of the Fig. 3: Left: The missile hitting the slab during test aircraft. Fig. 3 shows photographs taken during and after one of such tests. The TF11. Right: The folded missile after the test. missile used in this test, tagged as TF11, was an empty stainless steel one with inner diameter of 250 mm, slab thickness of 2 mm and mass of 50 kg. The impact velocity in the test was 108.3 m/s. The main parameters that are of interest and thus measured in this type of tests are the displacements, strains on the reinforcement rebars, strains on the front surface of the slab and the support forces. An example of the displacements measured at the back surface of the slab 400 mm to one or other side and 400 mm above the centre Fig. 4: Maximum displacements at x=400mm and point of the slab in some of the bending y=4000mm as a function of impact velocity in four soft behaviour tests are shown in Fig. 4. In missile tests. test TF13, the missile was filled with water (26.25 kg), in the other tests it was empty and of the type described in the previous chapter. The target was a 150 mm thick concrete slab with bending- and shear reinforcement and with simple supporting in two directions. The values of fct in the graph describe the splitting tensile strength of concrete in these tests. 2.4 Punching behaviour tests Punching resistance of a structure is of interest when the impacting object is considerably less deformable than the structure that it impacts against. This is the case for example with a motor of an aircraft impacting against a reinforced concrete slab. In testing, this type of impact is simulated using a thick walled steel pipe filled with light weight concrete and with a solid steel dome at the front and slightly thicker slab as a target than in the bending behaviour tests. The photographs shown in Fig. 5 present the missile and the slab just before the impact takes place
Fig. 5: Left: The missile and the slab just before the impact takes place. Right: The missile penetrating through the slab roughly 9 ms after the beginning of the impact.
Fig. 6: Slab AT2 after the test. Left: Backside of the slab with scabbing. Right: Horizontal cross section of the sawn lower left quarter of the slab.
in one of the punching behaviour tests (left) and the missile penetrating through the slab roughly 9 ms after the beginning of the impact in the same test (right). The photographs shown in Fig. 6 present the backside of the slab after one of the punching behaviour tests (left) and the horizontal cross section of the lower left quartile of the same slab after sawn (right). The effect of shear reinforcement can be seen from the crosssection where a shear cone starts to form only after the central plane of the slab. The main parameter that is of interest in the punching tests is the residual velocity of the missile in case it perforates the slab. The second important parameter is the area of scabbed concrete at the back surface of the slab. The graph in Fig. 7 shows the distance from perforation and the residual velocity of the missile in some of the tests with (tests AT1 and AT1R) and without shear reinforcement (A1 and A1R). The results are presented as a function of cubic unconfined compression strength of concrete. The missile mass was roughly 47.5 kg. The impact velocities are presented in the graph beside each marker.
The main goal of experimental tests described above is the verification and development of structural analysis methods. Generally, the same methods and tools are used for calculating structural behaviour in both the tests and real scenarios. In real and complex applications, numerical simulation is Fig. 7: Distance from perforation (tests A1 and AT1) and often the only option for obtaining residual velocity of the missile (tests A1R and AT1R). detailed results. Simplified methods can be used in preliminary design, for parametric studies, and also as a reliability check of numerical simulation. Two different types of loading cases are considered in this chapter, presenting different possible methods for solving the problem in case. 3.1 Bending behaviour tests Bending behaviour tests are simulated with both a simplified 2 degree of freedom (2DOF) model using Rayleigh damping and a quarter FE model. These are shown on the left and right in Fig. 8. Two degrees of freedom is the minimum requirement to study both bending and shear failure in a plate impacted by a missile. In the model shown in Fig. 8 spring 1 and mass 1 are connected with the global bending deformation and spring 2 and mass 2 are used in describing the local shear behaviour at the plate centre when the missile impacts. The 2DOF model is reported in more detail for example in the work by Saarenheimo et. al. . The FE model has been constructed using shell elements with the reinforcement modelled as smeared layers. Double symmetry has been utilized in the model. Symmetry boundary condition has been applied at the left and bottom edges of the model. Displacement in the plane of the figure
