Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Predicting the vulnerability of spacecraft components: modelling debris impact effects through vulnerable-zones

Predicting the vulnerability of spacecraft components: modelling debris impact effects through... The space environment around the Earth is populated by more than 130 million objects of 1 mm in size and larger, and future predictions shows that this amount is destined to increase, even if mitigation measures are imple- mented at a far better rate than today. These objects can hit and damage a spacecraft or its components. It is thus necessary to assess the risk level for a satellite during its mission lifetime. Few software packages perform this analysis, and most of them employ time-consuming ray-tracing methodology, where particles are randomly sampled from relevant distributions. In addi- tion, they tend not to consider the risk associated with the secondary debris clouds. The paper presents the development of a vulnerability assessment model, which relies on a fully statistical procedure: the debris uxes are di- rectly used combining them with the concept of vulnerable zone, avoiding the random sampling the debris uxes. A novel methodology is presented to predict damage on internal components. It models the interaction between the components and the secondary debris cloud through basic geometrical Corresponding author Email addresses: mirko.trisolini@polimi.it (Mirko Trisolini), H.G.Lewis@soton.ac.uk (Hugh G. Lewis), camilla.colombo@polimi.it (Camilla Colombo) Present address: Politecnico di Milano, Via La Masa 34, 20156, Milano, Italy Preprint submitted to Advances in Space Research March 13, 2020 arXiv:2003.05521v1 [astro-ph.IM] 10 Mar 2020 operations, considering mutual shielding and shadowing between internal components. The methodologies are tested against state-of-the-art software for relevant test cases, comparing results on external structures and internal components. Keywords: space debris, debris impact, debris cloud, vulnerable zone, spacecraft vulnerability, penetration probability 1. Introduction Since the launch of Sputnik 1 in 1957, the space around the Earth and beyond has been the theatre of remarkable achievements but has also su ered from continuous exploitation. Decommissioned satellites, spent upper stages, mission-related objects, and fragments generated by collisions and explosions of satellites and upper stages have polluted the space environment in the form of space debris. Space debris is thus considered as a major threat to space mission; in fact, debris of just 1 cm can cause the break-up of a satellite (Putzar and Sch afer, 2006; Stokes and Swinerd, 2005), and smaller particles can still have enough energy to produce failures on components critical to the mission operations and scienti c objectives. Recent studies have shown a constant increase in the population of space debris, and the amount of debris is expected to keep growing in the next years (Radtke et al., 2017; Lewis et al., 2017a,b, 2012). The major space-faring nations and international committees have thus proposed a series of debris mitigation measures to protect the space environment (O'Connor, 2008; Sch afer et al., 2005; Inter- Agency Space Debris Coordination Committee and Committee, 2007). The implementation of these mitigation measures can help to mitigate the impact of human activities on the space environment. Nonetheless, the danger posed by possible impacts of space debris on spacecraft structures and components has to be addressed and regarded as a mission design driver as a single impact can compromise an entire mission (Grassi et al., 2014; Stokes et al., 2012; Christiansen et al., 2009). This is particularly important for missions in speci c orbital regions, such as highly inclined orbits, which are more densely populated by space debris and im- pacts can be more frequent (Liou and Johnson, 2006). The space debris population accounts for about 29000 objects larger than 10 cm, 750000 ob- jects between 1 cm to 10 cm, and 166 million objects from 1 mm to 1 cm 2 (Space Debris Oce, 2017). In Low Earth Orbit (LEO) orbits, only particles larger than 10 cm can be tracked (Schaub et al., 2015). Consequently, only for these particles collision avoidance manoeuvres are e ective in preventing damage to satellites. For the other untraceable yet dangerous debris, it is necessary to adopt speci c measures to ensure the mission survivability. To do so we need to assess the amount of damage caused by space debris to spacecraft components and identify the most vulnerable ones. Few software packages are currently available to perform the vulnerability assessment of a mission. Examples are ESABASE2/DEBRIS (G ade and Miller, 2013), Na- tional Aeronautics and Space Administration (NASA) BUMPER (Bjorkman et al., 2014), European Space Agency (ESA) Debris Risk Assessment and Mitigation Analysis (DRAMA) (Gelhaus et al., 2014, 2013; Martin et al., 2005b,a), SHIELD (Stokes et al., 2000; Stokes and Swinerd, 2005; Stokes et al., 2012), and Particle Impact Risk and Vulnerability Analysis Tool (PI- RAT) (Gulde et al., 2016; Kempf et al., 2013, 2016). These codes have a common structure, where the satellite is modelled through a geometric rep- resentation (which can be more or less detailed). The geometry is usually schematised through a set of panels. Each panel is then assigned the proper- ties and characteristics required for the impact analysis, such as the material, the shielding con guration, the thickness, etc. (Grassi et al., 2014; G ade and Miller, 2013; Stokes et al., 2000; Welty et al., 2013). An environmental model (e.g. Meteoroid and Space Debris Terrestrial Environment Reference (MAS- TER)) is then used to generate the debris uxes a ecting the spacecraft. The nature and frequency of the impacts on the structure and components of the spacecraft is evaluated and the damage associated with them is computed using semi-empirical relationships (i.e. Ballistic Limit Equations). Finally, the results of the analysis can be obtained in the form of number of impacts, number of penetrations, impact probability, and penetration probability. These software packages perform the vulnerability analysis with di erent levels of detail and techniques. The two main aspects to be considered are the modelling of the satellite geometry and the computation of the damage on internal components. For instance, DRAMA is only used for preliminary analysis. It only allows the de nition of simple geometrical shapes through a limited number of panels, which are created specifying their area, nor- mal, and shielding properties. Also, the impact analysis only involves the outer structure of the satellite and does not consider internal components. BUMPER instead generates a high- delity Finite Element Model (FEM) of the spacecraft structure; however, it cannot take into account the im- 3 pacts on internal components. SHIELD and ESABASE2/DEBRIS both use a panelised model of the satellite structure, and a ray-tracing methodology to simulate the impact of debris particle on the spacecraft. They both can compute the e ects of debris impacts on internal components. The method- ologies used by the two models for this task are, however, very di erent. In SHIELD, after the rst impact on the external structure, the projectile is con- sidered still intact and following the same direction of the initial impact. In ESABASE2/DEBRIS the assessment for internal components is not directly supported and requires a workaround: the user has to create a panelized rep- resentation of the internal components, without the external structure. The contribution of the external structure is then taken into account assigning a multiple-wall Ballistic Limit Equation (BLE) to each panel constituting the internal components. A recent integration to this methodology has been proposed and is under development (Bunte et al., 2017), which considers the characteristics of the debris cloud that is formed inside the spacecraft after the rst impact. In this methodology, the characteristic distributions relative to the debris cloud velocity, mass, and spatial distribution are obtained from available experimental data. The distributions are then sampled to generate rays representing the cloud fragments. This type of methodology is, however, time consuming as it requires the generation of thousands of rays and the veri cation of the impact with the components for each one of them. Be- sides, as the rays are generated through sampling, several simulations must be performed to have statistically meaningful results. Finally, the tool PI- RAT by the Fraunhofer Institute, also relies on a complete 3D modelling of the satellite structure and internal components and on the use of environ- mental models such as MASTER and BUMPER. However, di erently from SHIELD and ESABASE2/DEBRIS, uses a deterministic approach instead of a probabilistic one (Kempf et al., 2016; Gulde et al., 2016; Welty et al., 2013). This approach is based on a three-wall geometrical assessment to de ne threat directions, which generate the BLE con guration for the components. In par- ticular, it uses the Schafer-Ryan-Lambert (SRL)-BLE, so that shielding and partial shielding can be taken into account, as well as primary and secondary impacts. It has also been applied in the Concurrent Design Facility at ESA for preliminary vulnerability analyses. The work presented in this paper is devoted to the introduction of a novel methodology to assess the penetration probability on internal spacecraft com- ponents in an intrinsically statistical fashion (Welty et al., 2013) by exploiting 4 and extending the concept of vulnerable zones (Putzar and Sch afer, 2006). The need for ray tracing methodology is removed and so is the need for several Monte Carlo runs to obtain robust results. Also, the implementa- tion of vulnerable zones is coupled with the formation of secondary ejecta clouds after the rst impact. The methodology relies then on geometrical operations to compute the interactions of the debris cloud with the internal components. The presented methodology thus evolves from SHIELD, where the particle remains intact, by considering the cloud contribution, and avoids computationally expensive ray-tracing procedure as in Bunte et al. (2017). The methodology has been developed as a part of a broader study (Trisolini et al., 2018a,b, 2016, 2015), where a fast assessment of the vulnerability of simpli ed spacecraft con gurations was required. We rst describe the over- all model developed, for then focusing on the speci cs of the computation of the vulnerability of internal components. The results are then compared with DRAMA and ESABASE2/DEBRIS for relevant test cases. 2. Model outline To properly describe the developed methodology, it is important to un- derstand the entire vulnerability model in all its parts. The model evaluates the vulnerability of simpli ed spacecraft con gurations, considering the ef- fects on both the external structure and on the internal components (Trisolini et al., 2018a,b). It takes as inputs the spacecraft con guration, the mission scenario, and the debris uxes from a space environment model, and returns the impact and penetration probability on the external structure and inter- nal components. The main novelties in the model are the schematization of the debris environment through vector ux elements (Section 2.2.1), the implementation of a statistical methodology to compute the impact and pen- etration probabilities through the use of vulnerable zones (Section 3.2), and the development of a fully geometrical methodology for the assessment of the mutual shielding between internal components (Section 4) without relying on ray tracing methodologies. 2.1. Spacecraft con guration The spacecraft con guration is de ned through a set of elementary shapes such as boxes, cylinders, spheres, and panels. For each element in the con- guration, its main properties must be provided, such as the shape, the size, 5 the mass, and the position inside the spacecraft. An example of the con- guration entries is provided in Table 1; the main body of the satellite is always the rst entry and is positioned at the centre of the reference frame. The x -axis points towards the direction of the orbital velocity (RAM direc- tion), the z -axis points away from the Earth, and the y -axis follows from the right-hand rule. Table 1: Example of spacecraft con guration structure used by the model. ID Name Shape m l r w h Pos. (kg) (m) (m) (m) (m) (m) 0 Spacecraft Box 2000 3.5 n/a 1.5 1.5 n/a 1 Tank Sphere 15 n/a 0.55 n/a n/a (-1,0,0) 2 BattBox Box 5 0.6 n/a 0.5 0.4 (0,0,1) ... ... ... ... ... ... ... ... ... The satellite structure can be de ned with a single material and a uni- form thickness; alternatively, di erent materials, thickness, and shielding types can be assigned to each face. Three types of shielding can be used in the current version of the model: single wall shields, Whipple shields, and honeycomb sandwich panels (Ryan and Christiansen, 2010). For the internal components, the position must be speci ed; the components can be free in- side the main structure, or they can be attached to the external panels. The orientation of a component can also be speci ed, but only along the main axes of the satellite (x, y, and z ). Once the geometry of the spacecraft has been nalised, the software extracts all the relevant information needed for the vulnerability assessment, such as the shield type, the stand-o distances, the vulnerable zones extents, etc. Using this information together with the vector ux elements (Section 2.2.1), and the ballistic limit equations (Section 4.2), the vulnerability of each internal component can be computed. 2.2. Space environment model To assess the impact of space debris on a satellite, it is necessary to know the characteristics of the environment the satellite has to face during its lifetime. The characteristics and number of debris particles impacting a spacecraft are dependent upon the mission scenario, mainly its altitude and inclination. The present model uses ESA MASTER 2009 (Flegel et al., 2009; Gelhaus et al., 2011) to obtain the description of the debris environment via 6 ux predictions on user-de ned target orbits (Trisolini et al., 2018a). The uxes provide the number of particles per unit area per year that impact the spacecraft, given the speci c mission scenario (i.e. orbital parameters). The distributions needed for the survivability computation are the ux vs particle diameter, ux vs impact elevation, ux vs impact elevation vs impact azimuth, and the ux vs impact velocity vs impact azimuth. This set of distributions is sucient to perform the spacecraft vulnerability analysis. Fig. 2 and 1 represent an example of the ux distributions given by MASTER, where the di erential ux as a function of the impact velocity and impact azimuth and the cumulative ux as a function of the particle diameter are presented. The uxes refer to a Sun-synchronous orbit (SSO) with semi- major axis of 7185 km, eccentricity of 0.001, and inclination of 98 for one year with starting epoch 1st May 2009. 1 0 1 0 1 0 1 0 -1 1 0 -2 1 0 -3 1 0 -4 1 0 -5 1 0 -6 1 0 -4 -3 -2 -1 1 0 1 0 1 0 1 0 Par t icle d iam et er (m ) Figure 1: Example of MASTER-2009 cumulative ux vs particle diameter distribution. Orbit characteristics: a = 7185 km, e = 0.001, i = 98 . Epoch: 1st May 2009. Mission lifetime: 1 year 2.2.1. Vector Flux Elements Despite MASTER-2009 provides the description of the debris environ- ment through uxes, this information cannot be directly used in the compu- tation of the vulnerability of a spacecraft. In fact, assessing the criticality of the impact of a debris fragment requires the knowledge of the particle Particle Flux (1/m yr) Figure 2: Example of MASTER-2009 di erential ux vs impact azimuth vs impact velocity distribution. diameter, velocity, and direction. These data can be obtained by directly sampling the ux distributions provided by MASTER to generate a set of impacting particles, which are shot towards the satellite. This procedure is at the core of ray tracing methods, which are computationally expensive and re- quire a large number of simulations in order to be statistically robust (Stokes et al., 2000; Stokes and Swinerd, 2005). The methodology implemented in this work, instead, follows a novel approach to bypass the use of such com- putationally expensive procedures, and at the same time have a suciently accurate description of the debris environment. The compromise is achieved by characterising the debris environment surrounding the spacecraft with vector ux elements. The role of vector ux elements is to summarise the information relative to a speci c angular sector of space around the satel- lite (expressed in terms of impact azimuth and elevation) as provided by MASTER-2009. In fact, each vector element is associated with a value of the particle ux, particle diameter, impact velocity, and impact direction (Fig. 3) and these values are then used for the entire sector. The complete procedure to generate the vector ux elements is described in detail in Trisolini et al. (2018a). In summary, it consists of subdividing the space around the satellite into a set of angular sectors of impact azimuth and 8 Figure 3: Vector ux elements representation. In this example each vector carries the value of the impact velocity for the angular sector it is referred to. elevation. Then, for each angular sector, the diameter, the impact velocity, and the impact direction are extracted from the MASTER-2009 distribu- tions. As in each angular sector these quantities have a distribution and not a single value, we need to use a criterion to select the value to be as- sociated to the vector ux element. The criterion we use is the weighted average of the quantity of interest, using as weights the corresponding values of the particle uxes. For a generic quantity x, inside the angular sector [Az ; Az ; El ; El ] we thus have i i+1 i i+1 x ' k k k=1 x  = i 2 f1; :::; N g; j 2 f1; :::; N g; (1) ij Az El k=1 where x  is the weighted average value of the quantity x referred to the ij ij angular sector, ' is the value of the ux corresponding to the quantity x , which is obtained from a histogram equivalent to the one of Fig. 1 but limited to the considered angular sector, and M is the number of subdivisions used to obtain the ux histogram from MASTER-2009. N , and N are Az El the number of angular sector subdivision for the azimuth and the elevation angle respectively. Using this procedure for all the quantities of interest we 9 obtain a vector in the form of Eq. (2), which represents all the properties de ning a vector ux element referred to an angular sector. '  v  Az El d f = (2) ij p;ij ij ij p;ij ij The use of the schematisation introduced by the vector ux elements aims at simplifying the computation and have a computationally ecient proce- dure. In fact, with this, we can limit the number of times we check for penetration using BLEs and, in particular, the number of geometrical oper- ations that are required to assess the contribution of the clouds associated to the secondary ejecta. A better precision could be achieved directly using the binned output generated by MASTER-2009. However, this would come at the cost of a higher computational time. 2.3. Ballistic Limit Equations When debris particles hit a spacecraft, the type of impact and its ef- fects need to be assessed. Such an assessment is performed using a set of experimentally derived analytical expressions referred to as Ballistic Limit Equations (BLEs) (Ryan and Christiansen, 2010, 2011; Christiansen et al., 2009; Sch afer et al., 2005). Given the impact speed, the impact angle, the projectile density, and the target characteristics, BLEs allow the computation of the critical particle diameter, which is the minimum diameter an impact- ing particle must have to produce damage on the speci ed target. In general, BLEs have di erent expressions as a function of the type of impact. Three di erent types of impacts are distinguished depending on the relative impact velocity: ballistic, hypervelocity, and shatter regime (Ryan and Christiansen, 2010; Ryan et al., 2008; Ryan and Christiansen, 2011). The expressions for the BLEs are further divided by the shielding type: single-wall, double-wall, triple-wall, and advanced shielding concepts all have dedicated expressions (Ryan and Christiansen, 2010). Single-wall BLEs are subdivided accord- ing to the material of the shield, from classical metallic materials such as aluminium, titanium, and stainless steel, to Carbon Fibre Reinforced Plas- tics (CFRP), breglass, glass, and polycarbonate. Dual-wall shields include two main con gurations: the standard metallic Whipple shield (Ryan and Christiansen, 2010), and the Honeycomb Sandwich Panel (HC-SP) (Ryan et al., 2008). A commonly used triple-wall BLE for the analysis of space- craft vulnerability is the SRL equation Eqs. (3-4) (Sch afer et al., 2008; Ryan et al., 2008; Putzar and Sch afer, 2006; Grassi et al., 2014; Welty et al., 2013). 