Technical Report Dipartimento di Ingegneria Strutturale - Politecnico di Milano

New NDT techniques for the assessment of fire damaged RC structures

Roberto Felicetti 1. Introduction
Concrete is known to exhibit a good behaviour at high temperature, thanks to its incombustible nature and low thermal diffusivity [1], which guarantee a slow propagation of thermal transients within the structural members. As a consequence, very strong temperature gradients are experienced by the reinforcement cover during a fire and the material thermal damage rapidly decreases from a maximum to nil in a few centimetres depth [2]. For this reason, assessing the residual capacity of concrete structures exposed to fire is quite a difficult task, because the traditional destructive or non-destructive testing techniques are generally not suitable for the inspection of such a highly heterogeneous material. The possible approaches to this problem generally involve the inspection of the average response of the concrete cover, a point by point analysis of small samples taken at different depths or some special techniques aimed to interpret the overall response of the concrete member after fire (Table 1). However, the majority of these methods are generally not practical for in situ applications, being either fast but sketchy (e.g. the rebound hammer) or detailed but time consuming (e.g. the point by point analyses). In order to overcome these limitations, an extensive research programme has been performed at Politecnico di Milano in the framework of the UPTUN European Research Project - WP4, aimed both to check the viability of some well-established NDT techniques and to propose at least one quick and easy method for the assessment of the damage experienced by the lining of a tunnel in consequence of a fire. Concerning this latter point, the most desirable feature to be met is the ability to assess the thermal damage profile in one single test, regardless of the quality of the pristine material and with no need for demanding laboratory analyses. This report contains a summary of the activities performed within the research programme, a detailed description of the proposed new method and the results of a series of verification tests which prove its reliability and viability. Average response of the concrete cover Schmidt rebound hammer Windsor probe Capo test BRE internal fracture Ultrasonic Pulse Velocity Point by point response of small samples Small scale mechanical tests Differential Thermal Analysis (DTA) ThermoGravimetric Analysis (TGA) Dilatometry (TMA) Thermoluminescence Porosimetry Colorimetry Micro-crack density analysis Chemical analysis Special interpretation techniques UPV indirect method Impact echo Sonic tomography Modal Analysis of Surface Waves (MASW) Electric Resistivity

Table 1 - Possible approaches to Non-Destructive assessment of fire damaged concrete structures.

2. Materials and thermal damage conditions

The main aspect to be considered in any experimental investigation on the response of thermally damaged concrete is the temperature field reached during the heating phase, which is a function of both the heating rate and the following spell at the maximum temperature. The maximum temperature experienced by the material at each point is usually of prime interest, being the mechanical decay almost totally irreversible and less affected by the cooling process. It has to be noted that in the case of a strong transient, which is the rule in real fires, the temperature of the inner material of a structural member keeps raising during the early cooling phase (Fig. 1). This is due to the heat stored in the external layer, which flows towards the colder part of the member.

ISO 834
30 60 90 120
Cooling phase temperature increase

Cooling

Fire duration Fig.1 - Standard heating and cooling fire curves and maximum temperature envelope within a concrete wall after the exposure to the complete temperature history.

In order to cover the whole range from the material characterization under homogeneous damage conditions to the strong gradients ensuing from real fires, a series of five testing conditions has been considered in this research programme. A) Calibration tests on homogeneously damaged concrete cubes An ordinary concrete and a structural lightweight concrete (average cubic strength Rcm = 50 N/mm2 - max aggregate size = 16 mm - cube side = 150 mm) have been tested in compression as they were or after a slow thermal cycle up to 200, 400, 600 and 800C (heating rate = 0.5C/min, cooling rate 0.2C/min). These concretes exhibited very similar compressive strength decays (Fig. 2), with a significant loss at temperatures higher than 400C, in accordance with Eurocode 2 (Part 1.2: General rules Structural fire design, draft October 2002). The same cubes have been used to calibrate the response of different Non-Destructive Testing techniques in order to ascertain their intrinsic sensitivity to the thermally induced strength loss. B) Concrete panels under a constant temperature gradient The same concretes adopted for calibration tests were used to prepare a couple of small panels (275x550x80 mm) which have been exposed to a marked thermal gradient (> 5C/mm) by heating them on the one side (Tmax = 750C) while keeping cold the opposite side with a fan (Fig. 3). The maximum temperature profile has been determined by means of three embedded thermocouples. These panels are intended as a first, well controlled benchmark for checking the response of different Non-Destructive Testing techniques in the case of thermal damage gradients.

R T / R 20 (%) c c
100 80 60 40 20 0 0 200 400 600 ordinary (R20 =50.4 N/mm 2) C lightweight (R20 =51.1 N/mm 2) C Eurocode 2

T (C)
800

Fig. 2 - Normal density and lightweight concrete cubes for calibration tests and relative decay of the residual compressive strength after a slow thermal cycle (Tmax = 200, 400, 600 and 800C).

800

C
600
Ordinary Lightweight

400

200

0 0 20 40 60 80

Depth (mm)

Fig. 3 - A concrete panel positioned as a replacement for the furnace door and profiles of the maximum temperature recorded within the panels during the thermal cycle.

