Advertisement

Impact behavior of triggered and non-triggered crash tubes with auxetic lattices

  • Fatih UstaEmail author
  • Osman F. Ertaş
  • Altuğ Ataalp
  • Halit S. Türkmen
  • Zafer Kazancı
  • Fabrizio Scarpa
Original Paper
  • 140 Downloads

Abstract

In this study, impact behavior of triggered and non-triggered tubes with and without auxetic filler is examined using numerical methods. Material properties of tubes made of aluminum alloy and auxetic lattices utilizing ABSplus plastics are determined using tensile tests. Finite element analyses are performed using LS-DYNA software at 5 m/s impact velocity. Two different trigger shapes are suggested and compared each other and discussed the advantages and disadvantages over non-triggered tubes. For these loading conditions, trigger mechanism provides lower peak forces and higher crash force efficiency (CFE), but lower specific energy absorption (SEA). In addition, the effects of using auxetic fillers in these triggered tubes are investigated in terms of crashworthiness characteristics.

Keywords

Crash tube Auxetic materials Impact loading Energy absorption 

1 Introduction

Thin-walled tubes convert part of the kinetic energy during collision to plastic energy, and therefore help to prevent the crash and damage. Besides, inertial loads, which occur at the beginning of the collision, can be also quite harmful for passengers during a dynamic crash. Metallic cylindrical tubes are widely used as structural components of automobiles (Abdollahpoor and Marzbanrad 2010).

In the literature, there are several studies on the impact behavior of crash box. The main objective of these studies is to increase the crashworthiness of crash boxes. For this purpose, geometrical modifications on tubes, various cross section types and material types have been examined to obtain better designs. For example, Usta et al. (2015, 2018); Usta and Turkmen (2017) studied on stepped and concentric circular crash tubes under axial impact loading using numerical, experimental and optimization techniques. Eyvazian et al. (2014) investigated the effects of corrugations on the deformation behavior, energy absorption, and failure mode of circular aluminum tube and show that corrugated tubes provided higher crashworthiness characteristics. Lee et al. (1999) analyzed metallic tubes including grooves at folding sites where could be pre-estimated by FEA analysis. Triggering improved the energy absorption performance, and half-dented tube was more effective than full-dented tube. On the contrary, energy absorption could be worsening using triggering without consideration of the peak location of the folding wave and inhomogeneous deformation. The geometry of the trigger mechanism could be optimized using various methods (Marsolek and Reimerdes 2004). Hage et al. (2005) investigated various types of trigger mechanism of metallic tubes such as chamfering, triangular hole pattern and geometric imperfection. Force displacement response and folding mechanism could be controlled by changing trigger types.

As folding material, foams and conventional honeycombs are generally used in the tube structures. Beside the traditional honeycombs and foams, auxetic lattices have been studied in recent years. Auxetic lattices are preferable, because anisotropic material properties increase the impact and indentation resistance and energy and shock absorption performance. A prior study on auxetic materials belongs to Lakes (1987) in the literature. Auxetic materials have several application fields such as crash helmets, body armor, and sports clothing. For example, Yang et al. (2018) analyzed the auxetic performance of a conventional and two auxetic honeycomb structures subjected to dynamic and static loading. They provided that using auxetic structures in the sandwich structures have higher shock absorption over non-auxetic structure.

Density and modulus of auxetic structures can be controlled during fabrication which provides to be obtained auxetic structure as required in terms of Poisson ratio (Duncan et al. 2018). Hou et al. (2018) compared a uniform cell design with a present functionally gradient auxetic cellular structure subjected to low speed collision. Impact tests demonstrated that gradient auxetic crash box has higher energy absorption and lower reaction force than uniform auxetic crash box. In another study, Jiang and Hu (2017) tested PU foam, auxetic and non-auxetic composite structures under low velocity impact and showed that auxetic composite has better energy absorption performance in medium strain range. Direnberger et al. (2013) proved the advantages of the auxetic materials in comparison with honeycombs. Yang et al. (2012) indicated that re-entrant honeycomb structure had good ability of energy absorption under compression load. Lim at al. (2014) investigated the auxetic behavior of polyurethane foam under impact loading. According to their work, auxetic foams could have no advantages over conventional foams under high velocity impact.