10 The SRL BLE allows the computation of the critical diameter for equipment placed inside the main structure of a satellite by considering the last plate of the triple-wall con guration as the face of the target equipment (Putzar and Sch afer, 2006). As this BLE is versatile and can be used for internal components, we will use it for the reminder of the work. The expression of the SRL BLE for the ballistic regime (v  V  cos ) is p LV 0:5 " # 18=19 1 0:5 y t + t  + t b ob K 40 3S d = (3) c;b 2=3 0:5 0:6 cos     v where v is the particle velocity relative to the spacecraft,  is the particle p p density,  is the impact angle (angle between the relative velocity vector and the normal to the impacted surface), V is the ballistic regime transition LV velocity, K and  are tting factors whose values are summarised in Table 3S A.7, t is the bumper plate thickness, t is the outer bumper thickness, t is b ob w the rear wall thickness, and  is the yield strength of the rear wall material. In the hypervelocity regime (v  V  cos ) instead we have p HV 1=3 2=3 1=3 1:155 S  t + K  t + K  S  t cos b tw w S2 1 2 w d = c;h 2=3 1=3 1=9 2=3 K      v  cos p p 3D ob (4) where V is the hypersonic regime transition velocity. K , K , K , HV tw S2 3D , , and are tting factors (Table A.7);  is the outer bumper density; S ob 1 and S are the spacing between the outer bumper and the bumper plate, and the space between the bumper plate and the rear wall respectively. Finally, in the shatter regime (V  cos   v  V  cos ) a linear interpolation LV p HV is used between the critical diameters obtained in the ballistic and in the hypervelocity regimes. The SRL BLE has been used throughout this work for the computation of the critical diameter for impacts on components insides the spacecraft. Single-wall and dual-wall BLEs can also be used by the model to predict impact damage on external structures and components. 2.4. Impact and penetration probability assessment The procedure for the computation of the impact probability (P ) and imp penetration probability (P ) depends on the characteristics of the environ- pen ment surrounding the satellite that are the particle ux, velocity, direction, 11 and diameter. In addition, the orientation of the surface considered plays an important role as it in uences the area that is visible by a speci c vector ux element. Finally, even the mission lifetime is an important parameter to be considered as the more a spacecraft resides in a speci c debris environment, the more it will be exposed to the debris uxes. The outline of the procedure is presented in Fig. 4 and the detailed description can be found in Trisolini et al. (2018b). Start Vector flux element impacts panel? No Yes Compute critical diameter Compute critical flux Compute panel Store panel vulnerability vulnerability Compute overall vulnerability End Figure 4: Flow diagram for the vulnerability computation. To evaluate the impact and penetration probabilities, the structure of the spacecraft is schematised as a set of panels. For each panel, it is necessary to check which vector ux element can actually impact the panel. This is achieved using a simple visibility which checks that v  n < 0, where v is i j i the i -th vector ux element and n is the normal to the j -th panel in the set. If a vector ux element impacts the panel, its contribution to the overall impact probability is then evaluated. By using Poisson statistics that is by assuming that debris impact events are statistically independent, the impact Next panel Next vector flux element Next vector flux element probability of a vector ux element on an a panel can be expressed as ij P = 1 exp '  S  T (5) i ij imp where ' is the ux associated to the i -th vector ux element, S is the i ij projected area of the j -th face on the direction of the i -th vector ux element, and T is the mission time in years. For the computation of the penetration probability, the classic approach involving BLEs can be adopted (Kuiper et al., 2010; Reimerdes and Wohlers, 2001; Welty et al., 2013; Stokes et al., 2000; Grassi et al., 2014; Bunte et al., 2009; Stokes et al., 2012). Substituting into the relevant BLE the velocity and direction of the vector ux elements, the critical diameters are computed. The critical diameter is then used to compute the corresponding critical ux i.e. the particle ux corresponding to diameters greater than the critical diameter. Figure 5: Graphical representation of the methodology for the computation of the critical ux. Orbit characteristics: a = 7185 km, e = 0.001, i = 98 . Epoch: 1st May 2009. Mission lifetime: 1 year To compute the critical ux, we use the MASTER-2009 distribution of the cumulative ux vs particle diameter, from where the it can be extracted (Fig. 5). Finally, the expression for the penetration probability is given by Particle Flux (1/m yr) ij P = 1 exp '  S  T (6) c;i ij pen where ' is the critical ux associated to the i -th vector ux element. c;i It is important to highlight the simpli cation introduced with this pro- cedure: as the global distribution of cumulative ux vs diameter is used for computing the critical diameter, the obtained ux is the overall ux for the entire range of azimuth and elevation angles. This ux cannot be directly used to compute the penetration probability associated to a single vector ux element as to each one of them is associated a value of the particle ux that is dependent upon the directionality, i.e. impact elevation and impact azimuth. It is thus assumed that the shape of the curve describing the relation between the particle ux and the particle diameter maintains the same shape through the di erent values of impact azimuth and impact elevation. By doing so, we obtain the critical ux for a vector ux element as a fraction of the total critical ux, which is proportional to the ux associated to that particular vector ux element. With this assumption, the critical ux associated to a vector ux element can be computed as follows ' = '  (7) c;i i tot where ' is the total debris ux and ' is the overall critical ux as tot c extracted from Fig. 5. Figs. 6 and 7 show the distribution of the particle ux for the particle diameter and the impact azimuth and elevation respectively. From these maps can be better observed that the assumption is not always satis ed: there are regions where the inverse exponential trend of the overall particle ux with the diameter (Fig. 5) is maintained. In these regions, the scaling assumption can be reasonable; however, it is less accurate for other regions. The simpli ed nature of the proposed method is aligned with the introduced simpli cation; nonetheless, future development of the methodol- ogy will revise this assumption, including a more accurate determination of the critical ux. Finally, the overall impact and penetration probabilities (Eqs. (8) and (9) can be computed iterating over the entire set of vector ux elements and spacecraft panels as follows: N N panels fluxes Y Y ij P = 1 1 P (8) imp imp j=1 i=1 14 Figure 6: Particle ux vs Impact elevation vs Object diameter. Orbit characteristics: a = 7185 km, e = 0.001, i = 98 . Epoch: 1st May 2009. Mission lifetime: 1 year Figure 7: Particle ux vs Impact elevation vs Object diameter. Orbit characteristics: a = 7185 km, e = 0.001, i = 98 . Epoch: 1st May 2009. Mission lifetime: 1 year N N panels fluxes Y Y ij P = 1 1 P (9) pen pen j=1 i=1 where N is the total number of vector ux elements and N is fluxes panels the total number of panels composing the structure considered. The outlined 15 procedure is general and can be used for both the external structure of the spacecraft and for the internal components, provided the relevant changes are taken into account. Section 3 provides a detailed description of such cor- rections with the complete procedure for the assessment of the vulnerability of internal components. 3. Vulnerability of internal components For a complete vulnerability assessment, it is also necessary to consider the e ects on internal components. When a debris particle with sucient size and velocity impacts the outer structure of the spacecraft, a secondary debris cloud is usually generated (Putzar and Sch afer, 2006; Depczuk and Schonberg, 2003). The particles belonging to this secondary cloud can impact the internal components; therefore, it is necessary to evaluate the probability that the secondary ejecta have to damage internal components. This can be achieved through a fully statistical procedure based on the concept of vulnerable zones (Trisolini et al., 2018b; Putzar and Sch afer, 2006; Welty et al., 2013), where the vulnerability of an internal component (V ) can comp be expressed as the product of three di erent contributions as follows V = P  P  P (10) comp struct cloud BLE where P is the probability of space debris hitting the spacecraft ex- struct ternal structure inside the vulnerable zone relative to the target component, P is the probability that the secondary cloud ejecta will hit the compo- cloud nent, and P is the probability that the particles in this cloud perforate BLE the component wall. Fig. 8 shows the schematics of the entire procedure. 3.1. Impact propagation Understanding how debris impacts propagate inside the spacecraft plays a crucial role in modelling and predicting the damage received by internal components due to secondary debris ejecta. Experimental campaigns on very high-speed impacts have shown a considerable di erence between bal- listic and hypervelocity behaviour. In the former case, the projectile stays almost intact after the impact and keeps propagating in a direction that al- most coincides with the impact direction. In the latter case, instead, both the projectile and the impacted plate su er a fragmentation, which produces two secondary clouds of debris (Fig. 9) (Schonberg, 2001; Depczuk and Schon- berg, 2003). One cloud exits almost perpendicularly to the impacted wall 16 Start Vector flux element impact panel? Compute impact probability on V.Z. (P ) struct Ballistic Hypervelocity Type of impact Shatter hyp ball P P comp comp shatter comp Store component impact probability (P ) comp Compute critical diameter (d ) Compute critical flux (� ) Compute penetration Next vector probability (P ) BLE flux element Compute panel Store panel vulnerability vulnerability Next vulnerable zone Compute object vulnerability End Figure 8: Flow diagram for the vulnerability computation of an internal component. and is referred to as the normal debris cloud. The second cloud instead fol- lows more closely the direction of the projectile and is identi ed as the inline debris cloud. It is assumed that the debris belonging to these clouds remains contained inside conic surfaces so that their behaviour can be modelled using only the characteristics of the cone (i.e. the direction of the cone axis and the spread angle of the cone). Eqs. (11-14) describes the relation between the Next vector flux element Projectile Bumper plate In-line cloud Normal cloud Figure 9: Geometry of secondary cloud ejecta. characteristics of the impact (impact velocity, impact angle, particle diame- ter, and wall material) and geometry of the cones (Schonberg, 2001; Depczuk and Schonberg, 2003). 0:049 0:054 v t 1 p s 1:134 = 0:471   cos () (11) C d 0:086 0:478 v t 2 p s 0:586 = 0:532   cos () (12) C d 0:907 0:195 v t p s 0:394 tan  = 1:318   cos () (13) C d 1:096 0:345 v t p s 0:738 tan  = 1:556   cos () (14) C d where  is the impact angle,  is the de ection angle of the normal debris cloud,  is the de ection angle of the inline debris cloud,  is the half-cone 2 1 angle of the normal cloud,  is the half-cone angle of the inline cloud. t is 2 s the impacted plane thickness, C is the speed of sound of the plate material, d is the particle diameter, and v is the relative particle impact velocity. p p 3.2. Vulnerable zones The vulnerable zone (Putzar and Sch afer, 2006) is de ned as the area on the external structure of the spacecraft that, if impacted, can lead to an 18 impact also on the considered component (Fig. 10). Any impact of a particle onto this area generates fragments that may hit the component in question, with a probability that depends on the geometry of the impact (secondary debris ejecta characteristics, stand-o distance, and shielding from other com- ponents). The lateral extent of the vulnerable zone (l ) is expressed as vz l = 2 s tan +  d + d (15) vz max target p;max where s is the spacing between the structure wall and the component front face (stand-o distance), d is the lateral extent of the target compo- target nent (the component of which we are evaluating the vulnerability), d p;max is the maximum projectile diameter, and is the maximum ejection an- max gle, which has a value of 63.15 degrees (Putzar and Sch afer, 2006). The parameter d instead represents a user-de ned value for the extent of the p;max particle diameter, which takes into account the contribution of the particle to the impact probability. Suggested values for d are 10 mm for vulner- p;max able components and 20 mm for components with higher impact resistance (Putzar and Sch afer, 2006). Fig. 11 shows a visual example of the vulnerable zones for a box-shaped component inside a cubic structure. The cyan regions represent the projection of the vulnerable zones extensions onto the external structure. Projectile l / 2 vz Structure wall max Target plane target Figure 10: Representation of the lateral extension of the vulnerable zone. 3.3. Penetration probability of internal components As previously introduced in Eq. (10), the computation of the penetration probability on an internal component is subdivided into the computation 19 Figure 11: Vulnerable zones of a box projected onto the faces of a cubic structure (the vulnerable zone of the back face is omitted for clarity). ij of three di erent contributions. The rst contribution, P , is calculated struct following the procedure outlined in Section 2.4. However, in this case, the surface to be considered is the one delimited by the vulnerable zone (Fig. 11). The resulting expression is ij ij P = 1 exp '  S  T (16) struct vz ij where P is the impact probability on the j -th vulnerable zone by the i -th struct vector ux element. ' is the ux relative to the i -th vector ux element, ij S is the projected area of the j -th vulnerable zone relative to the target vz component along the direction of the i -th vector ux element ; and T is the mission time in years. When a particle hits the vulnerable zone, it does not necessarily mean that the downrange fragments will damage the target component. The second term in Eq. (10) takes this into account. To prop- erly compute this contribution, it is necessary to consider the three di erent impact regimes (ballistic, hypervelocity, and shutter ) separately; in fact, they lead to di erent geometries for the impact ejecta and require di erent ex- pressions for the assessment of the critical diameter (Section 2.3). In case of an hypervelocity impact, the probability to impact the target component is 20 computed as the ratio between the extent of the cloud ejecta at the target component plane and the extent of the vulnerable zone of the component (Putzar and Sch afer, 2006). The resulting expression is given by ij ejecta ij P = (17) cloud;h vz;j ij where P is the probability that the cloud relative to the i -th vector cloud;h ux element, which has already impacted the j -th vulnerable zone, will hit the target component, lvz; j is the extent of the j -th vulnerable zone in the ij target plane, and d is the extent of the debris ejecta at the target plane ejecta relative to the j -th vulnerable zone (Fig. 10), which is expressed as ij d = 2 tan   s + 1=2 d (18) ij j target;j ejecta where s is the stand-o distance between the component and the external wall to which the j -th vulnerable zone is associated, d is the extent of target;j the target component along the direction of the j -th vulnerable zone, and ( ) is the ejection angle associated with the i -th vector ux element im- ij pacting on the j -th vulnerable zone, and can be computed with the following expression () =  + (19) which is obtained simplifying Eqs. (11-14), assuming that the ejection and spread angles are only a function of the impact angle  and that all the other parameters can be absorbed by a constant factor (Putzar and Sch afer, 2006). In case of an impact in the ballistic regime, as no fragmentation occurs, only the size of the projectile needs to be taken into account. ij ij d + d target ij p P = (20) cloud;b vz;j ij where d is the particle diameter relative to the i -th vector ux element im- pacting on the j -th vulnerable zone. This value is the most probable particle diameter for the i -th vector ux element (Section 2.2.1) and is extracted from the debris ux distributions from MASTER-2009. For the scatter regime, a linear interpolation (Eq. (21)) between the hypervelocity regime (Eq. (17)) 21 and the ballistic regime (Eq. (20)) is adopted. ij v V LV ij ij p ij P = P +  P (21) cloud;s cloud;b cloud;h V V HV LV The value of P used in Eq. (10) is then selected as follows cloud ij P if v  V  cos p LV cloud;b ij ij P = (22) P if v  V  cos p HV cloud cloud;h : ij P otherwise cloud;s Finally, the last contribution in Eq. (10) is related to the computation of the penetration probability on the front face of the target component. Similarly, the contribution can be expressed as ij P = 1 exp '  S  T (23) c;i vz;ij BLE ij where P is the penetration probability for the j -th vector ux element BLE on the component associated with the i -th vulnerable zone, and ' is the c;i critical ux that is the ux associated to the value of the critical diameter computed with BLEs. As speci ed in Section 2.3, we adopted the SRL BLE to compute the critical diameter relative to the target component. It is always assumed that the last wall of the shielding con guration corresponds to the face of the target component, while the other walls are representative of the outer structure. The overall vulnerability of an internal component can then be expressed as N N panels fluxes Y Y ij ij ij V = 1 1 P  P  P (24) comp struct cloud BLE j=1 i=1 4. Mutual shielding methodology The vulnerable zone methodology as described in Section 3 lacks the ca- pability of considering the mutual shielding between components. In fact, Eqs. (29) and (31) do not take into account the contribution of possible shadowing between the components, where additional shielding is provided, preventing part if not all the secondary debris ejecta from impacting them. 22 This is especially important considering the directional nature of the space debris uxes. For example, impacts coming from the RAM direction are more dangerous than from other directions because of a higher relative ve- locity. As the vulnerable zone approach has its simplicity and the possibility to avoid time-consuming Monte Carlo simulations among its advantages, the idea is to extend this approach by integrating a methodology that allows the evaluation of the mutual shielding while still maintaining the advantages of the original approach. Among the three di erent contributions to the vulnerability (Eq. (10)), the second term (associated with the impact prob- ability of the ejecta on the component) is the one connected to the mutual shielding. The idea is to consider the interaction between the debris cone developed after the impact and the shielding elements interposed between the outer faces and the component in a purely geometrical way. To do so, we rst consider the nature of the impacts on the spacecraft structure. In the standard vulnerable zone approach, a particle can impact anywhere on the vulnerable zone (no impact location is sampled or speci ed). The lack of a precise impact location results in an issue connected to the assessment of the mutual shielding. In fact, the shadowing between components depends on the impact location and on the characteristic of the debris produced after the impact. In case of a hypervelocity impact, a cone of debris is generated; depending on the cone axis and aperture angle, and on the impact location, di erent areas of the target component may be visible and di erent portions of the debris cone can be shielded by the interposing components. To over- come this issue it was decided to discretize the vulnerable zone using a grid (Fig. 12). As there must not be a preferred impact location, for each vector ux element, an impact is simulated assuming the centre of each cell in the grid to represent the impact location. At this point, for each impact location, the resulting debris cone is generated and its interaction with the target and the shielding components is evaluated. The overall e ect of a vector ux element impacting the considered vulnerable zone is then obtained averaging the contribution of each cell in the grid. The procedure is then repeated for each vector ux element and each vulnerable zone. In the case of a ballis- tic impact, the procedure must change. In this case, the projectile passes through in the same direction of the impact vector, and no secondary debris is produced. 