C) Concrete wall submitted to a ISO 834 standard fire (90 min) A more realistic benchmark for the effect of thermal gradients has been provided by a standard fire test on a concrete duct for electric cabling protection in railway tunnels (ISO 834 fire curve, 90 min duration - Fig. 4). The test was run in a vertical furnace, after closing the specimen in a low-grade reinforced-concrete box (siliceous aggregate concrete - fcm 25 N/mm2). As a consequence, the 0.2 m thick concrete wall on the back of the duct was partly exposed to the burners and partly protected by the tested specimen itself. Even not being the object of the fire test, this panel is an interesting example of the possible not uniform damage pattern resulting from a severe "realistic" fire. Hence, the temperature of the exposed portion has been monitored on both faces and at half thickness. Moreover, the experimental temperature field has been modelled and fit numerically, allowing to continuously plot the maximum temperature profile experienced by this concrete member (including the cooling phase).

1000 800

T (C)

back wall concrete duct

600 400

maximum temperature envelope at the burners turning off

200 0 0 20 40

depth (mm)
60

exposed area

protected area
rebound index
Fig. 4 - Setup for the fire test on a duct for electric cabling; back concrete wall partly exposed to the burners and partly protected by the tested specimen; maximum temperature envelope in the exposed part of the wall and rebound hammer response on both parts of the heated area.

D) Real fire in a precast RC structure The occasion for a first check on the viability of the NDT techniques in the case of a real fire has been provided by the thorough analysis of an industrial building surviving a 4 hour fire. Despite the actual thermal load experienced by each member of this structure is unknown, this case allowed to compare a number of investigation techniques in terms of sensitivity to the thermal damage, time needs for the implementation and in situ practicability. The original grade of the concrete of this structure is typical of precast RC structures (fcm = 45-50 N/mm2).

39 19

45 49

F3
15
cover splitting

42 16 24

Fig. 5 - View of the precast RC structure after the fire, detail of a significantly damaged column and rebound index around its cross-section.

E) Hydrocarbon-pool fire tests at the Virgolo tunnel (February 2005) The last thermal damage condition considered in this programme is strongly related to the tasks of the UPTUN European Research Project. Taking advantage of the lining renovation works in progress in the north channel of the Virgolo tunnel (Bolzano - Italy), the Brennero Highway management decided to run a series of tests on different active and passive fire protection systems and to compare the response of six different shotcrete mixes for lining repair. The fire load was provided by the diesel oil stored in a series of stainless steel tubs arranged next to the side wall of the tunnel. The vault temperature was accurately monitored during the test. A preliminary small scale test was run also, in order to check the behaviour of four additional concrete panels under a very severe fire. The results of the temperature monitoring and the mix designs of the ten concretes at issue have not been circulated so far, but they will be included in a document that is being released by the Brennero Highway management. Concerning the mechanical performances of these concretes, they range from very soft insulating materials to a high strength concrete mix (fcm 120 N/mm2) Besides the actual material features and thermal conditions, this tunnel fire tests allowed a further verification on the viability of different NDT techniques in the case of roughly finished materials (the six shotcrete lining samples) and under difficult operational conditions (hard to approach test sites, interfering research teams, short time available for testing).

Fig. 6 - The north channel of the Virgolo tunnel covered with six different lining systems and the "mini tunnel" containing four concrete panels during the small scale hydrocarbon-pool fire test.

3. NDT techniques and performed tests

On the whole, five NDT techniques have been considered in this study. Two of them are well established methods which have been validated under some of the cited thermal damage conditions: the rebound hammer and the Cut And Pull-Out test (CAPO test). The remaining three methods can be considered to be innovative either for the interpretation of the results (indirect ultrasonic method), for the technique of analysis (concrete colorimetry) or for the test principle itself (drilling resistance). Table 2 provides an overview of the investigated combinations of NDT techniques and thermal damage conditions. Based on its valuable features and good response in any thermal damage condition, the drilling resistance method turned out to be the most promising technique to be proposed as a new quick method for the assessment of fire damaged RC structures. Hence, only a brief description of the other techniques will be given in the following.

Ultrasonic Pulse Velocity

A) calibration tests B) concrete panels C) ISO 834 on a wall D) real fire E) Virgolo tunnel
the method has been successfully applied the method proved to be difficult to apply

Table 2 - Overview of the investigated combinations of NDT techniques and thermal damage conditions.

The rebound hammer test This well-known technique confirmed to be of valuable help for a first, quick monitoring of the severity of the effect of fire on a concrete structure. The simple inspection of the rebound index itself usually allows to recognise the most impaired parts of a member, with no need for specific correlations with the residual strength (see Figs. 4 and 5). Concerning the sensitivity of this method, the results available in the literature are quite dispersed, possibly because of the rebound enhancement due to concrete drying, which might mask the material weakening at the first stage. In the case of concrete members submitted to thermal gradients (test condition B), this method is expected to indicate the average response of a thin superficial layer of concrete. By transforming the experimental temperature profiles (Fig. 3) into rebound index profiles (after the curves of Fig. 7), it appears that the rebound index at the heated surface actually represents the response of the material located at about 1525 mm depth (Fig. 8).

