Influence of aggregate size and free water on the dynamic behaviour of concrete subjected to impact loading

Concrete is a material widely used in civil engineering. Thus the knowledge of its mech anical behaviour is a major safety issue to evaluate the ability of a stru cture to resist to an intense dynamic loading. In this study, two experimental techniques have been applied to a micro-concrete and a co mmon concrete to assess the influence of the aggregate size on the dynamic response. First, spalling tests o n dry and wet specimens have been performed to characterize the tensile strength of concrete at strain rates in the range 30 – 150/ s. Then, edge-on i mpact t ests in sarcophagus configuration have been conducted. The cracking pattern of the micro-concrete and the concrete plates in wet and dry conditions have been compared to appraise the influence of aggregate size and free water on the damaging process.


Introduction
Although concrete is widely used for construction all around the world, its mechanical behaviour is not well understood in dynamic conditions.The tensile properties of concretes constitute generally the main weakness o f t hese materials.T outlemonde [1] perf ormed direct te nsile te sts on diff erent concretes at relatively low strain rates (up to 1/s).The influence of aggregate size and th e free water were particularly i nvestigated.The results s howed that, in this ran ge of load ing rate, th e aggregate size presents a limited influence on the tensile strength while the free water appeared as a prominent parameter.An en hancement of tens ile stre ngth of 3 MPa w as observ ed bet ween qu asi-static v alue (obtained at 1 e-5/s) and t he dy namic on e ( at ab out 1/s) f or wet speci mens while dr y co ncrete samples present a limited improvement of about 1 MPa in the same range of loading rate.At higher rates of strain, spalli ng te sts are suitable to identify t he dy namic strength of co ncrete [2 -4].Klepaczko and B rara [2 ] a dapted th e Ho pkinson bars ap paratus to test co ncrete in d ynamic conditions b y re moving t he ou tput bar .Using th is tech nique, several experi mental data h ave been gathered on wet and dr y specimens.Unfortunately, no measurement was performed directly on the specimen.Based on t he same principle, Schuler et al. [3] used an acceleration se nsor placed on the rear face of the specimen to get the free surface velocity, and computed the tensile strength from this signal.W eerheijm and Va n Doormaal [ 4] realised e xperiments o n co ncrete using a si milar configuration.They u sed str ain g auges on the tested concrete specimen to obtai n local data concerning the loading field that allowed deducing the tensile strength.In this study, two concretes are co mpared: the MB 50 micro-concrete with a maximum aggregate size o f 2 mm and t he R30 A7 concrete, a co mmon concrete with a maximum grain size o f 8 mm.In the first part, these materials are presen ted and th eir m ain m echanical properties are g iven.A seco nd part is ded icated to th e experimental ca mpaign of spalling test s perf ormed to i nvestigate t he sensitivity o f th e tensile strength to the strain rate.In the third part, another technique is applied to investigate the response of a co ncrete targ et to a ballisti c i mpact: t he so-called edg e-on i mpact tests h ave bee n co nducted o n both co ncretes.Consequently, by crossc hecking all experimental data, the i nfluence o f ag gregate size and free water on the dynamic response of concrete has been appraised.

The MB50 micro-concrete
Due to their macroscopic heterogeneities, concretes are generally difficult to test in dynamic conditions.The MB50 micro-concrete has been designed to be representative of a standard concrete with a small maximum aggregate size (2 mm): the distribution of aggregates and the water to cement ratio are similar to a common concrete.This particularity allows reducing the tested volume which is convenient for laboratory testing.Its main properties are gathered in Table 1.Several experimental data are available for this material in direct tension [1], in bending [5] or in splitting [6].Moreover, its compressive behaviour has been studied in simple compression [7] and under high confining pressure [8][9][10].

The R30A7 concrete
The R30A7 concrete has been designed to be representative of a standard concrete.Oppositely to the MB50 micro-concrete, bigger inclusions are included: its m aximum aggregate size is 8 mm.Its behavior has already been studied in confined compression [11] and quasi-oedometric tests [12].Its composition and its main mechanical properties are reported in Table 1.

Tested specimens
The s pecimens used for s palling test s are c ylinders 46 mm in dia meter a nd 120 or 140 mm i n length.T o perf orm ed ge-on i mpact test s plates of 200 x 120 x 15 mm 3 w ere u sed.They were all obtained from large blocks (30 x 30 x 20 cm 3 ) b y d rilling, c utting and g rounding.Af ter th e machining processe s, t hey h ave been stored i n water saturated by li me to a void t he dis solution o f portlandite.On the on e hand, sat urated speci mens were pi cked u p from water le ss th an on e hour before testing and re gularly re-hydrated during their preparation.On the other hand, a sec ond set o f specimen was des iccated at 6 0°C du ring several weeks.T he los s o f water was reg ularly c hecked until the mass of the specimen stabilized.