For a conventional material, the dog bone specimen elongates in axial direction while its cross section area is shrinking. Therefore, the sign of Poisson’s ratio becomes positive. On the other hand, the sign of the Poisson’s ratio of the auxetic materials is negative. Some researchers examined the mechanical properties of auxetic tubular structures under compression and tension loads (Ren et al. (2016)) and resistance to kinking (Karnessis and Burriesci 2013).

In this work, circular crash tubes with and without auxetic lattices are examined under dynamic loads, numerically. The novel designs of trigger mechanism of crash tubes are introduced and it is discussed that the shape and location of the triggering with consideration of the re-entrant lattices. Two different trigger mechanisms are used for the tube system and compared with the conventional one. The effects of using triggering on tube structures are investigated. Besides, this paper focuses on the exploration the possibility of using auxetic materials with the combination of trigger mechanism in crashworthiness performance of metallic tubes which are not found in the pertinent literature. Impact behavior of re-entrant honeycomb filled tubes is also compared with empty tubes.

2 Determining material properties

2.1 Material properties of AL6063 aluminum alloy

The tubes considered here are made of aluminum alloy AA6063. Owing to insensitivity of aluminum on strain rate, tensile tests are carried out under quasi-static loading at 2 mm/min velocity (Ali et al. 2015). Six test specimens are cut using CNC machine according to ASTM E8/E8M test standard (see, Fig. 1). Biaxial strain gauges are stuck and extensometers are located to define material properties. Material properties of each sample are determined and listed in Table 1. Random errors are calculated according to average value of Young modulus, yield stress and ultimate tensile strength which are 67.69 GPa, 160 MPa, and 192 MPa, respectively. Poisson ratio and density are 0.3 and 2710 kg/m3, respectively. No significant differences are observed between the stress and strain curves of tested specimens. Engineering stress and strain curve is plotted in Fig. 2.
Fig. 1

Tensile test specimens of AL6063 aluminum alloy

Table 1

Mechanical properties and random errors of AL6063 aluminum alloy

Specimen no

Young modulus

(GPa)

Random error

(%)

Yield strength

(MPa)

Error

(%)

Ultimate tensile strength

(MPa)

Random error

(%)

1

70.44

4.06

145

9.47

182

5.29

2

65.33

3.49

165

3.02

195

1.47

3

65.49

3.25

151

5.72

180

6.33

4

66.12

2.32

172

7.39

202

5.12

5

71.27

5.29

173

8.01

210

9.28

6

67.49

0.30

155

3.23

184

4.25

Average

67.69

 

160

 

192

 
Fig. 2

Engineering stress and strain curve of AL6063 tube

2.2 Material properties of ABSplus plastics

The auxetic lattice is made of ABSplus plastics. Strain rate has significant effects on mechanical properties of ABSplus plastics (Rodríguez 2001). In this study, tensile tests of ABSplus plastic are revealed under quasi-static loading due to having no possibility of split-Hopkinson pressure bar test system. Four test specimens are produced with FDM (fused deposition modeling) technology using Dimension Elite 3D printer machine (see, Fig. 3). Specimens are printed at room temperature (effective temperature range 15–30 °C) and interior filling style is selected as solid which provides more durable and stronger parts. After performing tension tests, mechanical properties of each sample and random errors are determined (see, Table 2). Average Young modulus is 1.53 GPa, ultimate tensile strength is 23.25 MPa, yield stress is 19.73 MPa, density is 1040 kg/m3 and Poisson’s ratio is 0.36. Engineering stress and strain curve is plotted in Fig. 4.
Fig. 3

Tensile test specimens of ABSplus plastics

Table 2

Mechanical properties and random errors of ABSplus plastics

Specimen no

Young modulus

(GPa)

Random error

(%)

Yield strength

(MPa)

Error

(%)

Ultimate tensile strength

(MPa)

Random error

(%)

1

1.58

3.10

16.8

14.83

21.8

6.24

2

1.39

9.30

21.5

9.00

24.1

3.66

3

1.48

3.43

18.2

7.73

22.8

1.94

4

1.68

9.62

22.4

13.56

24.3

4.52

Average

1.53

 

19.73

 

