Bundesanstalt für Materialforschung und -prüfung

International Symposium (NDT-CE 2003)

Non-Destructive Testing in Civil Engineering 2003
Start > Contributions >Lectures > Pavement: Print


Cesare Sangiorgi , PhD student, University of Bologna, ITA, Facoltà di Ingegneria, DISTART, Strade e Geotecnica, V.le Risorgimento 2, Bologna, 40136, Italy, cesare.sangiorgi@mail.ing.unibo.it
Andrew C.Collop, Reader in Civil Engineering, Nottingham Centre for Pavement Engineering, University of Nottingham, UK, Andrew.Collop@nottingham.ac.uk
Nicholas H.Thom, Teaching & Research Fellow, Nottingham Centre for Pavement Engineering, University of Nottingham, UK, Nicholas.Thom@nottingham.ac.uk


A road pavement is commonly built as a multi-layered structure with a "bond" coat between layers to achieve the adhesion needed for fully bonded behaviour under load. In all pavement design methods, the assumption is made that full bond is provided between asphalt pavement layers. Theoretically, it has been shown that this bond is necessary and that, if not present, increased strains result, leading to shorter life. In practice, poor bond, often due to poor construction practice, has been identified as a major cause of premature failure on numerous occasions. Hence, detection of poor bond is a priority research task. Knowing the actual bond condition of an apparently bonded pavement can be useful either for construction quality assessment, or for maintenance design. This paper describes a method for the in situ determination of bond beneath a surface course using a Non-Destructive Impulse Hammer approach. It involves the measurement of the vertical acceleration of a pavement surface adjacent to a point where an impulse is applied. Results will be presented relating to various different bond conditions under different surface course applications. The accelerometer time histories correlate well with these differences when a repeatable impulse is applied. The usual Frequency-based approach is replaced by a data interpretation approach that utilises Fractal Theory.


Bond problems account for a proportion of known premature highway pavement failures [1,2] but, though theoretical studies can be made [3], the full extent of 'loss of performance' due to lack of bond is not known. In France, in 1986, it was estimated that 5% of the network suffered from bonding related problems [4]. Lepert et al. [5] noted that weak bonding between bituminous pavement layers may develop during construction due to: poor quality control, lack of tack coat, contamination of the lower layer or laying in cold weather. Current practice in the UK is to construct layered asphalt pavements with a bituminous 'bond coat' (or tack coat) between the layers, the (unstated) aim being to achieve a full bond, that is zero slip, between layers. Where asphalt overlies concrete a more substantial coat is applied. The aim is the same, but it is recognised that the task of achieving bond is more difficult. However, the actual consequences of poor bond and the development of bond with time and traffic are not well understood. Bond strength is hard to measure (especially with Non Destructive tests) and, inevitably, most of the existing evidence for its importance is qualitative. In addition, bond strength/stiffness measurement is not routinely carried out in the UK and no accepted technique has yet been developed. This lack of understanding inevitably gives rise to difficulties during construction since neither the contractor nor the engineer will have a clear idea of the real importance (or otherwise) of the correct application of bond coat, nor of the effect of dirt, dampness, etc.
Figure 1 shows the result of a typical debonding failure.

Fig 1: debonding on a road surfacing.

The situation has been brought more fully into focus as the technology of thin surfacings has become evermore widespread. Because these materials are laid in a thin layer (down to 15mm) there can be high shear stresses transmitted to the bond between the thin surfacing material and the layer below. Also, in the prevailing contractual climate, contractors are increasingly expected to take responsibility for the performance of the road, often under a Design, Build, Finance and Operate (DBFO) contract. This has led to rapid developments in bituminous material design, as responsibility has brought with it greater freedom of choice. However, the bond coat, though a small detail in terms of overall pavement cost, is an important component of the whole structure, for which there is no readily available test and little knowledge of the implications of any shortcomings in terms of pavement life.

It is clear that slippage and delamination of surfacings can be a serious problem [6]. Hence, the early detection of poor bond between the surfacing and the binder course is a priority research task. Previous has used the Falling Weight Deflectometer (FWD) to non-destructively assess bond condition [7,8,9]. In this approach a back-analysis method is used determine the elastic in-situ shear bond stiffness at the interface.

This paper is focussed on in-situ assessment of bond condition using a non-destructive hammer test. The research forms part of a larger programme at Nottingham to assess the importance of bond between pavement layers [10,11].


2.1 Hammer test device
Fig 2: Hammer test set-up.
The non-destructive technique under development at Nottingham to assess in-situ bond condition involves the use of an impulse hammer. The basic principle is to apply an impulsive loading to the pavement surface with the hammer and measure the vertical dynamic response using an accelerometer. The hypothesis is that structural distress in the form of loss of adhesion between pavement layers is reflected in the dynamic response of the pavement structure. Figure 2 shows the equipment required to undertake the test. The main piece is the impulse hammer that contains a calibrated dynamic load cell (Kistler piezoelectric ±120kN) fixed between two pre-tensioned bolts. A plastic interchangeable tip is screwed to the load cell by means of a thread and transfers to it the given impulse load. The whole tip is fixed to a steel handle bar. The pavement dynamic response is recorded using an accelerometer (Kistler piezoelectric ±50g) coupled to the tested surface with either blue tack or with a glued steel ring ( depending on the surface roughness). Both the applied force and the resulting surface acceleration signals are captured by a data acquisition system operating on a portable laptop computer. The pavement temperature at the time of testing should be taken and two operators are required.