Dynamic tensile testing: spalling experiments
To test th e concrete at high strain rates, one can conduct spalling tests.During this experiment, the impact of a projectile generates a compressive pulse propagating through an instrumented bar (cf.Fig. 1).A p art of th e wave i s trans mitted t o th e concrete cyli nder while th e oth er part is ref lected back in the bar.The compressive loading propagates through the specimen on which several strain gauges are placed.When it reaches the free surface, the incident compressive pulse is reflected into a tensile wave propag ating i n the oppos ite direction .A tensile f ield appears along the s pecimen leading to its dynamic failure.These tests are unu sual because t he mechanical eq uilibrium of t he speci men is n ever reached during th e e xperiment.In thi s stu dy, t he b asic setup co nsists of a proj ectile and a Ho pkinson bar diameter 46 mm a nd le ngth r espectively 75 an d 1200 mm m ade of aluminium alloy to redu ce the impedance mismatch with concrete.Several strain gauges are placed on the input bar to characterize the incident loading.Other gauges are glued directly on the specimen.Linked to the high frequency scope (Ban dwidth: 500 MHz, recordin g f requency: 10 MS/s), th ey allo w recording t he s tress field evolution during the spalling test.Moreover, a la ser extensometer (Bandwidth: 1.5 MHz) points out the rear f ace of the tested specimen to g et the free velocity signal.From this last experimental data the d ynamic te nsile stre ngth σ dyn is ide ntified using t he li near acou stic ap proximation of No vikov [13]: where ρ is the density, C 0 is the one-dimensional wave velocity and ΔV pb is the pullback velocity, i.e. the diff erence bet ween t he maximum v elocity reac hed at th e rear f ace and th e reb ound v alue of velocity (cf.Fig. 2).The experim ental ca mpaign w as co nducted on dry and wet speci mens of concrete and m icroconcrete.During the experiments, an ultra-high speed camera with a maximum frame rate of 1 Mfps has been used to study the fragmentation kinetics in both concretes.After the test, frames have been post-processed with the Correli Q4 data image correlation (DIC) software (LMT-Cachan) to perform quantitative measurements of th e di splacement field.An example of spalling test perform ed o n MB50 micro-concrete specimen is presen ted in Figure 3a.T he corresponding field of displace ment (in pixels) obtain ed b y usi ng Correli Q4 is sh own.No crack s are v isible on th e specimen at t his point.Nevert heless t he co rrelation results reveal several di scontinuities of displace ment.In Fi gure 3b th e displace ment ded uced f rom t he velocity signal o f laser ex tensometers is co mpared to th e average v alue of fragments, time t = 0 co rresponding to th e begi nning of ten sile stres ses in t he concrete s pecimen accord ing to s train gauges.A g ood ag reement bet ween these t wo tec hniques i s observed.This technique has been applied to spallin g test p erformed on R30A7.Figure 3c presents the displace ment field and s hows t hat t he detected discontinuities co rrespond to f racture planes of the specimen.Thus the DIC method allows detecting early the damage in the specimen and carrying out an accurate evaluation of the strain field.
The spalling strength sensitivity to the strain rate of the micro-concrete and the concrete has been studied bet ween 30 a nd 150/s b y v arying t he i mpact velocity o f t he proj ectile.Dynamic tensile strengths obtained are plotted in Figure 4.The trends of the MB50 micro-concrete are similar to data of R 30A7 con crete.T his obs ervation s upports th e hy pothesis t hat a ggregate s ize has a li mited influence even at high strain rates.4 Damage under impact: edge-on impact test

Principle of the EOI experiment
A localized dynamic loading on a co ncrete structure like a ballistic i mpact or a deto nation leads to characteristic outcomes.In order to i mprove the understanding of the fragmentation of concrete, the edge-on impact test was developed at the Ernst Mach Institute -Germany [14] and at the Centre Technique d'Arcueil -France [15].This experimental technique has been designed to reproduce in a two-dimensional configuration the loading of a ballistic impact.The growth of damage is visualized using a high speed ca mera at the surface o f t he target.T he principle of the e xperiment co nsists in projecting a stri ker on the edge of a plate composed of the material to be test ed.The loading wave generates a high co mpressive zo ne n ear t he i mpacted zo ne and th en spreads i nto t he t arget.T he radial displacement of matter induced by the passage of the incident pulse generates dynamic tensile stresses in the hoop directio n th at result i n intense f ragmentation composed of radial crack s.Ceramics [14], glasses [15], ultra-high strength concrete [16] have been tested with edge-on impact tests.In th is st udy, several tests were perf ormed on dry and wet speci mens of the M B50 m icro-concrete and of the R30A7 concrete in the so -called sarcophagus configuration: concrete plates are encapsulated in an aluminium box that keeps fragments clo se to th eir original position (cf.Fig. 5).After the test, a coloured hyperfluid resin is injected to highlight the damage pattern.Additionally, a dynamic confinement system [5] was used to locally increase the pressure in the projectile-specimen contact zo ne at the begi nning o f t he load ing, red ucing t his way t he co mpressive da mage a nd improving the spread of the incident wave in the tile (see Fig .5).