23.25

 
Fig. 4

Engineering stress and strain curve of ABSplus plastics

3 Tube designs and finite element modeling

3.1 Finite element modeling

The numerical studies are conducted using LS-DYNA software, which is based on explicit time integration method and desirable program especially for dynamic crash analysis. A rigid mass is dropped on the tubes with a speed of 5 m/s. Rigidwall_Planar_Moving_Forces is used as an impactor to model the impact behavior. The nodes at the bottom of the tube structure are clamped. A schematic view of tube and rigid impactor is plotted in Fig. 5. Three different tube structures with and without re-entrant honeycomb structure are investigated. The diameter, length and thickness of the tubes are 70, 100 and 2 mm, respectively. Auxetic filler type is chosen as re-entrant honeycomb owing to its wider capacity of negative Poisson ratio range in comparison with the other auxetic geometries (Elipe and Lantada 2012). Scarpa et al. (2000) show that geometric cell parameters of reentrant honeycomb have significant effects on the mechanical properties such as in-plane Poisson ratio and Young modulus. The cell parameters of re-entrant honeycomb \(\theta , h, l\) and t are 30°, 10 mm, 10 mm and 1 mm, respectively.
Fig. 5

a Schematic view of rigid impactor and tube, b re-entrant honeycomb cell parameters

Tubes and auxetic fillers are modeled using Belytschko–Tsay shell elements. This element formulation gives greater computational efficiency compared with other shell element formulations (Hallquist 2006). Optimum element size is defined as 2 mm × 2 mm, after comparing the results of analysis of crash tubes at different mesh qualities. Automatic single surface contact algorithms are identified for each specimen, because they have possibility of lapping after deformation. In addition, automatic surface to surface contact interface is defined between surfaces of tube and auxetic filler. The static and dynamic friction coefficients are chosen as 0.3 and 0.2, respectively, for each contact definition.

The material properties of aluminum tube and plastic auxetic lattice are defined with MAT24 (Piecewise Linear Plasticity) material model in LS-DYNA. Elastic material properties of tube and auxetic lattice are used in this material model. True stress and effective plastic strain curve is embedded into this material model for the plasticity of this material. Effective plastic strain is derived from the difference of true strain and the ratio of true stress over Young modulus.

3.2 Tube designs

The first model does not have any triggering on the tube structure so that it is abbreviated as NT (non-triggered) as shown in Fig. 6. Empty tube is abbreviated as “E” and auxetic filler is abbreviated as “A”. The number in the name of model denotes tube length in terms of millimeter.
Fig. 6

NT-E and NT-A models

Second and third models include imperfection on the tubes. Trigger mechanism of second model is an arc which is shown in Fig. 7. Due to circular shape, it is named as circular trigger (CT). Trigger shape of the third model is like in zigzag geometry which is shown in Fig. 8. Because it consists of lines, it is named as Line Triggered (LT). It is aimed to control the crash stability and collapse modes and to increase crashworthiness characteristics of tubes using these trigger shapes. The distance between two arcs and two zigzag lines is chosen as 10 mm and the corner points of them are in same level.
Fig. 7

CT-E and CT-A models

Fig. 8

LT-E and LT-A models

4 Results and discussion

4.1 Formulation of parameters used for comparisons

For the comparison study, some crashworthiness characteristics which are used are listed as below. Energy absorption (EA), peak crash force (PCF), specific energy absorption (SEA), mean crash force (MCF) and crash force efficiency (CFE) are the most common parameters to represent the crash behavior. Energy absorption of the tube under crash load could be expressed as:
$${\text{EA}} = \mathop \int \nolimits_{0}^{\delta } P\left( s \right)ds$$
(1)
where δ is the axial deformation of the structure and P(s) denotes the axial crash force. SEA is calculated as follows:
$${\text{SEA}} = \frac{\text{EA}}{m}$$
(2)
Peak crash force means maximum force value in the axial direction:
$${\text{PCF}} = \hbox{max} \left( {P\left( s \right)} \right)$$
(3)
MCF is calculated as follows:
$${\text{MCF}} = \frac{\text{EA}}{\delta } = \frac{{\mathop \smallint \nolimits_{0}^{\delta } P\left( s \right){\text{d}}s}}{\delta }$$
(4)
Crash force efficiency is calculated as follows:
$${\rm{CFE}} = \frac{{{\rm{MCF}}}}{{{\rm{PCF}}}}$$
(5)