has been restricted at the penultimate row and column of nodes near the top and the right edge of the drawing Loading has been applied as time dependent surface load on the elements located at the impact area (highlighted with a dark colour in the figure). Concrete damaged plasticity has been used as material model for concrete to model its non-linear behaviour. CowperFig.8. A simplified 2DOF model (left) and a quarter FE Symonds viscoplastic model has been used to take into account increase in shell model (right) of a reinforced concrete slab. yield strength of reinforcement at elevated strain rates. No damping or shear reinforcement has been applied in FE analyses. Analyses have been carried out using explicit time integration with Abaqus/Explicit FE code . The loading that is generally used is either the one obtained from a force plate test (if available), the one given by the Riera method  or one of those that take the crushing force of the missile into account by considering folding of the missile . In FE analyses, the missile can also be Fig.9. Central displacements measured in one the tests included into the model. As an example, together with the corresponding values computed with Fig. 9 shows the central displacements FE-model (on the left) and with a simplified 2DOF of a slab computed both with the FEmodel (on the right) with different loading assumptions. model and the simplified 2DOF model using loading assumptions of Riera and a one including also the folding of the missile. Displacement measured in a corresponding test with a two-way supported slab without shear reinforcement is included in the graph for comparison. 3.2 Punching behaviour tests In case of hard missile impact with high velocity, the local compression strength of concrete is exceeded and the missile penetrates into the target first by spalling, i.e. forming a crater, and then by tunneling. If the velocity is high enough, contact force surpasses the remaining shear capacity of the concrete and reinforcement, leading to formation of a punching cone or, more generally, to a concrete fracture in front of the missile. This can ultimately lead to perforation, with the missile possessing certain residual velocity. It is also expected that the concrete protective layer outside rear face bending reinforcement will be fragmented by scabbing mechanism. The following models and results are described in detail in the work by Tuomala et. al . The penetration/perforation theory of Forrestal et al.  concerns the high-velocity impact of hard missiles. This method is based on cavity expansion theory (CET). In this theory, the penetration/perforation process is divided into phases described in general level above. The onedimensional penetration model presented in the same article  has been developed for deep penetration of ogive-nosed projectiles and it is known to yield good results in these cases. Teland and Sjol  have considered projectiles of different nose shapes. They assume that in the case of a flat-nosed or partially flat-nosed projectile, the cavity expansion model used in the tunneling phase by Forrestal et al. is applicable from the beginning of impact. This model needs no additional empirical constants, making model application straightforward. The UK Atomic Energy Authority  has developed an equation for calculating missile perforation velocity. As an example, let us consider a square reinforced concrete slab with a span of 2 m and a thickness of 0.25 m, tested at VTT for IRIS_2010 benchmark . Bending reinforcement consists of bars
with a diameter of 10 mm and spacing of 90 mm on both faces and in each direction and yield strength of 560 MPa. The cylindrical unconfined compressive strength of concrete is 59 MPa. The rigid missile has a diameter of 0.168 m, mass of 47 kg and impact velocity of 135 m/s. The methods described above predict perforation. Assuming a shear cone angle of 60o, the residual velocity is 39 m/s. The corresponding measured values in three repetition tests were 34, 45 and 36 m/s. Solid finite elements have to be used for simulating perforation, as distinct from shell elements used in bending case in Chapter 3.1. In addition, element erosion has to be applied to overcome extensive element distortion and enable the missile to perforate the slab. Another possibility is to use so-called hydrocodes or discrete element method, for instance. A highly nonlinear dynamic analysis, in which the considered target slab (2 000 000 linear solid Fig 10. Deformed shape of the FE model 19 ms elements), its reinforcement bars, and the after the impact (left) and one of the damaged missile are all modelled as separate parts having contact definitions between each other, is test slabs (right). conducted with Abaqus/Explicit FE code version 6.11-1 . The deformed shape of the model in Fig. 10 shows that damage is fairly realistically simulated. The residual velocity, 50 m/s, is slightly overestimated.