Experiments conducted on the MB50 micro-concrete
Several numerical simulations were conducted to determine a test configuration (impact velocity and length of projectile) that creates a dy namic tensile loading in the concrete target comparable to that observed in spalling tests.It was established that a projectile 22.5 mm in diameter and 100 mm in length with an initial velocity of 50 m/s may generate a tensile strain rate of about 150/s at 50 mm from the impact spot [17].
First EOI tests were carried out on MB50 micro-concrete plates of 200 x 120 x 15 mm 3 .T he specimens were i nfiltrated post mortem and polis hed to reveal t he cracks.T he resu lts obtain ed on dry and wet specimens are reported in Figure 6.It can be remarked that a higher cracking density has developed during th e te st per formed on a dr y s pecimen.Moreover, th e e xisting crac ks in t he wet target see m t hinner a nd more clo sed.Con sequently, free water i n t he micro-concrete sh ows a significant i nfluence on the damage pattern.P erforming these test s o n co ncrete R30 A7 al lows determining the influence of aggregate size on damages due to an impact loading.14th International Conference on Experimental Mechanics

Experiments conducted on the R30A7 concrete
Standard c oncretes in clude generally ag gregates a t t he c entimetre sca le.Co nsequently, it i s necessary to evaluate the influence of aggregate size on the dynamic response of concrete.EOI tests were performed on dry and wet specimens of R30A7 concrete keeping the previous parameters: the projectile has a d iameter of 2 2.5 mm and a length of 100 mm and i s proj ected onto the target at a speed of 50 m/s.Like for the MB50 micro-concrete, the specimens were infiltrated post-mortem and polished to bring out the cracks.The damage patterns are presented Figure 7. Again, the influence of free water is obvious: the damage is much more pronounced in the dry tile.In the same way as the MB50 m icro-concrete, an int ense cracking develo ped n ear th e i mpacted area.Farth er, many long radial crack s are obs erved.Oppositely, in the wet tile the cracking network i s le ss dev eloped an d cracks are hardly v isible.B esides, on e can se e th at few ag gregates are brok en: crack s h ave circumvented the inclusions during their propagation.

Maximum aggregate size influence on damage under impact
Despite si gnificant diff erences i n ter ms o f microstructure bet ween MB50 micro-concrete and R30A7 co ncrete, da mage patterns produ ced b y i mpact ar e v ery s imilar: i n wet co nditions small cracks and li mited crack ope ning are n oted, whereas in d ry co nditions a pron ounced damage is observed.It is i nteresting to n ote t hat few a ggregates ar e brok en, crack ing being pre dominantly inter-granular: thus the matrix behaviour see ms to drive t he dynamic fragmentation.Co nsequently, the maximum size of ag gregates ap pears to play a limited role on th e crac king pattern of sa mples subjected to im pact.Otherwise, the free water ap pears as a m ore i mportant factor.T he differences between dry and wet samples are more pronounced than between micro-concrete and concrete.

Conclusions
To assess the importance of aggregate size on the tensile behaviour of concrete subjected to high speed dynamic loading, two experimental methods have been used.On the one hand, a campaign of spalling tests has been co nducted on a standard concrete with a maximum aggregate si ze of 8 m m and a micro-concrete (maximum grain size: 2 mm).For both materials these experiments have been carried ou t on dry an d wet specimens.T he d ynamic tens ile tests performed bet ween 30 an d 150/s showed very similar results for both concretes despite their difference of the microstructure size.The 39007-p.7 moisture has shown a great influence on the dynamic strength of both concretes.Moreover an ultrahigh speed ca mera has been used.T he acqu ired f rames have been post -processed with a Digital Image Correlation so ftware.It allo wed identi fying the fracture planes at th e earl y sta ge of damage and evaluating finely the strain field evolution during the spalling test.On the other hand, EOI tests have been co nducted to asse ss th e influence of aggregate size on th e da mage pattern of co ncrete when subjected to an impact loading.Similar trends than in spalling tests have been observed in both concretes: t he crack ing is predominantly i nter-granular.Agai n, th e free water changes more significantly th e da mage patt ern.T hese observ ations show th at t he res ults obtai ned f rom d ynamic tensile experiments on a micro-concrete, easier to test, can be transposed to a standard concrete.

Fig. 3 .
Fig. 3. (a) Displacement field measured by Digital Image Correlation (scale in pixels) in a spalling test (MB50 specimen -49Dry test), (b) Comparison of axial displacements deduced from DIC and from the laser extensometer signal (MB50 49Dry test), (c) Comparison of the displacement field (in pixels) and the post mortem cracking pattern of a R30A7 specimen (23Wet test).

Fig. 4 .
Fig. 4. Dynamic tensile strength of the MB50 and the R30A7 concretes obtained from spalling tests.

Fig. 6 .
Fig. 6.Cracking pattern of EOI tests performed on (a) a wet MB50 tile and (b) a dry MB50 tile.

Fig. 7 .
Fig. 7. Cracking pattern of EOI tests performed on (a) a wet R30A7 tile and (b) a dry R30A7 tile.

Table 1 .
Composition and main mechanical properties of the MB50 and the R30A7 concrete.