4.2 The effects of length of the tubes on crashworthiness characteristics

To investigate the effects of the length of the tubes on crashworthiness characteristics, all models are analyzed by changing tube length with the length of 100, 125 and 150 mm. The results are compared in terms of reaction force response in time and collapse mechanism. The results show that shorter tubes are better in terms of SEA and MCF, as shown in Table 3. Tube length has no remarkable effects on the crash force response as shown in Fig. 9. Deformation of re-entrant honeycomb of NT-A-100 becomes to shrink initially at the bottom side. But for the longest non-triggered model, collapse mechanism appears at bottom and top surfaces together. Shorter tubes give better SEA and MCF values for CT and LT models. In comparison with CT and LT models with re-entrant structure, all tubes begin to collapse from top surfaces as shown in Figs. 10 and 11. In contrast to non-triggered model, first buckling of CT and LT tubes reveals in the middle section. Therefore, it could be said that trigger system may affect the deformation behavior of filler.
Table 3

Crashworthiness parameters of tubes with auxetic filler

Models

PCF

Deflection

Mass

SEA

MCF

CFE

 

(kN)

(mm)

(kg)

(kJ/kg)

(kN)

 

NT-A-100

61.62

32.70

0.2258

5.491

37.912

0.615

CT-A-100

34.35

39.78

0.2560

4.838

31.141

0.907

LT-A-100

34.28

39.32

0.2512

4.934

31.524

0.920

NT-A-125

60.89

32.74

0.2817

4.396

37.827

0.621

CT-A-125

33.22

41.13

0.3180

3.911

30.235

0.910

LT-A-125

34.29

39.46

0.3132

3.967

31.483

0.918

NT-A-150

62.97

33.55

0.3375

3.676

36.983

0.587

CT-A-150

32.75

41.22

0.3821

3.242

30.049

0.917

LT-A-150

33.73

39.73

0.3750

3.313

31.265

0.927

Fig. 9

Comparison of deformation and force response of NT-A models

Fig. 10

Comparison of deformation and force response of CT-A models

Fig. 11

Comparison of deformation and force response of LT-A models

4.3 Comparison triggered and non-triggered tube designs

Another comparison is made between impact behavior of the triggered and non-triggered tubes with auxetic lattice. Crash force response of NT-A-100, CT-A-100 and LT-A-100 models is plotted in Fig. 12. There is a remarkable difference between force response of non-triggered and triggered tubes. Initial peak crash forces are almost 62 kN and 34 kN, respectively. Although two different peak forces are observed for non-triggered tubes, initial and subsequent force values of triggered tubes become similar. It can be strictly effective on reducing the inertial effects of crash loads on passenger. It is achieved a great reduction of this inertial loads owing to small difference between subsequent forces of triggered tubes.
Fig. 12

Comparison of force response of different samples with auxetic filler

In Table 3, results show that non-triggered tubes have advantages over triggered tubes in terms of MCF and SEA due to higher durability under buckling and bending loads and lower mass. On the other hand, in terms of peak force and crash force efficiency, triggered tubes are obviously better than non-triggered tubes. CFE is the final decisive parameter to determine the most efficient tube design due to including three important results in one formulation. The triggering changes the deformation mode and increases the energy absorption capacity of the axially compressed tubes.

In comparison with the results of triggered tubes, and peak crash force change of CT tubes is smoother than LT tubes. Therefore, it is more useful in terms of inertial loads. But CT tubes are deflected more than LT tubes in axial direction and also its arc geometry causes higher mass. Therefore, LT tubes are better than CT tubes in terms of MCF and CFE. Table 3 illustrates that LT-A-100/125/150 are superior to other types of tubes in terms of CFE.

4.4 The effects of using auxetic lattices on crashworthiness characteristics

In Table 4, the last comparison is shown between empty and filled tubes. It is shown that using re-entrant honeycomb has small effects on the PCF values, axial deformation and efficiency. For example, PCF values of non-triggered empty and auxetic filled tubes are 60.8 and 61.62 kN, crash force efficiency values are 0.614 and 0.615, respectively. Contrary to a conventional honeycomb, the cross-sectional area of the re-entrant honeycomb reduces under compression loads due to its negative Poisson’s ratio. Therefore, the contact surfaces could be reduced and contact forces between the aluminum tube and auxetic material are lower than between the tube and a traditional honeycomb folding. This contact tends to reduce the peak forces on the specimen. Besides, auxetic lattices tend to increase the stiffness of the tube structure during the compressive load phase and can absorb totally more energy than empty tubes. However, re-entrant honeycomb structure can be easily deformed and have a smaller effect on the deflection and brings scarcely any improvement in terms of MCF. This improvement cannot visibly overcome the decline in CFE due to higher peak forces. Accordingly, CFE values of tubes with and without auxetic filler are close to each other.
Table 4