23 Cell impact point Target component Grid Vulnerable zone Figure 12: Representation of the vulnerable zone grid and of the cell impact locations. 4.1. Hypervelocity regime After a hypervelocity impact, the impacting particle is destroyed together with the area of the panel subject to the impact, and two secondary clouds of fragments are generated (Section 3.1). In the standard vulnerable zone formulation (Section 3.2), the extent of the vulnerable zone is de ned using a single conical shape (Eq. (15)). To maintain this approach, the two secondary clouds are merged into a single conical shape. The characteristics of the cone (i.e. its axis and aperture angle) are determined using Eqs. (11-14). The cone half-aperture angle () can be computed as follows: 1 1 =  (  ) +  ( +  ) (25) 2 1 1 2 2 4 The axis of the cone is the bisector of the aperture angle and belongs to the impact plane, which is the plane containing the impact vector ux element and the normal to the impacted face (Fig. 13). It is oriented of an angle equal to   with respect to the vector normal to the external plane. The impact plane de nes the interaction between the debris cone and the target plane, which is used to de ne the debris cone section at the impact plane (Fig. 14). In fact, it is used to determine the position and orientation of the debris cone section: the intersection of the axis of the cone and the target 24 plane is the centre of ellipse and the line belonging to the impact plane and perpendicular to the cone axis will be its semi-major axis. The cone obtained is then used to determine the impact probability on the target component. To do so, the interaction of the debris cone with the target object and the shielding components must be evaluated. Impact plane Cone axis External panel Figure 13: Representation of the cone aperture angle in the impact plane. The standard approach computes the impact probability by adding the extent of the target section to the extent of the debris cone in the target plane (Eq. (17)). This mechanism is extended here, and the mutual shielding is accounted for in both the debris cone and of the target section. For the debris cone, a perspective projection of the components onto the target plane is carried out and the intersection of these projections with the section of the cone at the target plane is performed (Fig. 14). The area used in the computation is then the residual area of the cone after the shielding component projections have been subtracted. This area is referred to as the available cone area A . For each vector ux element, c;av this area is evaluated over all the grid cells subdividing the vulnerable zone. k k k k k A = A : A ; :::; A ; :::; A ; k 2 f1; :::; N g; (26) cell c;av c s;1 s;n s;N where A is the intersection (:) between the k -th debris cone with the target plane, A is the perspective projection of the n -th shielding com- s;n ponent onto the target plane with respect to the k -th grid cell, N is the number of shielding components between the target component and the j -th vulnerable zone, and N is the number of cells subdividing the vulnerable cell 25 Target plane Debris cone section Shielding component projection Shielding component Target component External panel Figure 14: Mutual shielding in the hypervelocity regime. zone. The procedure is repeated for each grid cell and the result averaged in order to obtain the average available cone area relative to the i -th vector ij ux element impacting the j -th vulnerable zone (A ) c;av cell ij k A =  A (27) c;av c;av cell k=1 As the projection of the components and the cone section at the target plane can both exceed the limits of the spacecraft envelope, all the computed areas are cropped with respect to the limits of the target plane, which is limited by the external structure of the spacecraft. All the boolean operations such as intersection and di erence between the projected areas have been performed using the Python package Shapely (Gillies et al., 2007{). The second contribution in the vulnerable zone equation is the target component length (Eq. (15)). In this extension of the methodology, the visible target area is computed by performing a Boolean di erence between the target section and the perspective projection of the shielding components onto the target plane (Fig. 15). Again, the operation is repeated over each grid cell, and the average over the overall grid is carried out to evaluate the average target 26 ij visible area (A ) as follows: t;av cell ij k k k A =  A : A ; :::; A ; :::; A (28) t;av s;1 s;n s;N cell k=1 The impact probability associated to the i -th vector ux element impacting on the j -th vulnerable zone for the hypervelocity regime is then obtained with the following expression ij ij A + A t;av c;av ij P = (29) cloud;h vz;j where A is the area of the j-th vulnerable zone. vz;j Target plane Target component Shielding component projection Shielding component Visible target area External panel Figure 15: Perspective projection of a shielding component onto the target plane. 4.2. Ballistic regime In case of a ballistic impact, the projectile passes through the panel with- out being destroyed. Therefore, no ejecta are produced and the consequence of the impact cannot be schematized with a debris cone. Instead, for each vector ux element and impact point on the vulnerable zone grid, a line is generated with a vertex on the centre of the cell and direction equivalent to the one of the vector ux element (Fig. 16). This line represents the tra- 27 Impact plane Debris direction External panel Figure 16: Trajectory of the vector ux elements after a ballistic impact. jectory of the debris after the impact. The interaction of this line with the target component and the shielding components is then evaluated. Again, this is to maintain the analogy with the standard formulation. The variables that need to be considered are the particle size and the target area. It is assumed that the particle is not a ected during the impact, maintaining its shape (spherical) and dimension. The equivalence is obtained by using the cross-section of the particle (Eq. (30)). A = 1=4   d (30) p;i p;i where d is the sample particle diameter associated with the i -th vector ux p;i element. For the target section, the same procedure used in the hypervelocity case is adopted and the computation of the average target visible area is performed using Eq. (28). Finally, the impact probability in the ballistic case is given as ij A + A p;i ij t;av P = (31) cloud;b vz;j Eqs. (29) and (31) can be used inside Eq. (24) to compute the impact prob- ability of an internal component. 4.3. Equivalence with standard vulnerable zone formulation Sections 4.1 and 4.2 describe the methodology for assessing the contri- bution of shielding components through a procedure that relies on the use of areas, as this is the more natural way of considering such contribution. 28 However, the standard vulnerable zone formulation adopts an approach that uses the linear extension of the vulnerable zone and of the target component (Section 3.2). As the developed methodology aims at extending the standard formulation, Eqs. (29) and (31) have been converted into a linearised version as follows ij ij d + d t;av c;av ij P = (32) comp;h vz;j ij d + d p;i ij t;av P = (33) comp;b vz;j where d and d maintain the same meaning. The conversions from t;av c;av the areas to the equivalent length is carried out using the following simpli ed expression d = 2 A= (34) For completeness, both the methodology using the areas (Eqs. (29) and (31)) and the one using the equivalent linear extent (Eqs. (32) and (33)) have been tested and compared with state-of-the-art software packages. The results are presented in Section 5. 4.4. Correction factor methodology The previously described methodology captures the geometrical proper- ties of the impact phenomenon and its propagation inside the spacecraft structure; however, it still requires a computational time that is not applica- ble to a multi-objective optimisation framework (around 6 seconds for a con- guration with two internal components for a Python implementation). As the presented model is meant to be used inside a multi-objective optimisation framework, its execution speed is of paramount importance. Consequently, a simpli ed methodology has be developed, which still takes into account the mutual shielding, while reducing the large number of expensive geomet- rical operations needed for the complete procedure of Sections 4.1 and 4.2. The method described in this section consists in applying a correction fac- tor whenever one or more components cover the target component. This approach maintains the simplicity and the short computational time of the standard one while providing a consistent way to account for the reduced im- 29 pact probability caused by the mutual shielding. Given the di erent nature of the impacts in the hypervelocity and in the ballistic regime, two correction factors are used. In the case of the hypervelocity regime, it computes the portion of the vulnerable zone that can actually lead to an impact onto the target component. This is the part of the vulnerable zone that is not covered by the perspective projections of the shielding components. As these pro- jections depend on the impact point, even in this case the vulnerable zone is subdivided into a grid and each grid cell centre is an impact point for the vector ux element. The procedure is analogous to the one described in Section 4.1; however, in this case, the perspective projections of the shielding components onto the target plane (Fig. 14) are directly subtracted from the vulnerable zone area. This methodology avoids generating the cone of the secondary debris ejecta and performing its intersection with the target plane that is the most computationally expensive part of the procedure. Eq. (35) ij gives the expression for the available vulnerable zone area A relative to vz;av the j -th vulnerable zone and the can be computed again averaging over the contributions computed for each grid cell. cell h ij j k k k A =  A : A ; :::; A ; :::; A (35) vz;av vz s;1 s;n s;N cell k=1 The expression for the hypervelocity correction factor is then ij vz;av ij CF = (36) vz;j A value of 1 of the correction factor corresponds to no shielding, while a value of 0 indicates that the target component is not visible by the impactor. In the case of the ballistic regime, the same approach cannot be used, as there is no ejecta generation. Looking at Eq. (20), the impact probability in the ballistic regime depends only on the size of the particle and on the extent of the target object. As the dimension of the particle cannot change, only the extent of the target can be changed in order to correct the impact probability. In the case of a ballistic impact, a corrected target extent is used, which can be referred to as the visible target area. Using an approach similar to the hypervelocity case, the section of the shielding components is projected onto the target plane (Fig. 15). At this point, if these projections intersect the target component, they are subtracted from it using Boolean 30 operations. The procedure is repeated over each grid cell and averaged as follows: cell ij k k k A =  A : A ; :::; A ; :::; A (37) t;j t;av s;1 s;n s;N cell k=1 ij where A is the average visible target area associated to the j -th vul- t;av nerable zone and the i -th vector ux element, A is the target component t;j section relative to the j -th vulnerable zone, and A is the perspective pro- s;n jection of the n -th shielding component onto the target plane with respect to the k -th grid cell. The expression for the ballistic correction factor is then ij ij t;av CF = (38) t;j Again, as in Section 4.3, the correction factors are converted to their linearised counterparts as lengths are used in the standard procedure for the computation of the impact probability (Eq. (17)). The linearised version is again obtained using Eq. (34), giving the following expressions for the correction factors in the hypervelocity and ballistic regimes. ij vz;av ij CF = (39) vz;j ij ij t;av CF = (40) ij target Finally, the hypervelocity and the ballistic correction factors can be ap- plied to the computation of the impact probabilities as follows: ij ij ejecta ij P =  CF (41) comp;h h vz;j ij ij CF  d + d p;i target ij b P = (42) comp;b vz;j 31 5. Comparison with DRAMA and ESABASE2/DEBRIS To verify the validity of the approach described in Sections 3 and 4, a comparison with two state-of-the-art software packages has been performed. ESABASE2/DEBRIS and ESA DRAMA have been considered. DRAMA performs a simpli ed vulnerability analysis through its dedicated module MASTER-based Impact Flux and Damage Assessment Software (MIDAS) (Gelhaus et al., 2014, 2013; Martin et al., 2005b,a). The analysis of MIDAS is limited to the outer structure of the spacecraft (no internal components can be considered). The user can analyse the debris and meteoroid uxes and damage for any user-de ned target orbit and particle size range. The analy- sis can be performed for a spherical target, a random tumbling plate, or up to ten oriented surfaces can be de ned. The user can select the orientation, area, density, and type of shielding of the panels. There are four hard-coded damage equations and up to 20 can be de ned. The debris population uses the uxes distributions provided by ESA-MASTER. The collision ux analy- sis performed by MIDAS provides information about impacting particles and the probability of collision for each of the de ned surfaces separately. For this analysis MIDAS uses a reference area S . This area is the cross-section ref for spherical objects and randomly tumbling plate. For oriented surfaces, the surface areas of each panel are the reference areas. Given the simulation time T , and the impact ux ' generated by MASTER, MIDAS computes the number of impacts N as follows imp N = ' S  T (43) imp ref From which follows the impact probability imp P = 1 e (44) imp The value of T is computed as the di erence between the start and the end epoch de ned by the user. Alongside the impact analysis, MIDAS also performs a damage assessment for oriented surfaces. To do so, it uses the failure ux provided by MASTER, which is the ux of particles penetrating the surface. Similarly to the collision analysis, the number of penetrations N and the penetration probability P are computed as follows pen pen 32 N = '  S  T (45) pen fail ref pen P = 1 e (46) pen The failure ux ' is generated in a special plug-in routine of MASTER fail where the BLEs are applied. ESABASE2/DEBRIS is a more complex software, which allows the user to build an arbitrarily complex structure. The methodology used by ESABASE2 is based on a ray-tracing method, but no debris-cloud propagation is taken into account. However, the vulnerability of internal components can still be analysed by using a particular workaround: the outer structure is removed and the user has to provide manually the type of shielding and the stand- o distances for each panel of the internal structure. The presented model, instead, automatically detects the characteristics of the outer structure and assigns the proper stand-o distance and shielding con gurations. 