Drilling Resistance

Colorimetry

Capo-test

Rebound hammer

relative rebound index 100% Politecnico di Milano

(150mm cubes)

rebound index 70% decay 60% 50% 40%

Ordinary

80%

Aston University

Lightweight OPC/BFS OPC/PFA OPC Granite Limestone

60%
Ordinary Lightweight OPC/BFS OPC/PFA OPC Granite Limestone

30%
Aston University
(100mm cubes)

40%

20% 10% 0%
0% 20% 40% 60% 80% 100%

20%

Politecnico di Milano

0% 0 200 400 600 800 Temperature [C]

-10% strength decay

Fig. 7 - Sensitivity to thermal damage of the rebound hammer method according to the tests performed at Aston University [3] and at Politecnico di Milano (test condition A).
Relative rebound index
100% 80% 60% 40% 20% 0% 0 20 40 60 80 expected rebound profile average rebound on surface

ordinary concrete

Relative rebound index

100% 80% 60% 40% 20% 0% 0

lightweight concrete

expected rebound profile average rebound on surface 20 40 60 80

Depth (mm)

Depth (mm)

Fig. 8 - Expected rebound index profiles and average response on the heated surface of the concrete panels exposed to thermal condition B. Given the pronounced surface roughness of the shotcrete surface, a preliminary grinding of the lining surface was required in the Virgolo tunnel. Hence, this method has not been used thoroughly in the case of test condition E. The Cut and Pull Out Test (CAPO test) Being based on a direct assessment of the mechanical response of the concrete cover via a tensile test up to failure, this method proved to be quite sensitive to the thermal damage undergone by the surface of the tested member (test condition A - Fig. 9). In the case of concrete panels exposed to a thermal gradient (thermal condition B) the experimental temperature profiles (Fig. 3) can be transformed into pull-out resistance profiles (after the curves of Fig. 9) and compared with the average pull-out force determined experimentally. Even if the dispersion of nominally identical repeated tests is not negligible, it appears that the CAPO test indicate the response of the most damaged material next to the surface of the tested member (Fig. 10). However this technique proved to be time consuming and its use has been limited to the laboratory tests.

relative pull-out resistance

100%

decay of the pull-out force

Ordinary Lightweight

relative pull-out resistance

100%

sensitivity of the method

80%

80%

Ordinary Lightweight

60%

60%

40%

40%

20%

20%
0%

0% 0 200 400 600 800 Temperature [C]

0%

50% strength decay

100%

Fig. 9 - Standard setup for the pull-out test and sample tested after uniform thermal damage (test condition A); sensitivity of the method to a thermally induced strength decay.

ordinary concrete
pull-out resistance (kN)

lightweight concrete
pull-out resistance (kN)
25 20 15 10 5 0

25 20 15 10 5 0 0 20 40 60 80

20

40

60

80

depth (mm)

depth (mm)

Fig. 10 - Expected profiles of the pull-out resistance and average response of three nominally identical tests on the heated surface of the concrete panels exposed to thermal condition B.

The Ultrasonic Pulse Velocity indirect method (UPV) As it is usually recognized, the velocity of sound in concrete is strongly affected by the thermal damage, thanks to the pronounced temperature sensitivity of the Young's modulus and to the effect of drying of pores (Fig. 11). However, it is quite a difficult task to detect the velocity profile within a member submitted to strong temperature gradients. Useful information on the damage depth and severity can be provided via the indirect UPV method. In this method the measurement of the pulse arrival time is performed by applying both the emitter and the receiver on the same face of the investigated element (Fig. 12). In case the material response improves at increasing depth (that is the rule after a fire), the path of sound waves corresponding to the minimum travel time is the best compromise between the covered distance the material velocity. The actual depth of the material involved in this pulse propagation is governed by the distance x between the probes. A number of repeated measurements of the pulse arrival time at increasing distance allow then to investigate deeper and deeper material layers. The outcome of the test series is a plot on the X-T axes (probe distance - arrival time) whose interpretation has been the object of different numerical methods proposed in the literature [4, 5]. In this research programme, a quick graphical method is proposed for the assessment of both the damage depth and severity on the basis of the shape of the X-T plots. The method is based on a couple of shape parameters (namely the initial slope and the intercept of the final asymptote) that proved to be strongly related to the maximum decay at the surface of the member and to the depth of the sizably damaged concrete. A series of numerical simulations of different thermal transients involving different concrete mixes showed the nice feature of these correlations of being scarcely affected by the temperature conditions and material sensitivity to high temperature. Then, no preliminary information on the member under investigation are needed for the application of this method.
relative UPV
100%

decay of the UP Velocity

Ordinary Lightweight Aston Univ.

UPV decay
100%

sensitivity of the method

80%

Handoo et al.

80%

60%

Politecnico di Milano
(150mm cubes)

60% Politecnico di Milano 40% Ordinary 20% Lightweight Aston Univ. Handoo et al. 0%

40%
(500x100x100mm beams)

20%
(100mm cubes)

0% 0 200 400 600 800

0%

20%

40%

60%

80%

100%

Temperature [C]

strength decay

Fig. 11 - Sensitivity to thermal damage of the Ultrasonic Pulse Velocity according to the tests performed at Politecnico di Milano and to other Authors' results [3].

X E
damaged material pristine material

R too slow too long

R1

R2

R3

R4

Velocity

V20 z

T = time 3 2 1 1

4 1 V20

X = distance

Z80 (mm) V = 80% V20

depth of the damaged layer
V20
representative thickness

Vmin

average min velocity maximum damage

9 8 7 6 5 4 3 2 1 0 0 200 400 600 800 1000 1200 1400

TV20 @ X = 100 mm
Fig. 12 - Minimum time path for the UPV indirect method and typical X-T curve for a layered material; intercept shape parameter and its relationship with the depth of damaged concrete for different concrete mixes; initial secant slope of the X-T curve and its correlation with the maximum damage at the member surface. The reliability and viability of this method have been tested under different damage patterns (testing conditions B, C, D). In the case of the heated panels, the temperature profile has been determined experimentally and the velocity profile can be easily plotted on the basis of the calibration tests (Fig. 13). Therefore, it can be proved that the main features of the velocity profile can be adequately detected via this interpretation method. It is worth to note that the whole thickness of these