Crashworthiness parameters of tubes with and without auxetic filler

Models

PCF

Deflection

Mass

SEA

MCF

CFE

 

(kN)

(mm)

(kg)

(kJ/kg)

(kN)

 

NT-E-100

60.80

33.43

0.1187

10.513

37.333

0.614

NT-A-100

61.62

32.70

0.2258

5.491

37.912

0.615

CT-E-100

33.01

40.62

0.1490

8.366

30.686

0.929

CT-A-100

34.35

39.78

0.2560

4.838

31.141

0.907

LT-E-100

33.62

40.25

0.1442

8.664

31.038

0.923

LT-A-100

34.28

39.32

0.2512

4.934

31.524

0.920

In addition, it brings two times higher mass than empty tubes which lead to lower SEA values and there is a notable decline in energy absorption capability.

5 Conclusion

In this study, triggered and non-triggered tubes with and without auxetic honeycomb structure are analyzed using LS-DYNA software under dynamic impact loading. The ability of two different trigger systems according to non-triggered tubes is investigated and the effects of using arc and line shape on triggered tubes are evaluated. Triggered tubes with auxetic honeycomb have not been studied before. This paper also focuses on advantages and disadvantages of using auxetic honeycomb structure in the triggered tubes under crash load. The effects of trigger mechanism, auxetic lattice and length of specimen are compared in terms of PCF, SEA, MCF and CFE values. The results are listed as follow:
  • Shorter tubes can provide higher SEA, MCF.

  • Triggered tubes are obviously better in terms of peak crash force and crash force efficiency.

Trigger system affects the deformation behavior of re-entrant structure. Peak crash force change of CT tubes is smoother than LT tubes. Therefore, it is more useful in terms of inertial loads. On the other hand, LT tubes are better than CT tubes in terms of MCF and CFE due to lower deformation in axial direction.
  • CFE values of tubes with and without auxetic filler are close to each other. Re-entrant honeycomb brings two times higher mass of specimen so that SEA values of specimens with auxetic lattices are lower than empty tubes.

  • Table 3 illustrates LT-A-100/125/150 are superior to other types of tubes with auxetic filler in terms of CFE which is our decisive parameter to choose the better design. This result indicates the line triggering is the best choice according to our study.

  • Desirable models can be obtained by changing triggering shape of tube design and geometry of auxetic structures. By optimizing the geometry further, it is believed that the properties of auxetics can be exploited. A parametric optimization study on the lattices, changing the parameters (i.e., h, l, t, s and teta) could achieve this.

Notes

Acknowledgements

Support for this work has been provided by the Scientific and Technological Research Council of Turkey under Project Number 115M465.

