Passive safety system for small unmanned aerial vehicles

In this paper a new air-bag prototype suitable for protecting valuable objects mounted on the drone is presented. The paper provides a complimentary study involving both numerical simulations and experimental study. The experimental research results are presented for typical air-bag's textile material and were used as a base for the material model calibration process. This model was used for the numerical simulations of the proposed air-bag prototype, which were carried out in the LS-Dyna environment. Based on the outcome of the study, the proposed prototype seems to be a suitable device for preventing the unmanned vehicle equipment from unexpected accidents.


Introduction
Unmanned aerial vehicles (UAVs) have been dynamically developing during recent years, especially for commercial applications. UAVs are being used not only for the simplest tasks such as film recording or capturing pictures but also are used for more complex and responsible actions such as structure inspections or even to package delivering. Teleoperated multirotors are also used by the firefighters or police. In order to perform such operations, UAVs often require the installation of fragile, sensitive and very expensive additional equipment.
Statistics show that failures of the multirotors in the air and in consequences their crash with a ground are common. The collisions are very often caused by the components failure. Despite the fact that the design process and testing methodology have been strongly developed recently [1], it is technically impossible to eliminate all the accidents. According to [2] during the flight the most dangerous stage is landing. Although there are some landing systems available for manned airplanes, such as used at major airports Instrument Landing Systems (ILS), which significantly limit the number of crashes, they are not popular for unmanned aerial vehicles because of its complexity and weight. The landing processes of UAVs are therefore generally controlled manually or supported by more or less sophisticated portable external aiding systems [3].
Another important issue is related to various mishaps occurring in the air, due to uncontrolled collisions with other flying or ground objects such as electric poles, transmission and telephone lines or even birds. Such situations generally lead to serious damages of UAVs and in consequences of the equipment attached to them. Despite the fact that various systems preventing air accidents are continuously improved and developed [4] there is still need to enhance their effectiveness.
An interesting solution reducing landing threats for UAVs is proposed in the paper [5], which presents a vision-based automatic landing system that uses a dome-shaped airbag. In the proposed solution the air-filled dome enables airplanes to land softly on the specially colored copula using advanced navigation sensors. Nevertheless, such a solution is insufficient, when the object was damaged in the air. Other interesting proposition for passive safety system was described in [6], which presents a parachute mounted on the UAV.
In this paper a new concept of passive safety system for unmanned aerial vehicles will be presented. The proposed solution consists of a special parachute and airbag attached to the UAV. The parachute protects the object in case of failure at high altitudes, whereas the airbag protects the device in case of damages just above the ground, where the parachute is ineffective. In this paper the authors focus on the design process of an air-bag. In subsequent sections the full concept of this innovatory system is presented in details.

Material testing
The concept of the discussed airbag was created with finite element method (FEM) using LS-DYNA environment. The most important stage in the airbag modeling process is calibrating a suitable material model, which must be backed up with experiments. The airbag is composed of a woven fabric which is rapidly inflated during a crash. For the analysis of the textile fabric it is necessary to perform a wide range of experiments. Beside the standard uniaxial tension test in various directions, it is necessary to conduct the shear test and additionally, what is highly recommended, biaxial tests. As far as the uniaxial tests is concerned, the experiment is relatively simple to carry out. Shear test and biaxial test are generally much more complex and generate more problems. Shearing tests are conducted in few different ways including bias extension test [7] or picture frame test [8,9]. Among popular biaxial test methodologies a hemisphere test is worth mentioning, biaxial test with cross-shape sample or inflated cylinder fabric test [10].
In this paper the results of simple uniaxial tests and bias extension tests of an airbag fabric PA 6.6 are presented. Samples used in uniaxial tests ( Fig. 1) were stretched on MTS testing machine, however strains were measured by additional non-contact 3D deformation measuring system (ARAMIS). During bias extension tests the strains were measured based on the testing machine yaws displacement. Fig. 2 shows the typical results of the simple uniaxial tests. In Fig. 3 the results of bias extension tests are depicted. Additionally the photos recorded by the ARAMIS system including calculated strain maps are presented in Fig. 4. More detailed description of the fabric material laboratory tests can be found in the previous papers of the authors [11,12].
The data acquired from the experiments were used in the next stage of modeling implemented to numerical model prepared in LS-Dyna software. This topic is described in details in the next part of this paper.

Experimental research of leakage behavior
The crucial role in predicting the response of impacting object plays a gas leakage that is caused by fabric porosity or vents located on airbag surface (Fig. 5).