relatively thin elements has been more or less affected by the heating process and that the residual UP velocity at the cold face actually corresponds to the final slope of the X-T curves. On the contrary, in the case of larger structural members it is still possible to detect the depth up to which the material has been significantly affected by the exposure to high temperature (Fig. 14 thermal conditions C and D). Summing up, the UPV indirect method provides useful information on the velocity profile of concrete members submitted to fire. However, the time needed for performing the whole series of measurements and the sensitivity to possible cracks within the concrete cover suggest to limit its application the detailed analysis of a few relevant points. Given the pronounced surface roughness of the shotcrete surface and the difficult operational conditions this method has not been used thoroughly in the Virgolo tunnel.

V1 V20
0 ,8

V1 V20
0,8 0,6 0,4

ordinary lightweight

0 ,6 0 ,4 0 ,2 0

Vmin
0 40 80

0,2 0

known temperature profiles

depth (mm)
representative thickness

40

depth (mm)

80

Fig. 13 - Actual UP velocity profiles within the heated concrete panels (thermal condition B) and assessed values: average minimum velocity next to the hot face and maximum velocity at the cold face.

VT V20

1,2 1 0,8 0,6 0,4 0,2 0 0

500
velocity profile at the burners turning off

400

velocity profile after cooling

time T ( s)

4140 m/s
300 intercept 200 100 0 0.0

assessed values

1150 m/s
0.2 0.4

column F3
0.6 0.8 1.0

50

100

150

200

distance X (m)

Fig. 14 - Actual UP velocity profile within the concrete wall submitted to the ISO 834 fire (thermal condition C) and assessed values for the minimum velocity at the hot face and for the damage depth; X-T curve for a column of the precast RC structure (thermal condition D), corresponding to a damage depth of about 60 mm.

Concrete colorimetry The colour of concrete generally changes at increasing temperature from normal to pink or red (300-600C), whitish grey (600-900C) and buff (900-1000C). The pink-red discolouration ensues from the presence of iron compounds in the fine or coarse aggregate, which dehydrate or oxidise in this temperature range. The strength of this colour change depends on the aggregate type and it is more pronounced for siliceous aggregates and less so for calcareous and igneous aggregates [6]. Detecting this first colour alteration is of great interest because its appearance usually coincides with the onset of a significant loss of concrete strength as a result of heating. In this research programme a simplified approach to colorimetry has been formulated, based on the analysis of the side picture of a concrete core, taken via a commonly available low-cost digital camera [7]. The starting point of this method is that digital pictures are usually not very accurate by the colorimetric point of view, but they still allow to recognize the colour variation of a material. As an example, four concrete cores have been taken from each concrete panel pertaining to test condition B and their colour variation profiles clearly reveal up to which depth the material has been significantly affected by high temperature (Fig. 15). The only limitation of this method is that a core have to be cut from the member (which has not been possible in the Virgolo tunnel after the fire tests).
1000 100%

C
800

R T / R 20 c c

LWC
80% 60% 40%

T
600 400 200 0 0

65%

ordinary concrete
470C

LWC depth (mm)

20 40 60

20% 0% 80

0.35

0.35
NCS atlas (c = 5 - w = 5) colorimeter camera
(flash - auto wb)

y
0.34

y
0.34 1

ordinary concrete
masked aggre gate 10x10 pi xe l / point

full image masked aggregate std dev ellipse (masked aggr.) Mont Blanc tunnel [4]

0.008 0.006 0.004

color variation (x - y) ordinary concrete

(masked aggregate)

average breakpoint

1 600 400

800C

flash - auto wb

0.33

D65 masked
standard deviation ellipses

D65
0.33 20

ordinary concrete
flash illuminant auto white balance

0.002 0.000 -0.002 0 20 40

full 0.32 0.30 0.31

x
0.33
0.31

200 0.32 0.33

depth (mm)
60 80

0.32

Fig. 15 - Column on the left: visible colour change of a heated ordinary siliceous concrete, statistical scattering of the CIE 1931 colour measure due to the inherent material heterogeneity and average colour change after heating (test condition A). Box on the right: connections among the maximum temperature and the residual strength profiles and the colour variation profiles for the concrete panels subjected to a thermal gradient (heating condition B).

0.006

color variation (x - y) lightweight concrete

(masked aggregate)