For about ten years, VTT has been one of the organizations developing a Computational Fluid Dynamics –based software for fire simulations, called Fire Dynamics Simulator . The FDS program solves the weakly compressible form of the Navier-Stokes equations using Large Eddy Simulation for turbulence and Eulerian-Lagrangian formulation of the two-phase flows. The FDS program was used by Luther and Müller  to investigate the spreading of flames and smoke around a generic NPP geometry. Instead of introducing the fuel into the computation in the form of liquid droplets, they described the fuel source as a gas-phase inflow boundary condition with a corresponding mass flow. With this method, it is possible to predict the overall heat and smoke transport, but the dynamics of the early flame movement cannot be captured because the effects of the high momentum fuel spray are not included. Our approach has been to describe the fuel inflow boundary condition using the Lagrangian droplets carrying the mass and momentum of the fuel. The main challenges of the approach have been the determination of the necessary input values, validation of the method, and the computational cost and difficulty of the high-speed two-phase flow. To tackle the two first challenges, some liquid-filled missiles have been included in the VTT’s impact experiments to investigate both the overall spray behaviour and the fine details of the spray . The applications of the simulations include the work of Sikanen et al.  where the FDS tool was used to determine the fraction of the released aircraft fuel that ends up in the pool as a function of the impact conditions. One impartant restriction of the model, in the context of the aircraft impacts, is the lacking capability to predict the increase of pressure by detonation. Two-phase flow is computed using Eulerian-Lagrangian concept where the liquid droplets have zero volume in Eulerian space. This means that the liquid-liquid interactions, that may have importance in dense sprays and early phase of the impact, cannot be taken into account. Ignoring buoyancy, lift and forces arising from fluid acceleration, the motion of single spherical droplet is governed by the equation of motion dmd vd mg 12 g C D Aeff vrel vrel . (1) dt where md and vd are the droplet mass and velocity, respectively, g is the gravitational acceleration, v d v g is the velocity of the droplet relative to the g is the density of the surrounding gas, v rel surrounding gas, Aeff is the projected surface area of the droplet, rd is the radius of the drop-let, and CD is the drag coefficient. The evaporation of the droplets is calculated using mass-transfer
correlations. The combustion of the evaporated fuel is simulated using a mixing-controlled assumption of the reaction rate. In an example simulation of the plant-scale impact, 10 ton of heptane was released during a 0.1 s period. The speed of the droplets was set to 250 m/s, which according to the experimental observations corresponds to an aircraft speed of 125 m/s. The direction pattern of the droplets was determined from video recordings of the Sandia National Laboratories full-scale aircraft impact tests  and the median drop size to 30 m. Instantaneous gas temperatures in the central plane of the fire ball are shown in Fig. t=1.000s t=2.000s 11. The thermal lift-off phase of the fire ball takes about five seconds from the impact. During that time, the heat fluxes to the structures are very high, but due to the short duration of the exposure, thermal impact remains quite modest. In practice, the massive structures of any industrial plant would have no problem to withstand the thermal impact from the fire ball. To get insight to the fraction of the fuel that continues burning in a local pool fire, a series of t=3.000s t=6.000s simulations was performed to investigate the Fig.11: Instantaneous temperature fields in the effect of the pooling fraction on some of the fire simulation. impact parameters. Fig. 12 shows the accumulated fraction as a function of the impact height. With very low impact heights, as much as half of the fuel ends up in the pool. When the height gets above 25 m, the fraction gets close to zero, meaning that all the fuel has sufficient time to evaporate before reaching the ground. However, the uncertainty of this prediction is very high, mainly due to the limited knowledge of the droplet diameters and the limited capability of the combustion model to describe the fast combustion spray. For instance, increasing the median drop size from 30 Fig. 12: Accumulated fraction of the initial fuel mass m to 100 m would increase the pooling in the pools as function of impact height. fraction with about 60 %. To overcome these uncertainties, we have recently performed more experiments where we have carried out detailed measurements of the drop size distributions. These results will, in the future, help to reduce the uncertainties of aircraft impact the fire risk assessment.