5.1. Test case: Cubic structure in SSO orbit The rst comparison (Trisolini et al., 2018a) is aimed at verifying the main building blocks of the model, such as the representation of the environ- ment through vector ux elements, the implementation of the ballistic limit equations, and the computation of the impact and penetration probabilities through the approach outlined in Section 2.4 and the use of the concept of critical ux. A standard scenario has been selected, where the impact and penetration probabilities are computed for an aluminium Al-6061-T6 cubic- shaped object with 1 m side length and 2 mm thickness. The characteristics of the material are summarised in Table 2. Table 2: Material properties for aluminium Al-6061-T6. (kg=m ) HB C(m=s)  (MPa) m y 2713 95 5100 276 The mission considered is a 1-year mission in a SSO with altitude equal to 802 km, inclination of 98.6 degrees, and eccentricity of 0.001 with starting st epoch on the 1 of January 2016. The ballistic limit equation used is the Cour-Palais thin wall (Ryan and Christiansen, 2010). The discretisation used for the vector ux elements is summarised in Table 3. Either the bin 33 location or the number of bins has been speci ed, alongside the considered boundaries. Table 3: Binning used for the generation of the vector ux elements . Variable Bounds Binning d 0.0001 0.1 m 200 v 0 20 km/s 40 Every 15 in [ -90 , -30 ) [ ( 30 , 90 ] El -90 90 Every 10 in [ -30 , 30 ] Every 30 in [ -180 , -90 ] [ [ 90 , 180 ] Az -180 180 Every 15 in ( -90 , -45 ] [ [ 45 , 90 ) Every 5 in ( -45 , 45 ) The resulting comparison for the number of impacts and penetration is summarised in Table 4 and Table 5 respectively. Table 4: Comparison between the numbers of impacts for a cubic structure. Panel DRAMA ESABASE Model Lead 69.473 80.58 69.47 Space 0.48114 2.176 1.078 Trail 0.032326 0.0222 0.032 Earth 0.54294 2.517 1.259 Right 19.196 21.78 19.17 Left 21.953 27.96 21.975 Total 111.678406 135.0352 112.998 A good agreement is observed between the developed model and DRAMA, with results that di er from about 1 impact and 0.1 penetrations per year, which are comparable to other analyses (Miller et al., 2017). The results of the ESABASE2 simulations instead show a higher number of both pre- dicted impacts and penetrations with respect to the other two software pack- ages. This di erence may be explained with ESABASE2 using a di erent MASTER-2009 population. Nonetheless, the higher number of penetrations in ESABASE2 can then be directly connected to the higher number of im- pacts. The di erence with DRAMA is not substantial but still deserves an 34 Table 5: Comparison between the numbers of penetrations for a cubic structure. Panel DRAMA ESABASE Model Lead 0.2887 0.2868 0.276 Space 1.71E-05 5.65E-05 2.12E-05 Trail 7.60E-11 5.26E-08 1.03E-06 Earth 1.73E-05 5.05E-05 1.99E-05 Right 0.01027 0.0164 0.0067 Left 0.01 0.0203 0.0076 Total 0.3090 0.3235 0.2913 analysis. In fact, this is focused only on two of the faces (Earth and Space), while the others have almost identical results. This di erence can be traced back to the discretisation employed in the generation of the vector ux ele- ments (Table 3) combined with the strongly directional nature of the uxes in SSO orbits. The impacts on these two faces are mainly a function of the impact elevation angle. As this has a stronger directionality than the az- imuth, it is more susceptible to the discretisation used. Given the adopted discretisation in the impact elevation, it may not be able to predict the uxes on the Earth and Space faces as well as for the other ones. 5.2. Test case: Single child component in SSO orbit Once the general procedure has been veri ed with a standard test case against DRAMA and ESABASE2/DEBRIS, the model has been tested also for the computation of the vulnerability of internal components. For this test case, only a comparison with ESABASE2/DEBRIS could be performed as DRAMA does not support such an analysis. First, the model is tested for a con guration with a single internal component. A cubic-shaped parent object made of aluminium Al-6061-T6 with a 1 m side length and a 1 mm wall thickness has been selected. The child component contained in the parent is box-shaped, made of the same material, with a side length of 40 cm, and a wall thickness of 1 mm. The mission scenario is the same as the previous test case (Section 5.1). The ballistic limit equation used is the ESABASE Double Wall (G ade and Miller, 2013). Fig. 17 shows the results of the comparison between ESABASE2/DEBRIS and di erent options of the developed model. The di erent options presented are the standard vulnerable zone approach (Section 3) identi ed with the label Standard, the correction 35 factor methodology (Section 4.4), the impact ejecta methodology considering the areas (Sections 4.1 and 4.2) identi ed by the label Shielding, and the linearised version of the impact ejecta method (Section 4.3) identi ed by the label Linear shielding. Given the di erence between the impacts predicted by this model and ESABASE2/DEBRIS (Table 4), it was decided to compare the ratio between the number of penetrations and the number of impacts on the outer structure, instead of directly comparing the number of penetrations. 5.32e-05 5.32e-05 0.00005 0.00004 3.77e-05 0.00003 2.7e-05 0.00002 1.2e-05 0.00001 0.00000 Figure 17: Comparison for one internal component. Fig. 17 show that all the models predict a penetration ratio that is of the same order of magnitude. The model predicting the largest amount of penetrations is the standard vulnerable zone methodology. This is expected as this methodology is an intrinsically conservative procedure. Also expected is the behaviour of the correction factor methodology, which replicates the results of the standard formulation as no shielding components are present in the test case, resulting in a value of 1 for the correction factors of Eqs. (39) and (40). The two versions of the ejecta model instead exhibit quite a dif- ferent behaviour, with the linearised version showing the best agreement with ESABASE2/DEBRIS, while the version with the areas clearly gives the lowest result. This di erence can be expected: both models rely on the same computational procedure to evaluate the interactions between the de- bris ejecta and the shielding components; however, the linearisation process increases the computed value of P as the ratio between the areas will cloud ESABASE Standard Correction Factor Shielding Linear shielding # of pen. / # of imp. parent always be lower than the ratio between the associated equivalent lengths. 5.3. Test case: Two children components in SSO orbit For the nal comparison, a parent structure with two internal components is considered. The parent structure is a 2m  1m  1m aluminium Al- 6061-T6 box with a 1 mm wall thickness. The two internal components are two identical box-shaped objects with a 40 cm side length and a 1 mm wall thickness. The rst box (Component 1 ) is in the centre of the parent structure, while the second box (Component 2 ), is positioned in front of the rst one along the RAM direction at a distance of 0.2 m from the outer face (Fig. 18). Again, the mission scenario is equivalent to the one of Section 5.1. y External structure Comp. 2 40 cm Comp. 1 20 cm 20 cm Figure 18: Schematics of the satellite geometry used for the mutual shielding comparison with ESABASE2/DEBRIS. This comparison is used to test the capability of the code to deal with shielding components (Section 4). For the comparison, the ratio between the penetrations on the lead face of Component 1 and Component 2 is presented in Fig. 19. Fig. 19 shows that, as expected, the standard formulation is not able to correctly model the vulnerability of internal components when shielding needs to be considered. In fact, Component 1 has even more penetrations than Component 2. In this methodology, the two components are treated separately and Component 1 receives more penetration because the overall extent of the vulnerable zone is larger than the one of Component 2 as it is 37 1.19 1.2 1.0 0.783 0.8 0.6 0.57 0.528 0.4 0.208 0.2 0.0 Figure 19: Comparison for two internal components. positioned further away from the external face of the structure. The remain- ing three methodologies, instead, are all capable of comping with mutual shielding. The results of Fig. 19 shows that the correction factor model is the most conservative as it predicts a higher fraction of penetration between the two components. Similarly to Fig. 17, the ejecta model using the ar- eas (Shielding ) is the least conservative and predicts a considerably lower amounts of penetrations for Component 1. Finally, the linearised version of the ejecta model more closely matches the results of ESABAS2/DEBRIS. An interesting aspect of both the comparisons presented in Figs. 17 and 19 is the signi cant di erence between the procedure using area intersection for the assessment of the mutual shielding and the correspondent linearised version. As previously mentioned, this behaviour is expected. What is less expected is the better agreement of the linearised version with state-of-the-art soft- ware packages, rather than a more natural implementation based on the intersection of sections. This suggests that an extent-based computational methodology should be favoured over an area-based one for the implementa- tion of a vulnerable zone-based computation of the vulnerability. This is in agreement with the idea of this work that is to extend the original vulnera- ble zone methodology, whose probability assessment ultimately relies on an extent-based procedure. ESABASE Standard Correction Factor Shielding Linear shielding # of pen. Comp. 1 / # of pen. Comp. 2 5.4. Execution time This section contains a summary of the execution time of the presented methodology for the test cases presented in Sections 5.1 to 5.3 (Table 6). The tests have been run on a system equipped with an i7-8700 CPU @ 3.2 GHz with 6 cores and 16 GB of RAM. Our model uses pre-computed ux distributions from MASTER, and its execution time is not included in the runtime. Table 6: Execution time for the presented test cases (average over 10 simulations). Test case Runtime Test 1 0.53 s Test 2 - Correction factor 0.59 s Test 2 - Linear shielding 1.16 s Test 3 - Correction factor 0.69 s Test 3 - Linear shielding 5.56 s 6. Conclusions and Discussion The paper has presented the development of a novel methodology to pre- dict the vulnerability of spacecraft con gurations to the impact with debris particles, with particular attention to devising a methodology for taking into account mutual shielding e ects between internal components. The method has its foundations in the concept of the vulnerable zone and builds on it to obtain a statistically robust procedure for survivability assessments. A complete description of the main building blocks and methodologies imple- mented in the model has been given. A novel procedure for the computation of the mutual shielding contribution to the damage to internal components has been presented, based on the geometrical interaction between the sec- ondary debris ejecta, modelled as conic shapes and the shielding and target components. The overall methodology has been tested and compared with state of the art software packages, showing very good agreement with tradi- tional impact assessment methodologies. Such a comparison served also as a veri cation test for the fully probabilistic approach adopted, where the pen- etration probability is computed as the product of the probabilities of three The execution time for the Shielding and Linear shielding cases are equivalent 39 separate events. Also, the mutual shielding capabilities have been compared with ESABASE2/DEBRIS, showing comparable results and demonstrating that it is possible to predict the mutual shielding between components using the interaction of geometrical shapes and avoiding ray tracing methodologies. Alongside the obtained results, it should be mentioned that the presented methodologies is in its initial level of development and has been veri ed against limited test cases and scenarios. Consequently, for future devel- opment, a more complete set of tests should be executed, including more complex internal con gurations, with an increasing number of components and di erent shapes. This is especially useful for verifying the adaptability and scalability of the mutual shielding procedure to more complicated ge- ometries. Moreover, di erent shielding methodologies should also be tested, speci cally including honeycomb sandwich panels as they are among the most used spacecraft structures. Acknowledgements This work was funded by EPSRC DTP/CDT through the grant number EP/K503150/1. References Bjorkman, M.D., Christiansen, E.L., Lear, D.M., 2014. Bumper 3 Software User Manual. Technical Report NASA/TM-2014-218559. NASA. Bunte, K., Farahvashi, E., Miller, A., 2017. Methods to reduce uncertainties in spacecraft vulnearbility predictions, in: 7th European Conference on Space Debris, ESOC, Darmstadt, Germany. Bunte, K.D., Destefanis, R., Drolshagen, G., 2009. Spacecraft Shielding Lay- out and Optimisation Using ESABASE2/Debris, in: Proc. 5th European Conference on Space Debris. Christiansen, E., Arnold, J., Corsaro, B., Davis, A., Giovane, F., Hyde, J., Ratli , M., 2009. Handbook for Designing MMOD Protection. Technical Report NASA/TM-2009-214785. NASA Johnson Space Center. Huston, Texas, USA. 40 Depczuk, D., Schonberg, W.P., 2003. Characterizing debris clouds created in oblique orbital debris particle impact. Journal of Aerospace Engineering 16, 177{190. Flegel, S., Gelhaus, J., Wiedemann, C., Vorsmann, P., Oswald, M., Stabroth, S., Klinkrad, H., Krag, H., 2009. The MASTER-2009 Space Debris En- vironment Model, in: 5th European Conference on Space Debris, ESOC, Darmstadt, Germany. G ade, A., Miller, A., 2013. ESABASE2/Debris Release 6.0 Technical De- scription. Technical Report R077-231. etamax space GmbH. Gelhaus, J., Kebschull, C., Braun, V., Sanchez-Ortiz, N., Parilla Endrino, E., de Oliveira, J.C., Dominguez-Gonzalez, R., 2014. Upgrade of ESA's Space Debris Mitigation Analysis Tool Suite - Final Report. Technical Report ESA contract 4000104977/11/D/SR. European Space Agency. Gelhaus, J., M ockel, M., Wiedemann, C., Kempf, D., Krag, H., 2011. MASTER-2009 Software User Manual. Technical Report. M09/MAS- SUM. Gelhaus, J., Sanchez-Ortiz, N., Braun, V., Kebschull, C., de Oliveira, J.C., Dominguez-Gonzalez, R., Wiedemann, C., Krag, H., V orsmann, P., 2013. Upgrade of DRAMA ESA's space debris mitigation analysis tool suite, in: 6th European Conference on Space Debris, Darmstadt, Germany. Gillies, S., et al., 2007{. Shapely: manipulation and analysis of geometric objects. URL: https://github.com/Toblerity/Shapely. Grassi, L., Tiboldo, F., Destefanis, R., Donath, T., Winterboer, A., Evans, L., Janovsky, R., Kempf, S., Rudolph, M., Sch afer, F., Gelhaus, J., 2014. Satellite vulnerability to space debris { an improved 3D risk assessment methodology. Acta Astronautica 99, 283{291. doi:10.1016/j.actaastro. 2014.02.006. Gulde, M., Kempf, S., Sch afer, F., 2016. Fast and Flexible Space Debris Risk Assessment for Satellites. Journal of Space Safety Engineering 3, 111{113. doi:10.1016/S2468-8967(17)30003-4. 41 Inter-Agency Space Debris Coordination Committee, Committee, I.A.S.D.C., 2007. IADC space debris mitigation guidelines. Inter-Agency Space Debris Coordination Committee. Kempf, S., Sch afer, F., Rudolph, M., Welty, N., Donath, T., Destefanis, R., Grassi, L., Janovsky, R., Evans, L., Winterboer, A., 2013. Risk and vulnerability analysis of satellites due to MM/SD with PIRAT, in: 6th European Conference on Space Debris, Darmstadt, Germany. pp. 22{25. Kempf, S., Sch afer, F.K., Cardone, T., Ferreira, I., Geren e, S., Destefanis, R., Grassi, L., 2016. Simpli ed spacecraft vulnerability assessments at component level in early design phase at the European Space Agency's Concurrent Design Facility. Acta Astronautica 129, 291{298. doi:10. 1016/j.actaastro.2016.08.014. Kuiper, W., Drolshagen, G., Noomen, R., 2010. Micro-meteoroids and space debris impact risk assessment for the ConeXpress satellite using ESABASE2/Debris. Advances in Space Research 45, 683{689. Lewis, H.G., Radtke, J., Beck, J., Bastida Virgili, B., Krag, H., 2017a. Self- induced collision risk analysis for large constellations, in: 7th European Conference on Space Debris, ESOC, Darmstadt, Germany. Lewis, H.G., Radtke, J., Rossi, A., Beck, J., Oswald, M., Anderson, P., Bastida Virgili, B., Krag, H., 2017b. Sensitivity of the space debris en- vironment to large constellations and small satellites, in: 7th European Conference on Space Debris, ESOC, Darmstadt, Germany. Lewis, H.G., White, A.E., Crowther, R., Stokes, H., 2012. Synergy of debris mitigation and removal. Acta Astronautica 81, 62{68. doi:10.1016/j. actaastro.2012.06.012. Liou, J.C., Johnson, N.L., 2006. Planetary science - Risks in space from orbiting debris. Science 311, 340{341. doi:10.1126/science1121337. Martin, C., Cheese, J., Brandmueller, C., Bunte, K., Fritsche, B., Lips, T., Klinkrad, H., Sanchez, N., 2005a. A debris risk assessment tool support- ing mitigation guidelines, in: 4th European Conference on Space Debris, ESA/ESOC, Darmstadt, Germany. pp. 345{352. 42 Martin, C.E., Cheeses, J.E., Sanchez-Ortiz, N., Klinkrad, H., Bunte, K., Hauptmann, S., Fritsche, B., Lips, T., 2005b. Introducing the ESA DRAMA tool. Science and Technology Series 110, 219{233. Miller, A., Zaake, M., Gromann-Ruh, F., Bunte, K., Millinger, M., Drol- shagen, G., 2017. Recent extensions of the esabase2/debris impact risk assessment tool, in: 7th European Conference on Space Debris, ESOC, Darmstadt, Germany. O'Connor, B., 2008. Handbook for Limiting Orbital Debris. NASA Handbook 8719.14. National Aeronautics and Space Administration, Washington, DC Putzar, R., Sch afer, F., 2006. Vulnerability of spacecraft equipment to space debris and meteoroids impacts. Technical Report Report No. I-15/06. Ernst-Mach-Institute, Freiburg. Germany. Radtke, J., Stoll, E., Lewis, H.G., Bastida Virgili, B., 2017. The impact of the increase in small satellite launch trac on the long-term evolution of the space debris environment, in: 7th European Conference on Space Debris, ESOC, Darmstadt, Germany. Reimerdes, H.G., Wohlers, W., 2001. Optimization of micrometeoroid and space debris protection systems, in: Space Debris, pp. 655{660. Ryan, S., Christiansen, E.L., 2010. Micrometeoroid and orbital debris (MMOD) shield ballistic limit analysis program. Technical Report NASA/TM-2009-214789. Houston. Ryan, S., Christiansen, E.L., 2011. A ballistic limit analysis programme for shielding against micrometeoroids and orbital debris. Acta Astronautica 69, 245{257. Ryan, S., Sch afer, F., Destefanis, R., Lambert, M., 2008. A ballistic limit equation for hypervelocity impacts on composite honeycomb sandwich panel satellite structures. Advances in Space Research 41, 1152{1166. Sch afer, F., Lambert, M., Christiansen, E., Kibe, S., Stokes, H., Reimerdes, H.G., Zengyao, H., 2005. The inter-agency space debris coordination com- mittee (IADC) protection manual, in: 4th European Conference on Space Debris, ESA/ESOC, Darmstadt, Germany. p. 39. 43 Sch afer, F.K., Ryan, S., Lambert, M., Putzar, R., 2008. Ballistic limit equa- tion for equipment placed behind satellite structure walls. International Journal of Impact Engineering 35, 1784{1791. Schaub, H., Jasper, L.E.Z., Anderson, P.V., McKnight, D.S., 2015. Cost and risk assessment for spacecraft operation decisions caused by the space debris environment. Acta Astronautica 113, 66{79. Schonberg, W.P., 2001. Characterizing secondary debris impact ejecta. In- ternational Journal of Impact Engineering 26, 713{724. Space Debris Oce, 2017. Space Debris by the Numbers. URL: http://www.esa.int/Our_Activities/Operations/Space_Debris/ Space_debris_by_the_numbers. Stokes, H., Cougnet, C., Gelhaus, J., Oswald, M., Schaefer, U., Theroude, C., 2012. A Detailed Impact Risk Assessment of Two Low Earth Orbiting Satellites, in: 65th International Astronautical Congress, Naples, Italy. Stokes, H., Swinerd, G., Walker, R., Wilkinson, J., 2000. Achieving cost e ec- tive debris protection of unmanned spacecraft using shield, in: Bendisch, J. (Ed.), Space Debris 2000, American Astronautical Society. pp. 175{186. URL: https://eprints.soton.ac.uk/21579/. series ISSN 0278-4017. Stokes, P.H., Swinerd, G.G., 2005. Debris protection optimisation of a re- alistic unmanned spacecraft using SHIELD, in: 4th European Conference on Space Debris, p. 515. Trisolini, M., Lewis, H.G., Colombo, C., 2015. Survivability and Demise Criteria for Sustainable Spacecraft Design, in: 66th International Astro- nautical Conference, Jerusalem. Trisolini, M., Lewis, H.G., Colombo, C., 2016. Demise and survivability crite- ria for spacecraft design optimization. Journal of Space Safety Engineering 3, 83{93. doi:https://doi.org/10.1016/S2468-8967(16)30023-4. Trisolini, M., Lewis, H.G., Colombo, C., 2018a. Demisability and surviv- ability sensitivity to design-for-demise techniques. Acta Astronautica 145, 357{384. doi:10.1016/j.actaastro.2018.01.050. 44 Trisolini, M., Lewis, H.G., Colombo, C., 2018b. Spacecraft design optimisa- tion for demise and survivability. Aerospace Science and Technology 77, 20. doi:10.1016/j.ast.2018.04.006. Welty, N., Rudolph, M., Sch afer, F., Apeldoorn, J., Janovsky, R., 2013. Computational methodology to predict satellite system-level e ects from impacts of untrackable space debris. Acta Astronautica 88, 35{43. doi:10. 1016/j.actaastro.2013.01.021. Appendix A. Material database In Table A.7 are summarised the values of the coecients for the SRL BLE for the two di erent cases of Aluminium outer bumper plate and CFRP outer bumper plate (Christiansen et al., 2009). Table A.7: Summary of the coecients for the SRL BLE Symbol Aluminium outer bumper CFRP outer bumper V 3 km/s 4.2 km/s LV V 7 km/s 8.4 km/s HV K 1.4 1.1 3S K 0.4 0.4 3D K 1.5 1 tw K 0.1 1 S2 2/3 1/3 4/3 if 45    65 4/3 4/3 if 45    65 8/3 if 45    65 10/4 if 45    65 1/3 2/3 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Physics arXiv (Cornell University)

Predicting the vulnerability of spacecraft components: modelling debris impact effects through vulnerable-zones

Physics , Volume 2020 (2003) – Mar 10, 2020

Loading next page...