Fig. 5. Schematic illustration of the gas leakage
The modeling of leakage by vents is rather simple and will not be considered in this work. The more complicated task is modeling the leakage by material porosity. LS-DYNA provides various options for the airbag leakage modeling but the model must comply with the experiments. For this reason the airbag was tested on the drop test stand. During the experiments the internal pressure and strain of the airbag fabric were measured. The investigated airbag was filled up to predetermined pressure value and after that the intake of structure was closed. As a result the gas could only flow by fabric pores. The internal pressure was continuously measured during the outflow process. At the same time the average values of strains, calculated from the testing surfaces (Fig. 6) were recorded by the ARAMIS system. As the process of filling the air bag is rapid, the limitations of video measuring system (60 frames per second) did not allow to measure the strains during this stage. Therefore, the strains depicted in the Fig. 6 are equivalent to the maximum pressure reached in the air bag. Fig. 7 depicts the investigated airbag mounted on the test stand with strain maps calculated by the measuring system. In Fig. 8 variations of the internal pressure in time function are shown. Three different values of predefined maximal pressures were taken into consideration in laboratory tests. The obtained results were used in the model validation process to verify numerical results with experiments. Fig. 9 depicts the recorded experimentally characteristics of pressure in function of average strain determined from the area shown in Fig. 6. Fig. 10 confirms the reproducibility of the experimental results obtained from the experiments performed at a maximum pressure in the airbag.

Modeling of material mechanical behavior
To model the real mechanical response of the considered fabric material in the LS-Dyna environment the material model MAT_FABRIC (#34) was used [13]. This material model has been developed for many years beginning with the simple linear constitutive law and up to the latest implementation with a nonlinear biaxial loading and unloading curves. The newest model of the material allows to simulate the porosity of the structure where the gas flows out [14].
In order to ensure that the investigated airbag material model works correctly the comparison between simulation and experimental research was performed. Based on the geometrical dimensions of the tested samples FE models were prepared. Before the experimental data were implemented to FE model it was necessary to transform it into II Piola-Kirchhoff stress tensor and Green-Lagrange strain tensor. For the tension case it can be calculated from Eq. 1 and 2. Similarly, for shear tests it can be obtained from (3) and (4). The derivation of equations (1-4) is precisely described in paper [11,12].
Where T 22 , T 21 , T 12 -elements of stress tensor, E 22 , E 21 , E 12 -elements of strain tensor, f -force measured by testing machine, Ɛ eng -engineering strain, A 0 -sample cross section, l oinitial length of the sample, W -initial width of the sample, t -thickness of the sample, dyaws displacement.
When FE model was completed the analysis were performed using explicit LS-Dyna code. As a result it was possible to verify the experimental and numerical results. To ensure that the model correctly describes the stress and strain state in the material two independent t sts w r p r orm . Th irst o was r at to th simp u iaxia t sio t st. I this prob th o atio o th samp corr spo to th mai mat ria ir ctio . Forc s i a strai u ctio w r t rmi um rica y a i th i a sta th r su ts w r compar with r a xp rim ta r spo s o th mat ria . s it is show i Fi . th mo pr icts pr cis y th b havior o th abric.
I a oth r xp rim t bias xt sio t st, th samp coor i at syst m was rotat by 45 r s a th t sti pi c was a ai strai . I this cas th obs rv i r c s b tw simu atio a xp rim ta r su ts ar mor si i ica t (Fi .
. It is worth m tio i , that co si ri th r a b havior o a airba , th sh ar compo t i str ss t sor o s ot p ay a si i ica t ro , co s qu t y th authors assum that mo is r iab or urth r um rica a a ys s. Num rica y ca cu at str ss i s rat i oa abric samp s i simp t si t st a bias xt sio t st ar pict i Fi s. 3 a 4 r sp ctiv y.   Th mo has two mat ria co ici ts (FLC (t , F C (p that hav to b i ti i xp rim ta y. Thos co ici ts ha a co sta t va u qua 6 a . r sp ctiv y. M asur m t o th porosity co ici t ca b a v ry i icu t task, but s v ra typ s o air p rm abi ity m asuri i strum ts ar avai ab or ra us i t xti t sti [9]. I ra th i a i vo v s co tro air ow by k ow abric ar a at iv pr ssur . I this pap r a i r t approach was propos . xp rim ts w r p r orm o th r a airba . I th irst st p th i v sti at obj ct was i ita iz by r v rs i ri m tho usi Photomo r. t rwar , th F mo was v op . Th ori i a co c pt o th as aka mo i proc ss is pr s t i Fi . 5. The unfolding process of the air bag was not considered in this paper. It was assumed that this process does not have a significant influence on the air bag behavior as well as for a whole system behavior. In order to improve the transparency of Fig. 21, the airbag was hidden.

Conclusion
In this paper the new concept of the safety system for UAVs was proposed, which combines the advantages of a parachute and the special airbag provided to protect an additional device transported by the flying object. The proposed concept seems to have a great potential for future engineering applications. In the paper not only the new concept of a passive safety system was presented but also the complex methodology that can be adopted for any design process for airbags was proposed. The conducted simulations revealed a technical capability of the proposed UAVs airbag project as all design assumptions were fulfilled. The project was developed based on the models verified with the experiments, what exactly fits in the contemporary way of design. It was shown how the mechanical properties of a textile material should be recorded and modeled. Furthermore, the original approach based on the reverse engineering method for the validation of the model leakage behavior was proposed.
The models of the drone and airbag proposed in this paper were to some extent simplified, as the real dynamics of such a complex rotating systems is very complicated. Consequently, the real behavior of the investigated system could be much more problematic. To fully confirm the correctness of the proposed approach to the discussed problem conducting more complex numerical simulations is required. In the final stage, the experimental research carried out on the real object is necessary to verify the proposed numerical approach.