0.004

0.002

0.000

-0.002 0 20 40

depth (mm)
60 80

4. The drilling resistance test

The measurement of the drilling resistance appears to be a promising and fast technique, which allows to continuously scan the material response at increasing depth. Several examples of this kind of approach to the assessment of construction materials are available in the literature. A first application [8] was based on the measurement of the thrust to be exerted on the drill to drive the bit at a constant rate in the tested material (bit = 4-8 mm, max hole depth = 1520 mm for concrete and mortar). Recently, this method has been proposed also as a means to validate the performance of the surface treatments on stone materials [9] and it is in the process of being standardized by CEN TC 246 - Natural Stones. An alternative indicator of the material response is provided by the resistant torque at constant turning and feed rates, which is currently adopted for ascertaining the preservation of wooden structures [10]. It is worth to note that the resistant torque actually corresponds to the power spent to drive the bit, being the drill turning rate almost constant. In principle, the nice advantage of focusing on the drilling work is that a change on the exerted thrust concurrently affects the power consumption (J/s) and the advancing rate (mm/s). Then, their ratio, namely the specific work that has to be spent to drill a unit deep hole (J/mm), is expected to be marginally affected by the thrust. This assumption was confirmed in a broad series of tests on the mortar layers of different brick masonry walls (bit = 46 mm, max hole depth = 510 mm [11]). A relationship between the drilling work and the material fracture properties was also proposed in the cited study. As a matter of fact, releasing the test method from an accurate control of either the thrust or the bit feed rate allows to considerably simplify the experimental apparatus. Concerning the application to fire damaged concrete structures, the thickness to be inspected usually extends to several centimetres and a hammer drill is generally recommended to prevent an excessive bit wearing and overheating. In this case, the sensitivity to the exerted force is partly masked by the hammering action [8] and the dissipated work further appears to be the most promising indicator of the material soundness. Once a constant drill bit performance is guaranteed via the hammering action, the most interesting feature of the drilling technique is that the deep virgin material is inspected in the final stage of the drilling process. Hence, a reference drilling resistance is available for each test and no special calibration curves should be needed for the evaluation of the thickness of damaged concrete. Experimental setup The drilling resistance has been measured by modifying a Hilti TE 6-A battery hammer drill in order to monitor the electrical power consumption, the bit rotation and the hole depth (Fig. 16). Thanks to the significant motor power (350W) and the effective electro-pneumatic hammering action (impact energy = 1.5 J) this tool allows to drill small diameter holes in good quality concrete at quite a fast rate (about 5-10 mm/s for = 6-10 mm). After proper transformation and analog filtering, the electrical signals are acquired by a PCMCIA A/D card (National Instruments - DAQ Card 6036E) and processed by a dedicated software, in order to work out different test parameters such as the motor rate and acceleration, the instantaneous total power consumption and the net drilling work per unit depth (J/mm - regarded as the drilling resistance hereafter - Fig. 17).

measured parameters chuck rotation (rad) - photodiode current I (A) - Hall effect transd. DC tension V (V) - Hall effect transd. hole depth d (mm) - potentiometer worked out parameters chuck rotation rate = d / dt total electric power Ptot = V I idle resistant torque Ti = A + B + C d/dt idle power Pi = Ti net drilling work Wnet = (Ptot - Pi) dt drilling resistance DR = Wnet / d Fig. 16 - The battery hammer drill fitted with the electronic circuits and the displacement transducer; list of the parameters directly measured and worked out during the test.
400 W J/mm 300 0.8 s/mm 0.6

Po wer (W) 400 300 200 100 0 0

drill start-up ( 2 mm)

pristine concrete
R cm 60 N/mm2
20

cutting edge sinking ( 3 mm)

total net

bit = 10 mm

net drilling power (W)

30%

200
Thrust = 170 N

drilling time (s/mm)

0.4

idle
Depth (mm)
20 40 60

100
drilling resistance (J/mm)

0.2

0 0

Depth (mm)
40 60

20

Fig. 17 - Total and net drilling power and definition of the drilling resistance (J/mm) as the product of the net drilling power (W = J/s) and the drilling time (s/mm). Test procedure Several preliminary tests on both a virgin and a thermally-damaged good quality concrete (Rcm 60 N/mm2 - max aggregate size = 16 mm - T = 20C and 600C) have been performed, aimed to define a simple test procedure able to guarantee repeatable results. In the final arrangement, the bit is firstly pointed against the sample to be tested and the drill is pushed to preload the hammering mechanism. Then the drill is activated at the maximum power, in order to ensure a constant performance during the whole process. In this way, only the first 2-3 mm are expected to be somehow influenced by the drill start-up and by the initial sinking of the bit cutting edge. Concerning the thrust to exert on the drill, all the tests have been performed downwards in the vertical direction, putting different weights on top of the drill. It has been found that the maximum total thrust should not exceed the value of 200 N, so as to limit the bit wearing and overheating. On the opposite side, a thrust of at least 50 N is needed to guarantee the effectiveness of the electropneumatic hammering action. Within this range, the thrust doesnt significantly affect the drilling resistance of pristine concrete, whereas the sensitivity to thermal damage improves approaching the upper limit (Fig. 18). The same conclusion is valid for the drilling time (i.e. the inverse of the feed rate), confirming the regularizing effect of the hammering action. Then, all the succeeding laboratory tests have been performed by adding a dead weight of 100 N to the self weight of the drill (about 70 N). In any case, the recommendation of keeping the thrust nearly constant during the test should guarantee consistent results for in-situ applications.

As for the bit diameter (Hilti TE-CX - = 614 mm), it has been found that a small bit (6 mm) exhibits a higher sensitivity to the material inherent heterogeneity (Fig. 19), whereas a large bit (14 mm) is prone to overheating problems and is too demanding for the drill motor. Moreover, looking at the torque at the chuck, it has been observed that the most regular response at increasing hole depth is obtained in the case of a 10 mm bit, which has been adopted for all the succeeding tests.
6.0

Drilling Resistance (J/mm)

Thrust

80

Net drilling work (kJ)

4.0

Rcm 60 N/mm 2
20

60

bit = 10 mm 1 mm DR (J/mm)

pristine concrete

Drilling Time (s/mm)