Discussion and conclusions
The topic of aircraft crash against a nuclear power plant structure has received increasing attention during the last decade. VTT has participated actively in discussion by carrying out series of medium-scale impact tests, funded and designed jointly together with domestic as well as foreign partners. Development work of computation methods and models has been carried out together with TUT. With these methods and models, the response of the structure to such an impact can be predicted. In addition, VTT has carried out development of Fire Dynamics Simulator code and carried out analyses with it simulating burning of fuel released during an aircraft crash. The research has mainly given confidence on the methods, but at the same time revealed some
deficiencies and inaccuracies in them. The civil engineering community is still in a learning phase regarding the numerical analysis of concrete structures impacted by a missile. The topic is multidisciplinary and the assessment should be based on several different types of methods.
Part of the work described in this paper has been carried out with the funding of Finish programme on nuclear safety (SAFIR 2014). This funding is gratefully acknowledged.
Nuclear energy agency, Committee on the safety of nuclear installations, “Improving Robustness Assessment Methodologies for Structures Impacted by Missiles (IRIS_2010). Final Report.” Nuclear energy agency, Committee on the safety of nuclear installations, NEA/CSNI/R(2011)8, 2011 SAARENHEIMO A., TUOMALA M., VÄLIKANGAS P., and VEPSÄ A., ”Sensitivity Studies on a Bending Wall of IRIS_2010 Benchmark Exercise”, Journal of Disaster Research, Vol. 7, No. 5, 2012, pp. 629-637 ABAQUS Theory Manual, version 6.10-1. Dassault Systemes, 2010. RIERA J.D., “On the stress analysis of structures subjected to aircraft impact forces”, Nuclear Engineering and Design, Vol 8, 1968, pp. 415-426. TUOMALA M., CALONIUS K., KUUTTI J., SAARENHEIMO A., VÄLIKANGAS P., ”Sensitivity Studies on a Punching Wall of IRIS_2010 Benchmark Exercise”, Journal of Disaster Research, Vol. 7, No. 6, 2012, pp. 775-785. FORRESTAL M.J., ALTMAN B.S., CARGILE J.D. and HANCHAK S.J., “An Empirical Equation for Penetration Depth of Ogive-nose projectiles into Concrete Targets”, International Journal of Impact Engineering, Vol. 15, 1994, pp. 395-405. TELAND J.A., and SJOL H., “Penetration into Concrete by Truncated Projectiles”, International Journal of Impact Engineering, Vol. 30, 2004, pp. 447-464. BARR P., “Guidelines for the Design and Assessment of Concrete Structures Subjected to Impact”, AEA, 1990. VEPSÄ A., SAARENHEIMO, A., TARALLO, F., RAMBACH, J.-M., and ORBOVIC N., “Impact Tests for IRIS_2010 Benchmark Exercise”, Journal of Disaster Research, Vol. 7, No. 5, 2012, pp. 619-628 MCGRATTAN K.B., HOSTIKKA S., FLOYD J.E., BAUM H., and REHM R., “Fire Dynamics Simulator (Version 5): Technical Reference Guide.”, NIST SP 1018-5. NIST Special Publication 1018-5. October 2007. Building and Fire Research Laboratory. National Institute of Standards and Technology. LUTHER W., MÜLLER W.C., “FDS Simulation of the Fuel Fireball From a Hypothetical Commercial Airliner Crash on a Generic Nuclear Power Plant”, Nuclear Engineering and Design, Vol. 239, 2009, pp. 2056-2069. SILDE A., HOSTIKKA S., and KANKKUNEN A., “Experimental and Numerical Studies of Liquid Dispersal From a Soft Projectile Impacting a Wall”, Nuclear Engineering and Design Vol. 241, 2011, pp. 617-624. doi: 10.1016/j.nucengdes.2010.07.033 SIKANEN, T., HOSTIKKA, S., and SILDE, A., ”Experimental and Numerical Simulations of Liquid Spreading and Fires After Aircraft Impact”, 21st International Conference on Structural Mechanics in Reactor Technology (SMiRT 21) - 12th International Pre-Conference Seminar on "FIRE SAFETY IN NUCLEAR POWER PLANTS AND INSTALLATIONS", 13-14.9.2011, München. SUGANO T., TSUBOTA H., KASAI Y., KOSHIKA, N. ORUI, S., VON RIESEMANN, W.A., BICKEL, D.C. PARKS, M.B. 1993. Full-scale aircraft impact test for evaluation of impact force. Nuclear Engineering and Design 140, pp. 373-385.