References

  1. Abdollahpoor A, Marzbanrad J (2010) Crashworthiness study of axial impact in cylindrical aluminium tubes. J Mech Eng Stroj Časopis 61:1Google Scholar
  2. Ali M, Ohioma E, Kraft F, Alam K (2015) Theoretical, numerical, and experimental study of dynamic axial crushing of thin walled pentagon and cross-shape tubes. Thin Walled Struct 94:253–272CrossRefGoogle Scholar
  3. Dirrenberger J, Forest S, Jeulin D (2013) Effective elastic properties of auxetic microstructures: anisotropy and structural applications. Int J Mech Mater Des 9(1):21–33CrossRefGoogle Scholar
  4. Duncan O, Allen T, Foster L, Gatt R, Grima JN, Alderson A (2018) Controlling density and modulus in auxetic foam fabrications—implications for impact and indentation testing. MDPI Proc 2(6):250Google Scholar
  5. El-Hage H, Mallick PK, Zamani N (2005) A numerical study on the quasi-static axial crush characteristics of square aluminum tubes with chamfering and other triggering mechanisms. Int J Crashworthiness 10(2):183–196CrossRefGoogle Scholar
  6. Elipe JCÁ, Lantada AD (2012) Comparative study of auxetic geometries by means of computer-aided design and engineering. Smart Mater Struct 21(10):105004CrossRefGoogle Scholar
  7. Eyvazian A, Habibi MK, Hamouda AM, Hedayati R (2014) Axial crushing behavior and energy absorption efficiency of corrugated tubes. Mater Des 1980–2015(54):1028–1038CrossRefGoogle Scholar
  8. Hallquist JO (2006) LS-DYNA theory manual. Livermore Softw Technol Corp 3:25–31Google Scholar
  9. Hou W, Yang X, Zhang W, Xia Y (2018) Design of energy-dissipating structure with functionally graded auxetic cellular material. Int J Crashworthiness 23(4):366–376CrossRefGoogle Scholar
  10. Jiang L, Hu H (2017) Low-velocity impact response of multilayer orthogonal structural composite with auxetic effect. Compos Struct 169:62–68CrossRefGoogle Scholar
  11. Karnessis N, Burriesci G (2013) Uniaxial and buckling mechanical response of auxetic cellular tubes. Smart Mater Struct 22(8):084008CrossRefGoogle Scholar
  12. Lakes RS (1987) US Patent No. 4,668,557. Washington, DC: US Patent and Trademark OfficeGoogle Scholar
  13. Lee S, Hahn C, Rhee M, Oh JE (1999) Effect of triggering on the energy absorption capacity of axially compressed aluminum tubes. Mater Des 20(1):31–40CrossRefGoogle Scholar
  14. Lim TC, Alderson A, Alderson KL (2014) Experimental studies on the impact properties of auxetic materials. Phys Status Solid B 251(2):307–313CrossRefGoogle Scholar
  15. Marsolek J, Reimerdes HG (2004) Energy absorption of metallic cylindrical shells with induced non-axisymmetric folding patterns. Int J Impact Eng 30(8–9):1209–1223CrossRefGoogle Scholar
  16. Ren X, Shen J, Ghaedizadeh A, Tian H, Xie YM (2016) A simple auxetic tubular structure with tuneable mechanical properties. Smart Mater Struct 25(6):065012CrossRefGoogle Scholar
  17. Rodríguez JF, Thomas JP, Renaud JE (2001) Mechanical behavior of acrylonitrile butadiene styrene (ABS) fused deposition materials, experimental investigation. Rapid Prototyp J 7(3):148–158CrossRefGoogle Scholar
  18. Scarpa F, Panayiotou P, Tomlinson G (2000) Numerical and experimental uniaxial loading on in-plane auxetic honeycombs. J Strain Anal Eng Des 35(5):383–388CrossRefGoogle Scholar
  19. Usta F, Turkmen HS (2017) Crash behavior of nested tube structures with various cross sections. In Recent Advances in Space Technologies (RAST), 8th International Conference on IEEE, 23–27)Google Scholar
  20. Usta F, Eren Z, Turkmen HS, Kazancı Z, Mecitoglu Z (2015) Numerical investigation of stepped concentric crash tubes subjected to axial impact: The effects of number of tubes, Recent Advances in Space Technologies (RAST), 2015 7th International Conference on. IEEE, 39–43Google Scholar
  21. Usta F, Eren Z, Kurtaran H, Türkmen HS, Kazancı Z, Mecitoglu Z (2018) Crashworthiness optimization of nested and concentric circular tubes using response surface methodology and genetic algorithm. Latin Am J Solids Struct 15(5):1–16Google Scholar
  22. Yang S, Qi C, Guo DM, Wang D (2012) Energy absorption of an re-entrant honeycombs with negative Poisson’s ratio. Appl Mech Mater 148:992–995 (Trans Tech Publications) Google Scholar
  23. Yang C, Vora HD, Chang Y (2018) Behavior of auxetic structures under compression and impact forces. Smart Mater Struct 27(2):025012CrossRefGoogle Scholar

Copyright information

© Springer Nature Switzerland AG 2018

Authors and Affiliations

  1. 1.Faculty of Aeronautics and AstronauticsIstanbul Technical UniversityIstanbulTurkey
  2. 2.Advanced Composites Research Group, School of Mechanical and Aerospace EngineeringQueen’s University BelfastBelfastUK
  3. 3.Bristol Composites Institute (ACCIS)University of BristolBristolUK

Personalised recommendations