70 N 170 N

pristine concrete

0.3

pristine concrete

40

heated up to T= 600C

0.2

heated up to T= 600C

2.0

heated up to T= 600C

20

0.1 Rcm 60 N/mm 2

20

Rcm 60 N/mm 2
20

0.0 0 20 40

Depth (mm)
60 80

bit = 10 mm 0 50 100

Thrust (N)
0
150 200

bit = 10 mm 50 100

Thrust (N)
150 200

Fig. 18 - Effect of the exerted thrust on the drilling resistance (J/mm) and time (s/mm) of both a pristine and a thermally damaged concrete.
Torque deviation - RMS (Nm)

6.0

pristine concrete 20 Rcm 60 N/mm 2 Thrust = 170 N

1.0 pristine concrete 0.8 0.6 max 0.4 0.2 0.0 6 8 min Bit diameter (mm) 10 12 14 Rcm 60 N/mm 2
20

Torque (Nm)

bit = 6 mm

4.0

thrust = 170 N

2.0

bit = 10 mm linear fit Depth (mm) 60 80

0.0 0 20 40

Fig. 19 - Variation of the net torque at the chuck due to the inherent material heterogeneity and torque deviation from linearity with different drill bit diameters.

DR (J/mm)
40 30 20 10 0 0 200 400 ordinary lightweight

thrust = 170 N bit = 10 mm

DT (s/mm)
0.20 0.15 0.10 0.05 ordinary lightweight 0.00 0 200 400

thrust = 170 N bit = 10 mm

T(C)
600 800

T(C)
600 800

Fig. 20 - Effect of the thermal damage on the drilling resistance DR and drilling time DT (thermal condition A).

DRT / DR20
550C 100% 400C ordinary 50% lightweight Drilling Resistance Drilling Time 0% 0 200 400 600 decay onset

Fig. 21 - Relative decay of the drilling parameters and critical temperatures corresponding to the onset of a recognizable material damage.

T(C)
800

Sensitivity to thermal damage Once the optimum bit diameter and drilling thrust have been defined, a series of tests on concrete cubes (150 mm) has been performed in order to ascertain the sensitivity of this method to different levels of thermal damage (test condition A). Despite the compressive strength (Fig. 2), the drilling resistance is an almost constant or even increasing function of temperature up to about 400C (Fig. 20), possibly because of the increased material deformability and nearly constant fracture energy, which are generally associated to the material thermal damage [12]. Nevertheless, a marked drilling resistance decrease takes place at higher temperatures, as soon as the severe strength decay offsets the improved material ductility (RcT < 0.5 0.7 Rc20C). Due to the softer aggregate, the lightweight concrete is definitely easier to be drilled, but the temperature effect is still recognizable in relative terms (Fig. 21). The same trends can be observed for the drilling time, although this parameter proved to be less sensitive to the thermal damage, especially in the case of lightweight concrete. For this reason, only the drilling resistance will be considered in the following sections. Other tests, not reported in this paper, proved the viability of this technique also in detecting voids, defects and layers of distinct materials (plaster, insulation, etc). Tests on concrete panels In order to ascertain the reliability of the proposed test method in assessing the damage gradient within a concrete member exposed to fire, the concrete panels submitted to a constant temperature gradient have been investigated (test condition B). The drilling tests clearly reveal the effect of the thermal gradient (Fig. 22), albeit the result is partially masked by the inherent material heterogeneity ascribable to the aggregate. However, this disturbance can be easily cleaned out by averaging the results of a few repeated tests. Moreover, the plots of the relative drilling resistance (i.e. referred to the innermost material response) further help in recognizing the external damaged layer, regardless of the initial value of the drilling parameters at room temperature (Fig. 23). The depth of the isotherm corresponding to the onset of the drilling resistance decay can be easily detected from these latter diagrams.

DR 120 (J/mm)

ordinary concrete

DRT / DR 20 100% 80% 60%

ordinary concrete

80

LWC

40 average 0 0 20 40 depth (mm) 60

40% 20% 0% 0 20 40 depth (mm) 60

DR 40 (J/mm) 30 20

lightweight concrete

800

T (C)
600
550C

ordinary concrete 400

400C

10 0 0 20 40 depth (mm) 60
0 0 20 40 200

LWC

depth (mm) 60

Fig. 22 - Drilling resistance profiles through the ordinary and lightweight concrete panels after heating the left side (thermal condition B). Fig. 23 - Average profiles of the relative drilling resistance and connections to the temperature and residual strength profiles.

100% LWC 80% 60%

53% 74%

ordinary concrete

40% 20% 0% 0 20 40 depth (mm) 60

Tests on a concrete wall submitted to a ISO 834 fire As discussed before, the standard fire test on a concrete duct for electric cabling protection provided a useful benchmark for the possible inhomogeneous damage ensuing from a real fire (Fig. 24 - test condition C). After cooling, the back wall of the test setup has been examined by drilling a series of holes along 5 rows (from A to E - 3 holes for each row). The average diagrams pertaining to each row clearly reveal which part of the structure went through a severe thermal exposure (rows from A to C) and which one was only marginally impaired during the fire test. It is worth to note that only about 5 minutes were needed to perform the whole series of tests and the results were immediately available for the interpretation thereafter. This is definitely the main benefit of this kind of NDT technique.
DR
30 (J/mm)

20 A B C D E

10

0 0 20 40

depth (mm)
60

Fig. 24 - Fire test setup including the concrete duct to be tested and the back wall which was subsequently examined via the drilling resistance technique (thermal condition C); average drilling resistance profiles recorded at five levels different on the back wall.