 
/lp/arxiv-cornell-university/predicting-the-vulnerability-of-spacecraft-components-modelling-debris-31uIhiU6ly
ISSN
0273-1177
eISSN
ARCH-3341
DOI
10.1016/j.asr.2020.03.010
Publisher site
See Article on Publisher Site

Abstract

The space environment around the Earth is populated by more than 130 million objects of 1 mm in size and larger, and future predictions shows that this amount is destined to increase, even if mitigation measures are imple- mented at a far better rate than today. These objects can hit and damage a spacecraft or its components. It is thus necessary to assess the risk level for a satellite during its mission lifetime. Few software packages perform this analysis, and most of them employ time-consuming ray-tracing methodology, where particles are randomly sampled from relevant distributions. In addi- tion, they tend not to consider the risk associated with the secondary debris clouds. The paper presents the development of a vulnerability assessment model, which relies on a fully statistical procedure: the debris uxes are di- rectly used combining them with the concept of vulnerable zone, avoiding the random sampling the debris uxes. A novel methodology is presented to predict damage on internal components. It models the interaction between the components and the secondary debris cloud through basic geometrical Corresponding author Email addresses: mirko.trisolini@polimi.it (Mirko Trisolini), H.G.Lewis@soton.ac.uk (Hugh G. Lewis), camilla.colombo@polimi.it (Camilla Colombo) Present address: Politecnico di Milano, Via La Masa 34, 20156, Milano, Italy Preprint submitted to Advances in Space Research March 13, 2020 arXiv:2003.05521v1 [astro-ph.IM] 10 Mar 2020 operations, considering mutual shielding and shadowing between internal components. The methodologies are tested against state-of-the-art software for relevant test cases, comparing results on external structures and internal components. Keywords: space debris, debris impact, debris cloud, vulnerable zone, spacecraft vulnerability, penetration probability 1. Introduction Since the launch of Sputnik 1 in 1957, the space around the Earth and beyond has been the theatre of remarkable achievements but has also su ered from continuous exploitation. Decommissioned satellites, spent upper stages, mission-related objects, and fragments generated by collisions and explosions of satellites and upper stages have polluted the space environment in the form of space debris. Space debris is thus considered as a major threat to space mission; in fact, debris of just 1 cm can cause the break-up of a satellite (Putzar and Sch afer, 2006; Stokes and Swinerd, 2005), and smaller particles can still have enough energy to produce failures on components critical to the mission operations and scienti c objectives. Recent studies have shown a constant increase in the population of space debris, and the amount of debris is expected to keep growing in the next years (Radtke et al., 2017; Lewis et al., 2017a,b, 2012). The major space-faring nations and international committees have thus proposed a series of debris mitigation measures to protect the space environment (O'Connor, 2008; Sch afer et al., 2005; Inter- Agency Space Debris Coordination Committee and Committee, 2007). The implementation of these mitigation measures can help to mitigate the impact of human activities on the space environment. Nonetheless, the danger posed by possible impacts of space debris on spacecraft structures and components has to be addressed and regarded as a mission design driver as a single impact can compromise an entire mission (Grassi et al., 2014; Stokes et al., 2012; Christiansen et al., 2009). This is particularly important for missions in speci c orbital regions, such as highly inclined orbits, which are more densely populated by space debris and im- pacts can be more frequent (Liou and Johnson, 2006). The space debris population accounts for about 29000 objects larger than 10 cm, 750000 ob- jects between 1 cm to 10 cm, and 166 million objects from 1 mm to 1 cm 2 (Space Debris Oce, 2017). In Low Earth Orbit (LEO) orbits, only particles larger than 10 cm can be tracked (Schaub et al., 2015). Consequently, only for these particles collision avoidance manoeuvres are e ective in preventing damage to satellites. For the other untraceable yet dangerous debris, it is necessary to adopt speci c measures to ensure the mission survivability. To do so we need to assess the amount of damage caused by space debris to spacecraft components and identify the most vulnerable ones. Few software packages are currently available to perform the vulnerability assessment of a mission. Examples are ESABASE2/DEBRIS (G ade and Miller, 2013), Na- tional Aeronautics and Space Administration (NASA) BUMPER (Bjorkman et al., 2014), European Space Agency (ESA) Debris Risk Assessment and Mitigation Analysis (DRAMA) (Gelhaus et al., 2014, 2013; Martin et al., 2005b,a), SHIELD (Stokes et al., 2000; Stokes and Swinerd, 2005; Stokes et al., 2012), and Particle Impact Risk and Vulnerability Analysis Tool (PI- RAT) (Gulde et al., 2016; Kempf et al., 2013, 2016). These codes have a common structure, where the satellite is modelled through a geometric rep- resentation (which can be more or less detailed). The geometry is usually schematised through a set of panels. Each panel is then assigned the proper- ties and characteristics required for the impact analysis, such as the material, the shielding con guration, the thickness, etc. (Grassi et al., 2014; G ade and Miller, 2013; Stokes et al., 2000; Welty et al., 2013). An environmental model (e.g. Meteoroid and Space Debris Terrestrial Environment Reference (MAS- TER)) is then used to generate the debris uxes a ecting the spacecraft. The nature and frequency of the impacts on the structure and components of the spacecraft is evaluated and the damage associated with them is computed using semi-empirical relationships (i.e. Ballistic Limit Equations). Finally, the results of the analysis can be obtained in the form of number of impacts, number of penetrations, impact probability, and penetration probability. These software packages perform the vulnerability analysis with di erent levels of detail and techniques. The two main aspects to be considered are the modelling of the satellite geometry and the computation of the damage on internal components. For instance, DRAMA is only used for preliminary analysis. It only allows the de nition of simple geometrical shapes through a limited number of panels, which are created specifying their area, nor- mal, and shielding properties. Also, the impact analysis only involves the outer structure of the satellite and does not consider internal components. BUMPER instead generates a high- delity Finite Element Model (FEM) of the spacecraft structure; however, it cannot take into account the im- 3 pacts on internal components. SHIELD and ESABASE2/DEBRIS both use a panelised model of the satellite structure, and a ray-tracing methodology to simulate the impact of debris particle on the spacecraft. They both can compute the e ects of debris impacts on internal components. The method- ologies used by the two models for this task are, however, very di erent. In SHIELD, after the rst impact on the external structure, the projectile is con- sidered still intact and following the same direction of the initial impact. In ESABASE2/DEBRIS the assessment for internal components is not directly supported and requires a workaround: the user has to create a panelized rep- resentation of the internal components, without the external structure. The contribution of the external structure is then taken into account assigning a multiple-wall Ballistic Limit Equation (BLE) to each panel constituting the internal components. A recent integration to this methodology has been proposed and is under development (Bunte et al., 2017), which considers the characteristics of the debris cloud that is formed inside the spacecraft after the rst impact. In this methodology, the characteristic distributions relative to the debris cloud velocity, mass, and spatial distribution are obtained from available experimental data. The distributions are then sampled to generate rays representing the cloud fragments. This type of methodology is, however, time consuming as it requires the generation of thousands of rays and the veri cation of the impact with the components for each one of them. Be- sides, as the rays are generated through sampling, several simulations must be performed to have statistically meaningful results. Finally, the tool PI- RAT by the Fraunhofer Institute, also relies on a complete 3D modelling of the satellite structure and internal components and on the use of environ- mental models such as MASTER and BUMPER. However, di erently from SHIELD and ESABASE2/DEBRIS, uses a deterministic approach instead of a probabilistic one (Kempf et al., 2016; Gulde et al., 2016; Welty et al., 2013). This approach is based on a three-wall geometrical assessment to de ne threat directions, which generate the BLE con guration for the components. In par- ticular, it uses the Schafer-Ryan-Lambert (SRL)-BLE, so that shielding and partial shielding can be taken into account, as well as primary and secondary impacts. It has also been applied in the Concurrent Design Facility at ESA for preliminary vulnerability analyses. The work presented in this paper is devoted to the introduction of a novel methodology to assess the penetration probability on internal spacecraft com- ponents in an intrinsically statistical fashion (Welty et al., 2013) by exploiting 4 and extending the concept of vulnerable zones (Putzar and Sch afer, 2006). The need for ray tracing methodology is removed and so is the need for several Monte Carlo runs to obtain robust results. Also, the implementa- tion of vulnerable zones is coupled with the formation of secondary ejecta clouds after the rst impact. The methodology relies then on geometrical operations to compute the interactions of the debris cloud with the internal components. The presented methodology thus evolves from SHIELD, where the particle remains intact, by considering the cloud contribution, and avoids computationally expensive ray-tracing procedure as in Bunte et al. (2017). The methodology has been developed as a part of a broader study (Trisolini et al., 2018a,b, 2016, 2015), where a fast assessment of the vulnerability of simpli ed spacecraft con gurations was required. We rst describe the over- all model developed, for then focusing on the speci cs of the computation of the vulnerability of internal components. The results are then compared with DRAMA and ESABASE2/DEBRIS for relevant test cases. 2. Model outline To properly describe the developed methodology, it is important to un- derstand the entire vulnerability model in all its parts. The model evaluates the vulnerability of simpli ed spacecraft con gurations, considering the ef- fects on both the external structure and on the internal components (Trisolini et al., 2018a,b). It takes as inputs the spacecraft con guration, the mission scenario, and the debris uxes from a space environment model, and returns the impact and penetration probability on the external structure and inter- nal components. The main novelties in the model are the schematization of the debris environment through vector ux elements (Section 2.2.1), the implementation of a statistical methodology to compute the impact and pen- etration probabilities through the use of vulnerable zones (Section 3.2), and the development of a fully geometrical methodology for the assessment of the mutual shielding between internal components (Section 4) without relying on ray tracing methodologies. 2.1. Spacecraft con guration The spacecraft con guration is de ned through a set of elementary shapes such as boxes, cylinders, spheres, and panels. For each element in the con- guration, its main properties must be provided, such as the shape, the size, 5 the mass, and the position inside the spacecraft. An example of the con- guration entries is provided in Table 1; the main body of the satellite is always the rst entry and is positioned at the centre of the reference frame. The x -axis points towards the direction of the orbital velocity (RAM direc- tion), the z -axis points away from the Earth, and the y -axis follows from the right-hand rule. Table 1: Example of spacecraft con guration structure used by the model. ID Name Shape m l r w h Pos. (kg) (m) (m) (m) (m) (m) 0 Spacecraft Box 2000 3.5 n/a 1.5 1.5 n/a 1 Tank Sphere 15 n/a 0.55 n/a n/a (-1,0,0) 2 BattBox Box 5 0.6 n/a 0.5 0.4 (0,0,1) ... ... ... ... ... ... ... ... ... The satellite structure can be de ned with a single material and a uni- form thickness; alternatively, di erent materials, thickness, and shielding types can be assigned to each face. Three types of shielding can be used in the current version of the model: single wall shields, Whipple shields, and honeycomb sandwich panels (Ryan and Christiansen, 2010). For the internal components, the position must be speci ed; the components can be free in- side the main structure, or they can be attached to the external panels. The orientation of a component can also be speci ed, but only along the main axes of the satellite (x, y, and z ). Once the geometry of the spacecraft has been nalised, the software extracts all the relevant information needed for the vulnerability assessment, such as the shield type, the stand-o distances, the vulnerable zones extents, etc. Using this information together with the vector ux elements (Section 2.2.1), and the ballistic limit equations (Section 4.2), the vulnerability of each internal component can be computed. 2.2. Space environment model To assess the impact of space debris on a satellite, it is necessary to know the characteristics of the environment the satellite has to face during its lifetime. The characteristics and number of debris particles impacting a spacecraft are dependent upon the mission scenario, mainly its altitude and inclination. The present model uses ESA MASTER 2009 (Flegel et al., 2009; Gelhaus et al., 2011) to obtain the description of the debris environment via 6 ux predictions on user-de ned target orbits (Trisolini et al., 2018a). The uxes provide the number of particles per unit area per year that impact the spacecraft, given the speci c mission scenario (i.e. orbital parameters). The distributions needed for the survivability computation are the ux vs particle diameter, ux vs impact elevation, ux vs impact elevation vs impact azimuth, and the ux vs impact velocity vs impact azimuth. This set of distributions is sucient to perform the spacecraft vulnerability analysis. Fig. 2 and 1 represent an example of the ux distributions given by MASTER, where the di erential ux as a function of the impact velocity and impact azimuth and the cumulative ux as a function of the particle diameter are presented. The uxes refer to a Sun-synchronous orbit (SSO) with semi- major axis of 7185 km, eccentricity of 0.001, and inclination of 98 for one year with starting epoch 1st May 2009. 1 0 1 0 1 0 1 0 -1 1 0 -2 1 0 -3 1 0 -4 1 0 -5 1 0 -6 1 0 -4 -3 -2 -1 1 0 1 0 1 0 1 0 Par t icle d iam et er (m ) Figure 1: Example of MASTER-2009 cumulative ux vs particle diameter distribution. Orbit characteristics: a = 7185 km, e = 0.001, i = 98 . Epoch: 1st May 2009. Mission lifetime: 1 year 2.2.1. Vector Flux Elements Despite MASTER-2009 provides the description of the debris environ- ment through uxes, this information cannot be directly used in the compu- tation of the vulnerability of a spacecraft. In fact, assessing the criticality of the impact of a debris fragment requires the knowledge of the particle Particle Flux (1/m yr) Figure 2: Example of MASTER-2009 di erential ux vs impact azimuth vs impact velocity distribution. diameter, velocity, and direction. These data can be obtained by directly sampling the ux distributions provided by MASTER to generate a set of impacting particles, which are shot towards the satellite. This procedure is at the core of ray tracing methods, which are computationally expensive and re- quire a large number of simulations in order to be statistically robust (Stokes et al., 2000; Stokes and Swinerd, 2005). The methodology implemented in this work, instead, follows a novel approach to bypass the use of such com- putationally expensive procedures, and at the same time have a suciently accurate description of the debris environment. The compromise is achieved by characterising the debris environment surrounding the spacecraft with vector ux elements. The role of vector ux elements is to summarise the information relative to a speci c angular sector of space around the satel- lite (expressed in terms of impact azimuth and elevation) as provided by MASTER-2009. In fact, each vector element is associated with a value of the particle ux, particle diameter, impact velocity, and impact direction (Fig. 3) and these values are then used for the entire sector. The complete procedure to generate the vector ux elements is described in detail in Trisolini et al. (2018a). In summary, it consists of subdividing the space around the satellite into a set of angular sectors of impact azimuth and 8 Figure 3: Vector ux elements representation. In this example each vector carries the value of the impact velocity for the angular sector it is referred to. elevation. Then, for each angular sector, the diameter, the impact velocity, and the impact direction are extracted from the MASTER-2009 distribu- tions. As in each angular sector these quantities have a distribution and not a single value, we need to use a criterion to select the value to be as- sociated to the vector ux element. The criterion we use is the weighted average of the quantity of interest, using as weights the corresponding values of the particle uxes. For a generic quantity x, inside the angular sector [Az ; Az ; El ; El ] we thus have i i+1 i i+1 x ' k k k=1 x  = i 2 f1; :::; N g; j 2 f1; :::; N g; (1) ij Az El k=1 where x  is the weighted average value of the quantity x referred to the ij ij angular sector, ' is the value of the ux corresponding to the quantity x , which is obtained from a histogram equivalent to the one of Fig. 1 but limited to the considered angular sector, and M is the number of subdivisions used to obtain the ux histogram from MASTER-2009. N , and N are Az El the number of angular sector subdivision for the azimuth and the elevation angle respectively. Using this procedure for all the quantities of interest we 9 obtain a vector in the form of Eq. (2), which represents all the properties de ning a vector ux element referred to an angular sector. '  v  Az El d f = (2) ij p;ij ij ij p;ij ij The use of the schematisation introduced by the vector ux elements aims at simplifying the computation and have a computationally ecient proce- dure. In fact, with this, we can limit the number of times we check for penetration using BLEs and, in particular, the number of geometrical oper- ations that are required to assess the contribution of the clouds associated to the secondary ejecta. A better precision could be achieved directly using the binned output generated by MASTER-2009. However, this would come at the cost of a higher computational time. 2.3. Ballistic Limit Equations When debris particles hit a spacecraft, the type of impact and its ef- fects need to be assessed. Such an assessment is performed using a set of experimentally derived analytical expressions referred to as Ballistic Limit Equations (BLEs) (Ryan and Christiansen, 2010, 2011; Christiansen et al., 2009; Sch afer et al., 2005). Given the impact speed, the impact angle, the projectile density, and the target characteristics, BLEs allow the computation of the critical particle diameter, which is the minimum diameter an impact- ing particle must have to produce damage on the speci ed target. In general, BLEs have di erent expressions as a function of the type of impact. Three di erent types of impacts are distinguished depending on the relative impact velocity: ballistic, hypervelocity, and shatter regime (Ryan and Christiansen, 2010; Ryan et al., 2008; Ryan and Christiansen, 2011). The expressions for the BLEs are further divided by the shielding type: single-wall, double-wall, triple-wall, and advanced shielding concepts all have dedicated expressions (Ryan and Christiansen, 2010). Single-wall BLEs are subdivided accord- ing to the material of the shield, from classical metallic materials such as aluminium, titanium, and stainless steel, to Carbon Fibre Reinforced Plas- tics (CFRP), breglass, glass, and polycarbonate. Dual-wall shields include two main con gurations: the standard metallic Whipple shield (Ryan and Christiansen, 2010), and the Honeycomb Sandwich Panel (HC-SP) (Ryan et al., 2008). A commonly used triple-wall BLE for the analysis of space- craft vulnerability is the SRL equation Eqs. (3-4) (Sch afer et al., 2008; Ryan et al., 2008; Putzar and Sch afer, 2006; Grassi et al., 2014; Welty et al., 2013). 