Tests on a precast RC building after a real fire The depth of damaged concrete of a concrete column surviving a real fire has been ascertained via the drilling resistance method (test condition D - see also Figs. 5 and 14). A series of three holes has been performed on the two most severely damaged sides of the column, confirming the first indications provided by the rebound index along the cross section perimeter. The damaged depth detected by this method (about 40mm) is comparable to the like indication provided by the UPV indirect method (about 60mm from Fig. 14 and 12), the latter being sligtly more sensitive to the damage onset.

39 19

45 49

F3
15 16
50 40 30 20 10 0 0 20 40 60

42 24
50

drilling resistance (J/mm)

drilling resistance (J/mm)

average

40 30 20 10 0 0 20 40 60 average

concrete column F3
(0.45 x 0.45 m)

concrete column F3
(0.45 x 0.45 m)

depth (mm)

depth (mm)

Fig. 25 - Drilling resistance profiles recorded on the two most severely damaged sides of the concrete column after real fire (thermal condition D).

Tests on the lining samples of the Virgolo tunnel This latter verification case (test condition E) concerns a wide range of cementitious materials that have been tested either as a lining repair layer on the tunnel wall (6 shotcrete samples - letters from A to F) or as small panels arranged in the mini tunnel structure (4 panels - numbers from 1 to 4). No information are presently available on these mix designs and only some results of ND testing before the fire tests can be provided to give an idea of the different mechanical properties of these composites. Due to practical difficulties, the drilling resistance test was the only technique that was possible to apply after the fire tests. In a very short time ( 1 hour) about one hundred drilling tests were performed on the different concrete samples, with no need for sample preparation nor special analysis of the results. Moreover, the results can be summarized in a few damage parameter (namely the damage depth and the minimum drilling resistance at the surface), by simply comparing the whole profile to the final, almost constant drilling resistance. It is then possible to compare the performance of the different mixes after fire exposure.

panel # 1 2 3 4

base thickness thickness (mm) (mm) 45 50 50 100 40 50 40 50

UPV (m/s) 2840 4530 1580 1620

lining Rebound standard sample index deviation A 35 3.7 B 19 5.0 C 20 1.8 D 31 3.6 E 33 2.7 F 13 0.83

Tables 3, 4 - NDT response of the 4 + 6 concrete mixes before the fire testing.

20

60

15
reference (before fire)

drilling resistance (J/mm)

drilling resistance (J/mm)

mini tunnel panel #1

mini tunnel panel #2

40

10

top middle bottom 0 10 20 30

base panel

20 top middle bottom 0 0 20 10 20 30 40 50

0 40 50

depth (mm)
20

depth (mm) drilling resistance (J/mm) mini tunnel panel #4

15

drilling resistance (J/mm)

mini tunnel panel #3

15

10

10

top middle bottom 0 10 20 30 40 50

top middle bottom 0 10 20 30 40 50

depth (mm)

depth (mm)

Fig. 26 - Profiles of the average drilling resistance measured at different heights on the panels after the mini tunnel fire test (thermal condition E - each panel comprises a base concrete layer and a 4050 mm coating, that is the object of the fire test).

60
mini tunnel panels (middle height)

20

mini tunnel panels

(middle height)

drilling resistance (J/mm)

50 40 30 20 10 0 0

drilling resistance (J/mm)

15

1 3 4

10

0 10 20 30 40 50 0 10 20 30 40 50

depth (mm)

depth (mm)

Fig. 27 - Comparison among the drilling resistance average profiles measured at middle height within the concrete panels after the mini tunnel fire test.
50

mini tunnel panels (middle height)

mini tunnel panels (middle height)

drilling resistance (J/mm)

damage depth (mm)

30

DR
ref

40 30 20 10
43% 31%

reference
(before fire)

20

DR at the surface damage depth

10

at the surface
(after fire)

0 1 2 3 4

0 1 2

51%

53%

panel #

panel #

Fig. 28 - Damage depth and minimum residual drilling resistance at the panel surface after the mini tunnel fire test.

40

40

drilling resistance (J/mm)

drilling resistance (J/mm)

before fire
(h = 1.2 m)

before fire
(h = 1.2 m)

30

reference
after fire
(h = 3.0 m)

30 after fire 20
(h= 3.0 m)

20

10

10 tunnel lining B 0 10 20 30 40 50

h
0 0 10 20 tunnel lining A 30 40 50

depth (mm)
20 40

depth (mm)
before fire
(h = 1.2 m)

drilling resistance (J/mm)

(h = 1.2 m)

15 after fire
(h = 3.0 m)

drilling resistance (J/mm)

before fire

30 after fire 20
(h = 3.0 m)

10

5 tunnel lining C 0 10 20 30 40 50

10 tunnel lining D 0 10 20 30 40 50

depth (mm)
30 20 before fire
(h = 1.2 m)

depth (mm)

drilling resistance (J/mm)

drilling resistance (J/mm)

15 after fire 10
(h = 3.0 m)

20 after fire
(h = 3.0 m)

10

before fire
(h = 1.2 m)

0 0 10 20

tunnel lining E 30 40 50

0 0 10 20

tunnel lining F 30 40 50

depth (mm)

depth (mm)

Fig. 29 - Profiles of the average drilling resistance in the tunnel lining samples before and after the fire test (at h = 1.2m and h = 3.0m respectively).

20

40

drilling resistance (J/mm)

tunnel lining (h = 3.0 m)

tunnel lining
(h = 3.0 m)

reference at the surface

damage depth (mm)

C A
10

D E

30

D E

20
55%

C F
31% 21% 53% 37%

10
24%

lining type

panel #

Fig. 30 - Damage depth and minimum residual drilling resistance at the surface in the tunnel lining samples after the fire test (at h = 3.0m).