10 The SRL BLE allows the computation of the critical diameter for equipment placed inside the main structure of a satellite by considering the last plate of the triple-wall con guration as the face of the target equipment (Putzar and Sch afer, 2006). As this BLE is versatile and can be used for internal components, we will use it for the reminder of the work. The expression of the SRL BLE for the ballistic regime (v  V  cos ) is p LV 0:5 " # 18=19 1 0:5 y t + t  + t b ob K 40 3S d = (3) c;b 2=3 0:5 0:6 cos     v where v is the particle velocity relative to the spacecraft,  is the particle p p density,  is the impact angle (angle between the relative velocity vector and the normal to the impacted surface), V is the ballistic regime transition LV velocity, K and  are tting factors whose values are summarised in Table 3S A.7, t is the bumper plate thickness, t is the outer bumper thickness, t is b ob w the rear wall thickness, and  is the yield strength of the rear wall material. In the hypervelocity regime (v  V  cos ) instead we have p HV 1=3 2=3 1=3 1:155 S  t + K  t + K  S  t cos b tw w S2 1 2 w d = c;h 2=3 1=3 1=9 2=3 K      v  cos p p 3D ob (4) where V is the hypersonic regime transition velocity. K , K , K , HV tw S2 3D , , and are tting factors (Table A.7);  is the outer bumper density; S ob 1 and S are the spacing between the outer bumper and the bumper plate, and the space between the bumper plate and the rear wall respectively. Finally, in the shatter regime (V  cos   v  V  cos ) a linear interpolation LV p HV is used between the critical diameters obtained in the ballistic and in the hypervelocity regimes. The SRL BLE has been used throughout this work for the computation of the critical diameter for impacts on components insides the spacecraft. Single-wall and dual-wall BLEs can also be used by the model to predict impact damage on external structures and components. 2.4. Impact and penetration probability assessment The procedure for the computation of the impact probability (P ) and imp penetration probability (P ) depends on the characteristics of the environ- pen ment surrounding the satellite that are the particle ux, velocity, direction, 11 and diameter. In addition, the orientation of the surface considered plays an important role as it in uences the area that is visible by a speci c vector ux element. Finally, even the mission lifetime is an important parameter to be considered as the more a spacecraft resides in a speci c debris environment, the more it will be exposed to the debris uxes. The outline of the procedure is presented in Fig. 4 and the detailed description can be found in Trisolini et al. (2018b). Start Vector flux element impacts panel? No Yes Compute critical diameter Compute critical flux Compute panel Store panel vulnerability vulnerability Compute overall vulnerability End Figure 4: Flow diagram for the vulnerability computation. To evaluate the impact and penetration probabilities, the structure of the spacecraft is schematised as a set of panels. For each panel, it is necessary to check which vector ux element can actually impact the panel. This is achieved using a simple visibility which checks that v  n < 0, where v is i j i the i -th vector ux element and n is the normal to the j -th panel in the set. If a vector ux element impacts the panel, its contribution to the overall impact probability is then evaluated. By using Poisson statistics that is by assuming that debris impact events are statistically independent, the impact Next panel Next vector flux element Next vector flux element probability of a vector ux element on an a panel can be expressed as ij P = 1 exp '  S  T (5) i ij imp where ' is the ux associated to the i -th vector ux element, S is the i ij projected area of the j -th face on the direction of the i -th vector ux element, and T is the mission time in years. For the computation of the penetration probability, the classic approach involving BLEs can be adopted (Kuiper et al., 2010; Reimerdes and Wohlers, 2001; Welty et al., 2013; Stokes et al., 2000; Grassi et al., 2014; Bunte et al., 2009; Stokes et al., 2012). Substituting into the relevant BLE the velocity and direction of the vector ux elements, the critical diameters are computed. The critical diameter is then used to compute the corresponding critical ux i.e. the particle ux corresponding to diameters greater than the critical diameter. Figure 5: Graphical representation of the methodology for the computation of the critical ux. Orbit characteristics: a = 7185 km, e = 0.001, i = 98 . Epoch: 1st May 2009. Mission lifetime: 1 year To compute the critical ux, we use the MASTER-2009 distribution of the cumulative ux vs particle diameter, from where the it can be extracted (Fig. 5). Finally, the expression for the penetration probability is given by Particle Flux (1/m yr) ij P = 1 exp '  S  T (6) c;i ij pen where ' is the critical ux associated to the i -th vector ux element. c;i It is important to highlight the simpli cation introduced with this pro- cedure: as the global distribution of cumulative ux vs diameter is used for computing the critical diameter, the obtained ux is the overall ux for the entire range of azimuth and elevation angles. This ux cannot be directly used to compute the penetration probability associated to a single vector ux element as to each one of them is associated a value of the particle ux that is dependent upon the directionality, i.e. impact elevation and impact azimuth. It is thus assumed that the shape of the curve describing the relation between the particle ux and the particle diameter maintains the same shape through the di erent values of impact azimuth and impact elevation. By doing so, we obtain the critical ux for a vector ux element as a fraction of the total critical ux, which is proportional to the ux associated to that particular vector ux element. With this assumption, the critical ux associated to a vector ux element can be computed as follows ' = '  (7) c;i i tot where ' is the total debris ux and ' is the overall critical ux as tot c extracted from Fig. 5. Figs. 6 and 7 show the distribution of the particle ux for the particle diameter and the impact azimuth and elevation respectively. From these maps can be better observed that the assumption is not always satis ed: there are regions where the inverse exponential trend of the overall particle ux with the diameter (Fig. 5) is maintained. In these regions, the scaling assumption can be reasonable; however, it is less accurate for other regions. The simpli ed nature of the proposed method is aligned with the introduced simpli cation; nonetheless, future development of the methodol- ogy will revise this assumption, including a more accurate determination of the critical ux. Finally, the overall impact and penetration probabilities (Eqs. (8) and (9) can be computed iterating over the entire set of vector ux elements and spacecraft panels as follows: N N panels fluxes Y Y ij P = 1 1 P (8) imp imp j=1 i=1 14 Figure 6: Particle ux vs Impact elevation vs Object diameter. Orbit characteristics: a = 7185 km, e = 0.001, i = 98 . Epoch: 1st May 2009. Mission lifetime: 1 year Figure 7: Particle ux vs Impact elevation vs Object diameter. Orbit characteristics: a = 7185 km, e = 0.001, i = 98 . Epoch: 1st May 2009. Mission lifetime: 1 year N N panels fluxes Y Y ij P = 1 1 P (9) pen pen j=1 i=1 where N is the total number of vector ux elements and N is fluxes panels the total number of panels composing the structure considered. The outlined 15 procedure is general and can be used for both the external structure of the spacecraft and for the internal components, provided the relevant changes are taken into account. Section 3 provides a detailed description of such cor- rections with the complete procedure for the assessment of the vulnerability of internal components. 3. Vulnerability of internal components For a complete vulnerability assessment, it is also necessary to consider the e ects on internal components. When a debris particle with sucient size and velocity impacts the outer structure of the spacecraft, a secondary debris cloud is usually generated (Putzar and Sch afer, 2006; Depczuk and Schonberg, 2003). The particles belonging to this secondary cloud can impact the internal components; therefore, it is necessary to evaluate the probability that the secondary ejecta have to damage internal components. This can be achieved through a fully statistical procedure based on the concept of vulnerable zones (Trisolini et al., 2018b; Putzar and Sch afer, 2006; Welty et al., 2013), where the vulnerability of an internal component (V ) can comp be expressed as the product of three di erent contributions as follows V = P  P  P (10) comp struct cloud BLE where P is the probability of space debris hitting the spacecraft ex- struct ternal structure inside the vulnerable zone relative to the target component, P is the probability that the secondary cloud ejecta will hit the compo- cloud nent, and P is the probability that the particles in this cloud perforate BLE the component wall. Fig. 8 shows the schematics of the entire procedure. 3.1. Impact propagation Understanding how debris impacts propagate inside the spacecraft plays a crucial role in modelling and predicting the damage received by internal components due to secondary debris ejecta. Experimental campaigns on very high-speed impacts have shown a considerable di erence between bal- listic and hypervelocity behaviour. In the former case, the projectile stays almost intact after the impact and keeps propagating in a direction that al- most coincides with the impact direction. In the latter case, instead, both the projectile and the impacted plate su er a fragmentation, which produces two secondary clouds of debris (Fig. 9) (Schonberg, 2001; Depczuk and Schon- berg, 2003). One cloud exits almost perpendicularly to the impacted wall 16 Start Vector flux element impact panel? Compute impact probability on V.Z. (P ) struct Ballistic Hypervelocity Type of impact Shatter hyp ball P P comp comp shatter comp Store component impact probability (P ) comp Compute critical diameter (d ) Compute critical flux (� ) Compute penetration Next vector probability (P ) BLE flux element Compute panel Store panel vulnerability vulnerability Next vulnerable zone Compute object vulnerability End Figure 8: Flow diagram for the vulnerability computation of an internal component. and is referred to as the normal debris cloud. The second cloud instead fol- lows more closely the direction of the projectile and is identi ed as the inline debris cloud. It is assumed that the debris belonging to these clouds remains contained inside conic surfaces so that their behaviour can be modelled using only the characteristics of the cone (i.e. the direction of the cone axis and the spread angle of the cone). Eqs. (11-14) describes the relation between the Next vector flux element Projectile Bumper plate In-line cloud Normal cloud Figure 9: Geometry of secondary cloud ejecta. characteristics of the impact (impact velocity, impact angle, particle diame- ter, and wall material) and geometry of the cones (Schonberg, 2001; Depczuk and Schonberg, 2003). 0:049 0:054 v t 1 p s 1:134 = 0:471   cos () (11) C d 0:086 0:478 v t 2 p s 0:586 = 0:532   cos () (12) C d 0:907 0:195 v t p s 0:394 tan  = 1:318   cos () (13) C d 1:096 0:345 v t p s 0:738 tan  = 1:556   cos () (14) C d where  is the impact angle,  is the de ection angle of the normal debris cloud,  is the de ection angle of the inline debris cloud,  is the half-cone 2 1 angle of the normal cloud,  is the half-cone angle of the inline cloud. t is 2 s the impacted plane thickness, C is the speed of sound of the plate material, d is the particle diameter, and v is the relative particle impact velocity. p p 3.2. Vulnerable zones The vulnerable zone (Putzar and Sch afer, 2006) is de ned as the area on the external structure of the spacecraft that, if impacted, can lead to an 18 impact also on the considered component (Fig. 10). Any impact of a particle onto this area generates fragments that may hit the component in question, with a probability that depends on the geometry of the impact (secondary debris ejecta characteristics, stand-o distance, and shielding from other com- ponents). The lateral extent of the vulnerable zone (l ) is expressed as vz l = 2 s tan +  d + d (15) vz max target p;max where s is the spacing between the structure wall and the component front face (stand-o distance), d is the lateral extent of the target compo- target nent (the component of which we are evaluating the vulnerability), d p;max is the maximum projectile diameter, and is the maximum ejection an- max gle, which has a value of 63.15 degrees (Putzar and Sch afer, 2006). The parameter d instead represents a user-de ned value for the extent of the p;max particle diameter, which takes into account the contribution of the particle to the impact probability. Suggested values for d are 10 mm for vulner- p;max able components and 20 mm for components with higher impact resistance (Putzar and Sch afer, 2006). Fig. 11 shows a visual example of the vulnerable zones for a box-shaped component inside a cubic structure. The cyan regions represent the projection of the vulnerable zones extensions onto the external structure. Projectile l / 2 vz Structure wall max Target plane target Figure 10: Representation of the lateral extension of the vulnerable zone. 3.3. Penetration probability of internal components As previously introduced in Eq. (10), the computation of the penetration probability on an internal component is subdivided into the computation 19 Figure 11: Vulnerable zones of a box projected onto the faces of a cubic structure (the vulnerable zone of the back face is omitted for clarity). ij of three di erent contributions. The rst contribution, P , is calculated struct following the procedure outlined in Section 2.4. However, in this case, the surface to be considered is the one delimited by the vulnerable zone (Fig. 11). The resulting expression is ij ij P = 1 exp '  S  T (16) struct vz ij where P is the impact probability on the j -th vulnerable zone by the i -th struct vector ux element. ' is the ux relative to the i -th vector ux element, ij S is the projected area of the j -th vulnerable zone relative to the target vz component along the direction of the i -th vector ux element ; and T is the mission time in years. When a particle hits the vulnerable zone, it does not necessarily mean that the downrange fragments will damage the target component. The second term in Eq. (10) takes this into account. To prop- erly compute this contribution, it is necessary to consider the three di erent impact regimes (ballistic, hypervelocity, and shutter ) separately; in fact, they lead to di erent geometries for the impact ejecta and require di erent ex- pressions for the assessment of the critical diameter (Section 2.3). In case of an hypervelocity impact, the probability to impact the target component is 20 computed as the ratio between the extent of the cloud ejecta at the target component plane and the extent of the vulnerable zone of the component (Putzar and Sch afer, 2006). The resulting expression is given by ij ejecta ij P = (17) cloud;h vz;j ij where P is the probability that the cloud relative to the i -th vector cloud;h ux element, which has already impacted the j -th vulnerable zone, will hit the target component, lvz; j is the extent of the j -th vulnerable zone in the ij target plane, and d is the extent of the debris ejecta at the target plane ejecta relative to the j -th vulnerable zone (Fig. 10), which is expressed as ij d = 2 tan   s + 1=2 d (18) ij j target;j ejecta where s is the stand-o distance between the component and the external wall to which the j -th vulnerable zone is associated, d is the extent of target;j the target component along the direction of the j -th vulnerable zone, and ( ) is the ejection angle associated with the i -th vector ux element im- ij pacting on the j -th vulnerable zone, and can be computed with the following expression () =  + (19) which is obtained simplifying Eqs. (11-14), assuming that the ejection and spread angles are only a function of the impact angle  and that all the other parameters can be absorbed by a constant factor (Putzar and Sch afer, 2006). In case of an impact in the ballistic regime, as no fragmentation occurs, only the size of the projectile needs to be taken into account. ij ij d + d target ij p P = (20) cloud;b vz;j ij where d is the particle diameter relative to the i -th vector ux element im- pacting on the j -th vulnerable zone. This value is the most probable particle diameter for the i -th vector ux element (Section 2.2.1) and is extracted from the debris ux distributions from MASTER-2009. For the scatter regime, a linear interpolation (Eq. (21)) between the hypervelocity regime (Eq. (17)) 21 and the ballistic regime (Eq. (20)) is adopted. ij v V LV ij ij p ij P = P +  P (21) cloud;s cloud;b cloud;h V V HV LV The value of P used in Eq. (10) is then selected as follows cloud ij P if v  V  cos p LV cloud;b ij ij P = (22) P if v  V  cos p HV cloud cloud;h : ij P otherwise cloud;s Finally, the last contribution in Eq. (10) is related to the computation of the penetration probability on the front face of the target component. Similarly, the contribution can be expressed as ij P = 1 exp '  S  T (23) c;i vz;ij BLE ij where P is the penetration probability for the j -th vector ux element BLE on the component associated with the i -th vulnerable zone, and ' is the c;i critical ux that is the ux associated to the value of the critical diameter computed with BLEs. As speci ed in Section 2.3, we adopted the SRL BLE to compute the critical diameter relative to the target component. It is always assumed that the last wall of the shielding con guration corresponds to the face of the target component, while the other walls are representative of the outer structure. The overall vulnerability of an internal component can then be expressed as N N panels fluxes Y Y ij ij ij V = 1 1 P  P  P (24) comp struct cloud BLE j=1 i=1 4. Mutual shielding methodology The vulnerable zone methodology as described in Section 3 lacks the ca- pability of considering the mutual shielding between components. In fact, Eqs. (29) and (31) do not take into account the contribution of possible shadowing between the components, where additional shielding is provided, preventing part if not all the secondary debris ejecta from impacting them. 22 This is especially important considering the directional nature of the space debris uxes. For example, impacts coming from the RAM direction are more dangerous than from other directions because of a higher relative ve- locity. As the vulnerable zone approach has its simplicity and the possibility to avoid time-consuming Monte Carlo simulations among its advantages, the idea is to extend this approach by integrating a methodology that allows the evaluation of the mutual shielding while still maintaining the advantages of the original approach. Among the three di erent contributions to the vulnerability (Eq. (10)), the second term (associated with the impact prob- ability of the ejecta on the component) is the one connected to the mutual shielding. The idea is to consider the interaction between the debris cone developed after the impact and the shielding elements interposed between the outer faces and the component in a purely geometrical way. To do so, we rst consider the nature of the impacts on the spacecraft structure. In the standard vulnerable zone approach, a particle can impact anywhere on the vulnerable zone (no impact location is sampled or speci ed). The lack of a precise impact location results in an issue connected to the assessment of the mutual shielding. In fact, the shadowing between components depends on the impact location and on the characteristic of the debris produced after the impact. In case of a hypervelocity impact, a cone of debris is generated; depending on the cone axis and aperture angle, and on the impact location, di erent areas of the target component may be visible and di erent portions of the debris cone can be shielded by the interposing components. To over- come this issue it was decided to discretize the vulnerable zone using a grid (Fig. 12). As there must not be a preferred impact location, for each vector ux element, an impact is simulated assuming the centre of each cell in the grid to represent the impact location. At this point, for each impact location, the resulting debris cone is generated and its interaction with the target and the shielding components is evaluated. The overall e ect of a vector ux element impacting the considered vulnerable zone is then obtained averaging the contribution of each cell in the grid. The procedure is then repeated for each vector ux element and each vulnerable zone. In the case of a ballis- tic impact, the procedure must change. In this case, the projectile passes through in the same direction of the impact vector, and no secondary debris is produced. 