Fig. 31 - View of the four concrete panels after the mini tunnel fire test.

5. Conclusions
The drilling resistance test proved to be a viable and quick method for the assessment of the thermal damage undergone during a broad range of thermal loads. The tests performed within this research project, allow to formulate the following set of conclusions: A hammer drill is recommended in order to quickly inspect the concrete cover preventing an excessive bit wearing and overheating. In this case, the sensitivity to the exerted thrust is markedly reduced and no special control of either the drilling force or the feed rate is needed. A medium size drill bit ( 10mm) exhibits a regular response despite of the inherent material heterogeneity (max aggr. size 16 mm). The dissipated work per unit drilling depth (J/mm) appears to be the most sensitive indicator of the material soundness. A correlation between this parameter and the material compressive strength cannot be easily worked out, given the strong influence of other properties like the fracture energy and the aggregate hardness. However, the drilling resistance keeps its significance in relative terms and the comparison with the inner virgin material provides meaningful information on the thickness of the outer fire damaged concrete. Even if only a sizeable thermal damage can be detected via the drilling resistance method (RcT < 0.5 0.7 Rc20C), it should be noted that similar damage levels are considered in the popular Reduced cross-section method for the design of concrete structures under thermal loads and for the evaluation of the residual capacity after a fire (critical temperature = 500C). The drilling resistance test proved to be a fast and reliable method also in the case of in-situ application and realistic fire conditions. The immediate availability of the results is expected to be of valuable help in the assessment of concrete structures surviving complicated fire scenarios. In case of difficult operational conditions (which are typical of tunnels), this handy tool turned out to be the only viable technique in terms of time needs and implementation efforts. Being based on the mechanical response of the material under investigation, this technique is likely to provide useful results also in the case of different lining materials (stones, brickwork, etc), regardless of their more or less pronounced sensitivity to high temperature.

Acknowledgements
A grateful acknowledgement goes to all the students who lively cooperated in carrying out the tests in partial fulfilment of their MS degree requirements: Massimiliano Bondesan and Gianluca Pizzigoni (drilling tests), Gian Andrea Basilico and Davide Cabrini (concrete colorimetry), Angela Faccoli and Luigi Marzorati (indirect UPV interpretation). A particular acknowledgement to Matteo Colombo for sharing the difficulty of running the ND tests in the Virgolo tunnel. Special thanks also to Daniele Bonetti (D. Bonetti & Co. - Dalmine - Italy) for the valuable suggestions concerning the signal conditioning and data acquisition of the modified drill. The author is also indebted to MS Engr. Paolo Mele and Giuseppe Grella of CSI (Italian Experimental Center - Bollate, Italy) for their support to the experiments following the fire test on the concrete wall. Last but not least, the author wishes to acknowledge Hilti Corp. (Schaan - Principality of Liechtenstein) for making available the drill and its accessories.

References
[1] FELICETTI R. and GAMBAROVA P.G., "High-Performance Light-Weight Concrete: Material and Sectional Properties during and after a Fire", Proc. Int. Conf. on Advances in Concrete Structures, Xuzhou (China), V.1, 2003, pp. 89-99. CIB W14 Report, "Repairability of Fire Damaged Structures", DRYSDALE D.D. and U. SCHNEIDER U. (Editors), Fire Safety Journal, V.16, 1990, pp. 251-336. SHORT N.R., PURKISS J.A. and GUISE S.E., "Assessment of fire-damaged concrete using crack density measurements", Structural Concrete, V. 3, 2002. BENEDETTI A., "On the ultrasonic pulse propagation into fire damaged concrete", ACI Structural Journal, May-June, 1998, pp.257-271. ABRAHAM O. and DROBERT X., "Non-destructive testing of fired tunnel walls: the Mont-Blanc Tunnel case study", NDT&E International, n.36, 2003, pp. 411-418. SHORT N.R., PURKISS J.A. and GUISE S.E., "Assessment of fire damaged concrete using colour image analysis", Construction and Building Materials, n. 15, 2001, p. 9-15. FELICETTI R. , "Digital camera colorimetry for the assessment of fire damaged concrete", Proc. Fib Task Group 4.3 Workshop Fire Design of Concrete Structures: What now? What next? - Milan, Dec. 2-4, 2004, p.211-220. CHAGNEAU F. and LEVASSEUR M., "Contrle des matriaux de construction par dynamostratigraphie", Materials and Structures, V.22, 1989, pp. 231-236. TIANO P. and VIGGIANO A., "A new diagnostic tool for the evaluation of the hardness of natural and artificial stones", Int. Journal for Restoration of Buildings and Monuments, V.6, n.5, pp. 555-566.

[2] [3] [4] [5] [6] [7]

[8] [9]

[10] RINN F., "Eine neue Bohrmethode zur Holzuntersuchung", Holz-Zentralblatt, V. 34, n.115, 1989, pp. 529-530. [11] GUCCI N. and BARSOTTI R., "A non-destructive technique for the determination of mortar load capacity in situ", Materials and Structures, V.28, 1995, pp. 276-283. [12] FELICETTI R. and GAMBAROVA P.G,. "On the Residual Properties of High Performance Siliceous Concrete Exposed to High-Temperature", A Volume in Honour of Prof. Z.P. Bazants 60th Birthday, G. Pijaudier-Cabot, Z. Bittnar and Bruno Gerard, Hermes Science Publ., Paris,1999, pp.167-186.