23 Cell impact point Target component Grid Vulnerable zone Figure 12: Representation of the vulnerable zone grid and of the cell impact locations. 4.1. Hypervelocity regime After a hypervelocity impact, the impacting particle is destroyed together with the area of the panel subject to the impact, and two secondary clouds of fragments are generated (Section 3.1). In the standard vulnerable zone formulation (Section 3.2), the extent of the vulnerable zone is de ned using a single conical shape (Eq. (15)). To maintain this approach, the two secondary clouds are merged into a single conical shape. The characteristics of the cone (i.e. its axis and aperture angle) are determined using Eqs. (11-14). The cone half-aperture angle () can be computed as follows: 1 1 =  (  ) +  ( +  ) (25) 2 1 1 2 2 4 The axis of the cone is the bisector of the aperture angle and belongs to the impact plane, which is the plane containing the impact vector ux element and the normal to the impacted face (Fig. 13). It is oriented of an angle equal to   with respect to the vector normal to the external plane. The impact plane de nes the interaction between the debris cone and the target plane, which is used to de ne the debris cone section at the impact plane (Fig. 14). In fact, it is used to determine the position and orientation of the debris cone section: the intersection of the axis of the cone and the target 24 plane is the centre of ellipse and the line belonging to the impact plane and perpendicular to the cone axis will be its semi-major axis. The cone obtained is then used to determine the impact probability on the target component. To do so, the interaction of the debris cone with the target object and the shielding components must be evaluated. Impact plane Cone axis External panel Figure 13: Representation of the cone aperture angle in the impact plane. The standard approach computes the impact probability by adding the extent of the target section to the extent of the debris cone in the target plane (Eq. (17)). This mechanism is extended here, and the mutual shielding is accounted for in both the debris cone and of the target section. For the debris cone, a perspective projection of the components onto the target plane is carried out and the intersection of these projections with the section of the cone at the target plane is performed (Fig. 14). The area used in the computation is then the residual area of the cone after the shielding component projections have been subtracted. This area is referred to as the available cone area A . For each vector ux element, c;av this area is evaluated over all the grid cells subdividing the vulnerable zone. k k k k k A = A : A ; :::; A ; :::; A ; k 2 f1; :::; N g; (26) cell c;av c s;1 s;n s;N where A is the intersection (:) between the k -th debris cone with the target plane, A is the perspective projection of the n -th shielding com- s;n ponent onto the target plane with respect to the k -th grid cell, N is the number of shielding components between the target component and the j -th vulnerable zone, and N is the number of cells subdividing the vulnerable cell 25 Target plane Debris cone section Shielding component projection Shielding component Target component External panel Figure 14: Mutual shielding in the hypervelocity regime. zone. The procedure is repeated for each grid cell and the result averaged in order to obtain the average available cone area relative to the i -th vector ij ux element impacting the j -th vulnerable zone (A ) c;av cell ij k A =  A (27) c;av c;av cell k=1 As the projection of the components and the cone section at the target plane can both exceed the limits of the spacecraft envelope, all the computed areas are cropped with respect to the limits of the target plane, which is limited by the external structure of the spacecraft. All the boolean operations such as intersection and di erence between the projected areas have been performed using the Python package Shapely (Gillies et al., 2007{). The second contribution in the vulnerable zone equation is the target component length (Eq. (15)). In this extension of the methodology, the visible target area is computed by performing a Boolean di erence between the target section and the perspective projection of the shielding components onto the target plane (Fig. 15). Again, the operation is repeated over each grid cell, and the average over the overall grid is carried out to evaluate the average target 26 ij visible area (A ) as follows: t;av cell ij k k k A =  A : A ; :::; A ; :::; A (28) t;av s;1 s;n s;N cell k=1 The impact probability associated to the i -th vector ux element impacting on the j -th vulnerable zone for the hypervelocity regime is then obtained with the following expression ij ij A + A t;av c;av ij P = (29) cloud;h vz;j where A is the area of the j-th vulnerable zone. vz;j Target plane Target component Shielding component projection Shielding component Visible target area External panel Figure 15: Perspective projection of a shielding component onto the target plane. 4.2. Ballistic regime In case of a ballistic impact, the projectile passes through the panel with- out being destroyed. Therefore, no ejecta are produced and the consequence of the impact cannot be schematized with a debris cone. Instead, for each vector ux element and impact point on the vulnerable zone grid, a line is generated with a vertex on the centre of the cell and direction equivalent to the one of the vector ux element (Fig. 16). This line represents the tra- 27 Impact plane Debris direction External panel Figure 16: Trajectory of the vector ux elements after a ballistic impact. jectory of the debris after the impact. The interaction of this line with the target component and the shielding components is then evaluated. Again, this is to maintain the analogy with the standard formulation. The variables that need to be considered are the particle size and the target area. It is assumed that the particle is not a ected during the impact, maintaining its shape (spherical) and dimension. The equivalence is obtained by using the cross-section of the particle (Eq. (30)). A = 1=4   d (30) p;i p;i where d is the sample particle diameter associated with the i -th vector ux p;i element. For the target section, the same procedure used in the hypervelocity case is adopted and the computation of the average target visible area is performed using Eq. (28). Finally, the impact probability in the ballistic case is given as ij A + A p;i ij t;av P = (31) cloud;b vz;j Eqs. (29) and (31) can be used inside Eq. (24) to compute the impact prob- ability of an internal component. 4.3. Equivalence with standard vulnerable zone formulation Sections 4.1 and 4.2 describe the methodology for assessing the contri- bution of shielding components through a procedure that relies on the use of areas, as this is the more natural way of considering such contribution. 28 However, the standard vulnerable zone formulation adopts an approach that uses the linear extension of the vulnerable zone and of the target component (Section 3.2). As the developed methodology aims at extending the standard formulation, Eqs. (29) and (31) have been converted into a linearised version as follows ij ij d + d t;av c;av ij P = (32) comp;h vz;j ij d + d p;i ij t;av P = (33) comp;b vz;j where d and d maintain the same meaning. The conversions from t;av c;av the areas to the equivalent length is carried out using the following simpli ed expression d = 2 A= (34) For completeness, both the methodology using the areas (Eqs. (29) and (31)) and the one using the equivalent linear extent (Eqs. (32) and (33)) have been tested and compared with state-of-the-art software packages. The results are presented in Section 5. 4.4. Correction factor methodology The previously described methodology captures the geometrical proper- ties of the impact phenomenon and its propagation inside the spacecraft structure; however, it still requires a computational time that is not applica- ble to a multi-objective optimisation framework (around 6 seconds for a con- guration with two internal components for a Python implementation). As the presented model is meant to be used inside a multi-objective optimisation framework, its execution speed is of paramount importance. Consequently, a simpli ed methodology has be developed, which still takes into account the mutual shielding, while reducing the large number of expensive geomet- rical operations needed for the complete procedure of Sections 4.1 and 4.2. The method described in this section consists in applying a correction fac- tor whenever one or more components cover the target component. This approach maintains the simplicity and the short computational time of the standard one while providing a consistent way to account for the reduced im- 29 pact probability caused by the mutual shielding. Given the di erent nature of the impacts in the hypervelocity and in the ballistic regime, two correction factors are used. In the case of the hypervelocity regime, it computes the portion of the vulnerable zone that can actually lead to an impact onto the target component. This is the part of the vulnerable zone that is not covered by the perspective projections of the shielding components. As these pro- jections depend on the impact point, even in this case the vulnerable zone is subdivided into a grid and each grid cell centre is an impact point for the vector ux element. The procedure is analogous to the one described in Section 4.1; however, in this case, the perspective projections of the shielding components onto the target plane (Fig. 14) are directly subtracted from the vulnerable zone area. This methodology avoids generating the cone of the secondary debris ejecta and performing its intersection with the target plane that is the most computationally expensive part of the procedure. Eq. (35) ij gives the expression for the available vulnerable zone area A relative to vz;av the j -th vulnerable zone and the can be computed again averaging over the contributions computed for each grid cell. cell h ij j k k k A =  A : A ; :::; A ; :::; A (35) vz;av vz s;1 s;n s;N cell k=1 The expression for the hypervelocity correction factor is then ij vz;av ij CF = (36) vz;j A value of 1 of the correction factor corresponds to no shielding, while a value of 0 indicates that the target component is not visible by the impactor. In the case of the ballistic regime, the same approach cannot be used, as there is no ejecta generation. Looking at Eq. (20), the impact probability in the ballistic regime depends only on the size of the particle and on the extent of the target object. As the dimension of the particle cannot change, only the extent of the target can be changed in order to correct the impact probability. In the case of a ballistic impact, a corrected target extent is used, which can be referred to as the visible target area. Using an approach similar to the hypervelocity case, the section of the shielding components is projected onto the target plane (Fig. 15). At this point, if these projections intersect the target component, they are subtracted from it using Boolean 30 operations. The procedure is repeated over each grid cell and averaged as follows: cell ij k k k A =  A : A ; :::; A ; :::; A (37) t;j t;av s;1 s;n s;N cell k=1 ij where A is the average visible target area associated to the j -th vul- t;av nerable zone and the i -th vector ux element, A is the target component t;j section relative to the j -th vulnerable zone, and A is the perspective pro- s;n jection of the n -th shielding component onto the target plane with respect to the k -th grid cell. The expression for the ballistic correction factor is then ij ij t;av CF = (38) t;j Again, as in Section 4.3, the correction factors are converted to their linearised counterparts as lengths are used in the standard procedure for the computation of the impact probability (Eq. (17)). The linearised version is again obtained using Eq. (34), giving the following expressions for the correction factors in the hypervelocity and ballistic regimes. ij vz;av ij CF = (39) vz;j ij ij t;av CF = (40) ij target Finally, the hypervelocity and the ballistic correction factors can be ap- plied to the computation of the impact probabilities as follows: ij ij ejecta ij P =  CF (41) comp;h h vz;j ij ij CF  d + d p;i target ij b P = (42) comp;b vz;j 31 5. Comparison with DRAMA and ESABASE2/DEBRIS To verify the validity of the approach described in Sections 3 and 4, a comparison with two state-of-the-art software packages has been performed. ESABASE2/DEBRIS and ESA DRAMA have been considered. DRAMA performs a simpli ed vulnerability analysis through its dedicated module MASTER-based Impact Flux and Damage Assessment Software (MIDAS) (Gelhaus et al., 2014, 2013; Martin et al., 2005b,a). The analysis of MIDAS is limited to the outer structure of the spacecraft (no internal components can be considered). The user can analyse the debris and meteoroid uxes and damage for any user-de ned target orbit and particle size range. The analy- sis can be performed for a spherical target, a random tumbling plate, or up to ten oriented surfaces can be de ned. The user can select the orientation, area, density, and type of shielding of the panels. There are four hard-coded damage equations and up to 20 can be de ned. The debris population uses the uxes distributions provided by ESA-MASTER. The collision ux analy- sis performed by MIDAS provides information about impacting particles and the probability of collision for each of the de ned surfaces separately. For this analysis MIDAS uses a reference area S . This area is the cross-section ref for spherical objects and randomly tumbling plate. For oriented surfaces, the surface areas of each panel are the reference areas. Given the simulation time T , and the impact ux ' generated by MASTER, MIDAS computes the number of impacts N as follows imp N = ' S  T (43) imp ref From which follows the impact probability imp P = 1 e (44) imp The value of T is computed as the di erence between the start and the end epoch de ned by the user. Alongside the impact analysis, MIDAS also performs a damage assessment for oriented surfaces. To do so, it uses the failure ux provided by MASTER, which is the ux of particles penetrating the surface. Similarly to the collision analysis, the number of penetrations N and the penetration probability P are computed as follows pen pen 32 N = '  S  T (45) pen fail ref pen P = 1 e (46) pen The failure ux ' is generated in a special plug-in routine of MASTER fail where the BLEs are applied. ESABASE2/DEBRIS is a more complex software, which allows the user to build an arbitrarily complex structure. The methodology used by ESABASE2 is based on a ray-tracing method, but no debris-cloud propagation is taken into account. However, the vulnerability of internal components can still be analysed by using a particular workaround: the outer structure is removed and the user has to provide manually the type of shielding and the stand- o distances for each panel of the internal structure. The presented model, instead, automatically detects the characteristics of the outer structure and assigns the proper stand-o distance and shielding con gurations. 5.1. Test case: Cubic structure in SSO orbit The rst comparison (Trisolini et al., 2018a) is aimed at verifying the main building blocks of the model, such as the representation of the environ- ment through vector ux elements, the implementation of the ballistic limit equations, and the computation of the impact and penetration probabilities through the approach outlined in Section 2.4 and the use of the concept of critical ux. A standard scenario has been selected, where the impact and penetration probabilities are computed for an aluminium Al-6061-T6 cubic- shaped object with 1 m side length and 2 mm thickness. The characteristics of the material are summarised in Table 2. Table 2: Material properties for aluminium Al-6061-T6. (kg=m ) HB C(m=s)  (MPa) m y 2713 95 5100 276 The mission considered is a 1-year mission in a SSO with altitude equal to 802 km, inclination of 98.6 degrees, and eccentricity of 0.001 with starting st epoch on the 1 of January 2016. The ballistic limit equation used is the Cour-Palais thin wall (Ryan and Christiansen, 2010). The discretisation used for the vector ux elements is summarised in Table 3. Either the bin 33 location or the number of bins has been speci ed, alongside the considered boundaries. Table 3: Binning used for the generation of the vector ux elements . Variable Bounds Binning d 0.0001 0.1 m 200 v 0 20 km/s 40 Every 15 in [ -90 , -30 ) [ ( 30 , 90 ] El -90 90 Every 10 in [ -30 , 30 ] Every 30 in [ -180 , -90 ] [ [ 90 , 180 ] Az -180 180 Every 15 in ( -90 , -45 ] [ [ 45 , 90 ) Every 5 in ( -45 , 45 ) The resulting comparison for the number of impacts and penetration is summarised in Table 4 and Table 5 respectively. Table 4: Comparison between the numbers of impacts for a cubic structure. Panel DRAMA ESABASE Model Lead 69.473 80.58 69.47 Space 0.48114 2.176 1.078 Trail 0.032326 0.0222 0.032 Earth 0.54294 2.517 1.259 Right 19.196 21.78 19.17 Left 21.953 27.96 21.975 Total 111.678406 135.0352 112.998 A good agreement is observed between the developed model and DRAMA, with results that di er from about 1 impact and 0.1 penetrations per year, which are comparable to other analyses (Miller et al., 2017). The results of the ESABASE2 simulations instead show a higher number of both pre- dicted impacts and penetrations with respect to the other two software pack- ages. This di erence may be explained with ESABASE2 using a di erent MASTER-2009 population. Nonetheless, the higher number of penetrations in ESABASE2 can then be directly connected to the higher number of im- pacts. The di erence with DRAMA is not substantial but still deserves an 34 Table 5: Comparison between the numbers of penetrations for a cubic structure. Panel DRAMA ESABASE Model Lead 0.2887 0.2868 0.276 Space 1.71E-05 5.65E-05 2.12E-05 Trail 7.60E-11 5.26E-08 1.03E-06 Earth 1.73E-05 5.05E-05 1.99E-05 Right 0.01027 0.0164 0.0067 Left 0.01 0.0203 0.0076 Total 0.3090 0.3235 0.2913 analysis. In fact, this is focused only on two of the faces (Earth and Space), while the others have almost identical results. This di erence can be traced back to the discretisation employed in the generation of the vector ux ele- ments (Table 3) combined with the strongly directional nature of the uxes in SSO orbits. The impacts on these two faces are mainly a function of the impact elevation angle. As this has a stronger directionality than the az- imuth, it is more susceptible to the discretisation used. Given the adopted discretisation in the impact elevation, it may not be able to predict the uxes on the Earth and Space faces as well as for the other ones. 5.2. Test case: Single child component in SSO orbit Once the general procedure has been veri ed with a standard test case against DRAMA and ESABASE2/DEBRIS, the model has been tested also for the computation of the vulnerability of internal components. For this test case, only a comparison with ESABASE2/DEBRIS could be performed as DRAMA does not support such an analysis. First, the model is tested for a con guration with a single internal component. A cubic-shaped parent object made of aluminium Al-6061-T6 with a 1 m side length and a 1 mm wall thickness has been selected. The child component contained in the parent is box-shaped, made of the same material, with a side length of 40 cm, and a wall thickness of 1 mm. The mission scenario is the same as the previous test case (Section 5.1). The ballistic limit equation used is the ESABASE Double Wall (G ade and Miller, 2013). Fig. 17 shows the results of the comparison between ESABASE2/DEBRIS and di erent options of the developed model. The di erent options presented are the standard vulnerable zone approach (Section 3) identi ed with the label Standard, the correction 35 factor methodology (Section 4.4), the impact ejecta methodology considering the areas (Sections 4.1 and 4.2) identi ed by the label Shielding, and the linearised version of the impact ejecta method (Section 4.3) identi ed by the label Linear shielding. Given the di erence between the impacts predicted by this model and ESABASE2/DEBRIS (Table 4), it was decided to compare the ratio between the number of penetrations and the number of impacts on the outer structure, instead of directly comparing the number of penetrations. 5.32e-05 5.32e-05 0.00005 0.00004 3.77e-05 0.00003 2.7e-05 0.00002 1.2e-05 0.00001 0.00000 Figure 17: Comparison for one internal component. Fig. 17 show that all the models predict a penetration ratio that is of the same order of magnitude. The model predicting the largest amount of penetrations is the standard vulnerable zone methodology. This is expected as this methodology is an intrinsically conservative procedure. Also expected is the behaviour of the correction factor methodology, which replicates the results of the standard formulation as no shielding components are present in the test case, resulting in a value of 1 for the correction factors of Eqs. (39) and (40). The two versions of the ejecta model instead exhibit quite a dif- ferent behaviour, with the linearised version showing the best agreement with ESABASE2/DEBRIS, while the version with the areas clearly gives the lowest result. This di erence can be expected: both models rely on the same computational procedure to evaluate the interactions between the de- bris ejecta and the shielding components; however, the linearisation process increases the computed value of P as the ratio between the areas will cloud ESABASE Standard Correction Factor Shielding Linear shielding # of pen. / # of imp. parent always be lower than the ratio between the associated equivalent lengths. 5.3. Test case: Two children components in SSO orbit For the nal comparison, a parent structure with two internal components is considered. The parent structure is a 2m  1m  1m aluminium Al- 6061-T6 box with a 1 mm wall thickness. The two internal components are two identical box-shaped objects with a 40 cm side length and a 1 mm wall thickness. The rst box (Component 1 ) is in the centre of the parent structure, while the second box (Component 2 ), is positioned in front of the rst one along the RAM direction at a distance of 0.2 m from the outer face (Fig. 18). Again, the mission scenario is equivalent to the one of Section 5.1. y External structure Comp. 2 40 cm Comp. 1 20 cm 20 cm Figure 18: Schematics of the satellite geometry used for the mutual shielding comparison with ESABASE2/DEBRIS. This comparison is used to test the capability of the code to deal with shielding components (Section 4). For the comparison, the ratio between the penetrations on the lead face of Component 1 and Component 2 is presented in Fig. 19. Fig. 19 shows that, as expected, the standard formulation is not able to correctly model the vulnerability of internal components when shielding needs to be considered. In fact, Component 1 has even more penetrations than Component 2. In this methodology, the two components are treated separately and Component 1 receives more penetration because the overall extent of the vulnerable zone is larger than the one of Component 2 as it is 37 1.19 1.2 1.0 0.783 0.8 0.6 0.57 0.528 0.4 0.208 0.2 0.0 Figure 19: Comparison for two internal components. positioned further away from the external face of the structure. The remain- ing three methodologies, instead, are all capable of comping with mutual shielding. The results of Fig. 19 shows that the correction factor model is the most conservative as it predicts a higher fraction of penetration between the two components. Similarly to Fig. 17, the ejecta model using the ar- eas (Shielding ) is the least conservative and predicts a considerably lower amounts of penetrations for Component 1. Finally, the linearised version of the ejecta model more closely matches the results of ESABAS2/DEBRIS. An interesting aspect of both the comparisons presented in Figs. 17 and 19 is the signi cant di erence between the procedure using area intersection for the assessment of the mutual shielding and the correspondent linearised version. As previously mentioned, this behaviour is expected. What is less expected is the better agreement of the linearised version with state-of-the-art soft- ware packages, rather than a more natural implementation based on the intersection of sections. This suggests that an extent-based computational methodology should be favoured over an area-based one for the implementa- tion of a vulnerable zone-based computation of the vulnerability. This is in agreement with the idea of this work that is to extend the original vulnera- ble zone methodology, whose probability assessment ultimately relies on an extent-based procedure. ESABASE Standard Correction Factor Shielding Linear shielding # of pen. Comp. 1 / # of pen. Comp. 2 5.4. Execution time This section contains a summary of the execution time of the presented methodology for the test cases presented in Sections 5.1 to 5.3 (Table 6). The tests have been run on a system equipped with an i7-8700 CPU @ 3.2 GHz with 6 cores and 16 GB of RAM. Our model uses pre-computed ux distributions from MASTER, and its execution time is not included in the runtime. Table 6: Execution time for the presented test cases (average over 10 simulations). Test case Runtime Test 1 0.53 s Test 2 - Correction factor 0.59 s Test 2 - Linear shielding 1.16 s Test 3 - Correction factor 0.69 s Test 3 - Linear shielding 5.56 s 6. Conclusions and Discussion The paper has presented the development of a novel methodology to pre- dict the vulnerability of spacecraft con gurations to the impact with debris particles, with particular attention to devising a methodology for taking into account mutual shielding e ects between internal components. The method has its foundations in the concept of the vulnerable zone and builds on it to obtain a statistically robust procedure for survivability assessments. A complete description of the main building blocks and methodologies imple- mented in the model has been given. A novel procedure for the computation of the mutual shielding contribution to the damage to internal components has been presented, based on the geometrical interaction between the sec- ondary debris ejecta, modelled as conic shapes and the shielding and target components. The overall methodology has been tested and compared with state of the art software packages, showing very good agreement with tradi- tional impact assessment methodologies. Such a comparison served also as a veri cation test for the fully probabilistic approach adopted, where the pen- etration probability is computed as the product of the probabilities of three The execution time for the Shielding and Linear shielding cases are equivalent 39 separate events. Also, the mutual shielding capabilities have been compared with ESABASE2/DEBRIS, showing comparable results and demonstrating that it is possible to predict the mutual shielding between components using the interaction of geometrical shapes and avoiding ray tracing methodologies. Alongside the obtained results, it should be mentioned that the presented methodologies is in its initial level of development and has been veri ed against limited test cases and scenarios. Consequently, for future devel- opment, a more complete set of tests should be executed, including more complex internal con gurations, with an increasing number of components and di erent shapes. This is especially useful for verifying the adaptability and scalability of the mutual shielding procedure to more complicated ge- ometries. Moreover, di erent shielding methodologies should also be tested, speci cally including honeycomb sandwich panels as they are among the most used spacecraft structures. Acknowledgements This work was funded by EPSRC DTP/CDT through the grant number EP/K503150/1. References Bjorkman, M.D., Christiansen, E.L., Lear, D.M., 2014. Bumper 3 Software User Manual. Technical Report NASA/TM-2014-218559. NASA. Bunte, K., Farahvashi, E., Miller, A., 2017. Methods to reduce uncertainties in spacecraft vulnearbility predictions, in: 7th European Conference on Space Debris, ESOC, Darmstadt, Germany. Bunte, K.D., Destefanis, R., Drolshagen, G., 2009. Spacecraft Shielding Lay- out and Optimisation Using ESABASE2/Debris, in: Proc. 5th European Conference on Space Debris. Christiansen, E., Arnold, J., Corsaro, B., Davis, A., Giovane, F., Hyde, J., Ratli , M., 2009. Handbook for Designing MMOD Protection. Technical Report NASA/TM-2009-214785. NASA Johnson Space Center. Huston, Texas, USA. 40 Depczuk, D., Schonberg, W.P., 2003. Characterizing debris clouds created in oblique orbital debris particle impact. Journal of Aerospace Engineering 16, 177{190. Flegel, S., Gelhaus, J., Wiedemann, C., Vorsmann, P., Oswald, M., Stabroth, S., Klinkrad, H., Krag, H., 2009. The MASTER-2009 Space Debris En- vironment Model, in: 5th European Conference on Space Debris, ESOC, Darmstadt, Germany. G ade, A., Miller, A., 2013. ESABASE2/Debris Release 6.0 Technical De- scription. Technical Report R077-231. etamax space GmbH. Gelhaus, J., Kebschull, C., Braun, V., Sanchez-Ortiz, N., Parilla Endrino, E., de Oliveira, J.C., Dominguez-Gonzalez, R., 2014. Upgrade of ESA's Space Debris Mitigation Analysis Tool Suite - Final Report. Technical Report ESA contract 4000104977/11/D/SR. European Space Agency. Gelhaus, J., M ockel, M., Wiedemann, C., Kempf, D., Krag, H., 2011. MASTER-2009 Software User Manual. Technical Report. M09/MAS- SUM. Gelhaus, J., Sanchez-Ortiz, N., Braun, V., Kebschull, C., de Oliveira, J.C., Dominguez-Gonzalez, R., Wiedemann, C., Krag, H., V orsmann, P., 2013. Upgrade of DRAMA ESA's space debris mitigation analysis tool suite, in: 6th European Conference on Space Debris, Darmstadt, Germany. Gillies, S., et al., 2007{. Shapely: manipulation and analysis of geometric objects. URL: https://github.com/Toblerity/Shapely. Grassi, L., Tiboldo, F., Destefanis, R., Donath, T., Winterboer, A., Evans, L., Janovsky, R., Kempf, S., Rudolph, M., Sch afer, F., Gelhaus, J., 2014. Satellite vulnerability to space debris { an improved 3D risk assessment methodology. Acta Astronautica 99, 283{291. doi:10.1016/j.actaastro. 2014.02.006. Gulde, M., Kempf, S., Sch afer, F., 2016. Fast and Flexible Space Debris Risk Assessment for Satellites. Journal of Space Safety Engineering 3, 111{113. doi:10.1016/S2468-8967(17)30003-4. 41 Inter-Agency Space Debris Coordination Committee, Committee, I.A.S.D.C., 2007. IADC space debris mitigation guidelines. Inter-Agency Space Debris Coordination Committee. Kempf, S., Sch afer, F., Rudolph, M., Welty, N., Donath, T., Destefanis, R., Grassi, L., Janovsky, R., Evans, L., Winterboer, A., 2013. Risk and vulnerability analysis of satellites due to MM/SD with PIRAT, in: 6th European Conference on Space Debris, Darmstadt, Germany. pp. 22{25. Kempf, S., Sch afer, F.K., Cardone, T., Ferreira, I., Geren e, S., Destefanis, R., Grassi, L., 2016. Simpli ed spacecraft vulnerability assessments at component level in early design phase at the European Space Agency's Concurrent Design Facility. Acta Astronautica 129, 291{298. doi:10. 1016/j.actaastro.2016.08.014. Kuiper, W., Drolshagen, G., Noomen, R., 2010. Micro-meteoroids and space debris impact risk assessment for the ConeXpress satellite using ESABASE2/Debris. Advances in Space Research 45, 683{689. Lewis, H.G., Radtke, J., Beck, J., Bastida Virgili, B., Krag, H., 2017a. Self- induced collision risk analysis for large constellations, in: 7th European Conference on Space Debris, ESOC, Darmstadt, Germany. Lewis, H.G., Radtke, J., Rossi, A., Beck, J., Oswald, M., Anderson, P., Bastida Virgili, B., Krag, H., 2017b. Sensitivity of the space debris en- vironment to large constellations and small satellites, in: 7th European Conference on Space Debris, ESOC, Darmstadt, Germany. Lewis, H.G., White, A.E., Crowther, R., Stokes, H., 2012. Synergy of debris mitigation and removal. Acta Astronautica 81, 62{68. doi:10.1016/j. actaastro.2012.06.012. Liou, J.C., Johnson, N.L., 2006. Planetary science - Risks in space from orbiting debris. Science 311, 340{341. doi:10.1126/science1121337. Martin, C., Cheese, J., Brandmueller, C., Bunte, K., Fritsche, B., Lips, T., Klinkrad, H., Sanchez, N., 2005a. A debris risk assessment tool support- ing mitigation guidelines, in: 4th European Conference on Space Debris, ESA/ESOC, Darmstadt, Germany. pp. 345{352. 42 Martin, C.E., Cheeses, J.E., Sanchez-Ortiz, N., Klinkrad, H., Bunte, K., Hauptmann, S., Fritsche, B., Lips, T., 2005b. Introducing the ESA DRAMA tool. Science and Technology Series 110, 219{233. Miller, A., Zaake, M., Gromann-Ruh, F., Bunte, K., Millinger, M., Drol- shagen, G., 2017. Recent extensions of the esabase2/debris impact risk assessment tool, in: 7th European Conference on Space Debris, ESOC, Darmstadt, Germany. O'Connor, B., 2008. Handbook for Limiting Orbital Debris. NASA Handbook 8719.14. National Aeronautics and Space Administration, Washington, DC Putzar, R., Sch afer, F., 2006. Vulnerability of spacecraft equipment to space debris and meteoroids impacts. Technical Report Report No. I-15/06. Ernst-Mach-Institute, Freiburg. Germany. Radtke, J., Stoll, E., Lewis, H.G., Bastida Virgili, B., 2017. The impact of the increase in small satellite launch trac on the long-term evolution of the space debris environment, in: 7th European Conference on Space Debris, ESOC, Darmstadt, Germany. Reimerdes, H.G., Wohlers, W., 2001. Optimization of micrometeoroid and space debris protection systems, in: Space Debris, pp. 655{660. Ryan, S., Christiansen, E.L., 2010. Micrometeoroid and orbital debris (MMOD) shield ballistic limit analysis program. Technical Report NASA/TM-2009-214789. Houston. Ryan, S., Christiansen, E.L., 2011. A ballistic limit analysis programme for shielding against micrometeoroids and orbital debris. Acta Astronautica 69, 245{257. Ryan, S., Sch afer, F., Destefanis, R., Lambert, M., 2008. A ballistic limit equation for hypervelocity impacts on composite honeycomb sandwich panel satellite structures. Advances in Space Research 41, 1152{1166. Sch afer, F., Lambert, M., Christiansen, E., Kibe, S., Stokes, H., Reimerdes, H.G., Zengyao, H., 2005. The inter-agency space debris coordination com- mittee (IADC) protection manual, in: 4th European Conference on Space Debris, ESA/ESOC, Darmstadt, Germany. p. 39. 43 Sch afer, F.K., Ryan, S., Lambert, M., Putzar, R., 2008. Ballistic limit equa- tion for equipment placed behind satellite structure walls. International Journal of Impact Engineering 35, 1784{1791. Schaub, H., Jasper, L.E.Z., Anderson, P.V., McKnight, D.S., 2015. Cost and risk assessment for spacecraft operation decisions caused by the space debris environment. Acta Astronautica 113, 66{79. Schonberg, W.P., 2001. Characterizing secondary debris impact ejecta. In- ternational Journal of Impact Engineering 26, 713{724. Space Debris Oce, 2017. Space Debris by the Numbers. URL: http://www.esa.int/Our_Activities/Operations/Space_Debris/ Space_debris_by_the_numbers. Stokes, H., Cougnet, C., Gelhaus, J., Oswald, M., Schaefer, U., Theroude, C., 2012. A Detailed Impact Risk Assessment of Two Low Earth Orbiting Satellites, in: 65th International Astronautical Congress, Naples, Italy. Stokes, H., Swinerd, G., Walker, R., Wilkinson, J., 2000. Achieving cost e ec- tive debris protection of unmanned spacecraft using shield, in: Bendisch, J. (Ed.), Space Debris 2000, American Astronautical Society. pp. 175{186. URL: https://eprints.soton.ac.uk/21579/. series ISSN 0278-4017. Stokes, P.H., Swinerd, G.G., 2005. Debris protection optimisation of a re- alistic unmanned spacecraft using SHIELD, in: 4th European Conference on Space Debris, p. 515. Trisolini, M., Lewis, H.G., Colombo, C., 2015. Survivability and Demise Criteria for Sustainable Spacecraft Design, in: 66th International Astro- nautical Conference, Jerusalem. Trisolini, M., Lewis, H.G., Colombo, C., 2016. Demise and survivability crite- ria for spacecraft design optimization. Journal of Space Safety Engineering 3, 83{93. doi:https://doi.org/10.1016/S2468-8967(16)30023-4. Trisolini, M., Lewis, H.G., Colombo, C., 2018a. Demisability and surviv- ability sensitivity to design-for-demise techniques. Acta Astronautica 145, 357{384. doi:10.1016/j.actaastro.2018.01.050. 44 Trisolini, M., Lewis, H.G., Colombo, C., 2018b. Spacecraft design optimisa- tion for demise and survivability. Aerospace Science and Technology 77, 20. doi:10.1016/j.ast.2018.04.006. Welty, N., Rudolph, M., Sch afer, F., Apeldoorn, J., Janovsky, R., 2013. Computational methodology to predict satellite system-level e ects from impacts of untrackable space debris. Acta Astronautica 88, 35{43. doi:10. 1016/j.actaastro.2013.01.021. Appendix A. Material database In Table A.7 are summarised the values of the coecients for the SRL BLE for the two di erent cases of Aluminium outer bumper plate and CFRP outer bumper plate (Christiansen et al., 2009). Table A.7: Summary of the coecients for the SRL BLE Symbol Aluminium outer bumper CFRP outer bumper V 3 km/s 4.2 km/s LV V 7 km/s 8.4 km/s HV K 1.4 1.1 3S K 0.4 0.4 3D K 1.5 1 tw K 0.1 1 S2 2/3 1/3 4/3 if 45    65 4/3 4/3 if 45    65 8/3 if 45    65 10/4 if 45    65 1/3 2/3

Journal

PhysicsarXiv (Cornell University)

Published: Mar 10, 2020

There are no references for this article.