2.2 Trial pavement
In order to develop the Hammer test equipment and data analysis procedures, a trial site, located in a quarry in Leicestershire, was used. The objective of the trials was to investigate the hammer test and data analysis procedures under realistic (but closely controlled) conditions. The existing pavement was more than 60 meters long and 3 meters wide. The idea was to overlay a new wearing course on the existing material (HDM) using different bond interface conditions. Three different surface treatments were used to achieve different levels of bond at the interface:

  • A normal amount of tack coat (K1-40),
  • A swept surface with no tack coat,
  • A clay slurry applied to the receiving layer.

The pavement was divided into two sections. One section was overlaid with a proprietary surfacing known as Thinpave (20mm in thickness) and the other section was overlaid with a standard 10mm Close Graded Macadam (CGM, 30mm in thickness). Each section was divided into three sub-sections as shown schematically in Figure 3.

Fig 3: Trial Road layout.

Figure 4 shows the main construction operations taking place. The first two photographs show the interface being prepared and the last two photographs show the surfacings being laid.

Fig 4: Trial Road construction operations.

2.3 Hammer test procedure and data acquisition
Previous experience has shown that it is necessary to apply three repeated hammer blows at each of three equally spaced locations around the accelerometer to characterise the dynamic response of the pavement structure at a particular test location. Typically, the magnitude of the applied load is between 15 and 18kN. Previous studies have shown that the optimum location of the accelerometer is between 9 and 10cm from the applied load. Data acquisition was undertaken using SignalCalc software. In a previous study, a Frequency (spectral) based data analysis approach was used (see [11] for further details).

The first phase of hammer testing was undertaken approximately one month after the surfacing was laid. The air temperature at the time of testing was 4°C and the pavement surface temperature was between 7 and 8°C. In total, 18 test locations where identified on the site, 3 for each interface area. After Hammer testing, 18 no. 150 mm cores were taken from nominally the same locations that were tested to investigate the actual bond conditions in the laboratory for correlation with results from the hammer testing. All the cores from the areas where the clay slurry had been applied split into two parts indicating that, if there had been some bond, it was destroyed by the coring operation.

A second phase of Hammer tests was carried out 6 month later when the air temperature had risen to approximately 18°C and the pavement first interface temperature was 23°C. The locations tested were more than 20 cm away from the old core holes. No new cores were taken for laboratory testing.


Figure 5 shows typical accelerometer time histories taken from bonded and de-bonded areas. It can clearly be seen from this figure that, qualitatively, there is a significant difference between the signals.

Fig 5: Typical Debonded (left) and Bonded (right) time histories.

It can be seen from the response of the bonded system that the acceleration is heavily damped and decays to zero after approximately 1.5ms. Conversely, the response of the de-bonded system shows significant high frequency vibrations (>1kHz) in the measured acceleration response and the acceleration does not decay to zero until approximately 6ms. This is due to vibrations that are induced in the surfacing layer because it is not bonded to the layer below. A large number of laboratory and in-situ test have been undertaken and this pattern is repeatably observed from the results. A number of different techniques are available for quantifying the differences between these acceleration curves. In this paper the Theory of Fractals is used. Other approaches have also been investigated as part of this research and are detailed elsewhere [11].

Fig 6: Box Counting Fractal Dimension for the Koch fractal.

The use of Fractals Theory in scientific problems is today very common. There is a wide range of applications that involves fractals and nature is full of quasi-fractal objects (e.g. coastlines). A fractal is an object with an infinite nesting of structure at all scales with its own unique Fractal Dimension (F.D.). It is possible to apply the fractal theory to "non-mathematically" fractal objects such as a recorded acceleration time history and calculate a Fractal Dimension: a single number between 1 and 2. An elementary technique for determining the F.D. is known as "Box Counting" (see Figure 6). The following procedure (coded in Fortran90) was adopted to determine the F.D. for each measured acceleration time history by steps. An initial square box dimension (Linit) is chosen and the number (N) of boxes required to fully cover the complete signal is determined. In the following step (i) the box dimension is reduced (L(i)) and the number of new (smaller) boxes required to fully cover the complete signal is re-calculated. The procedure is repeated for proportionately smaller box dimensions in the next steps. The number of boxes (N) is then plotted against (s), the ratio between the initial box dimension (Linit) and the box dimension at each (i) step (L(i)) on double logarithmic scales. A straight line is then fitted to the data and the gradient of the line is the F.D.. Figures 6 and 7 show how the Fractal Dimension is deducted from the log(N)-log(s) graph for a classical Koch fractal curve and for a measured accelerometer time history recorded from a well bonded section of the trial pavement.

Fig 7: Box Counting Fractal Dimension for a Bonded location.

Since the input force time histories for consecutive blows are similar, but not identical, the acceleration time histories have been normalised by the area under the input force-time curve (i.e. impulse magnitude). The initial box side Linit was chosen to be 4 units for each application to fully encompass the measured time histories. Other initial box dimensions were tried and it was found that the calculated F.D. values were not particularly sensitive to the initial box dimension Linit.

Fig 8: Fractal Dimension values for the December hammer testing.

For each of the 36 measurement points (from the two testing phases) an average F.D. was calculated from the 9 blows around each point. Figure 8 shows the F.D.s calculated using data from the first phase of hammer testing. It can clearly be seen that the areas where a good bond would be expected, display a relatively low F.D. and the areas where a poor (or no) bond would be expected display a relatively high F.D. The minimum F.D. calculated for an area that was classified as "debonded" is 1.19. (An area was classified as debonded if samples taken for Leutner testing fell apart during coring.) Consequently, it can be assumed that higher values are representative of debonding whereas lower values (e.g. locations 4 to 15) are indicative of some degree of bond. It is difficult to identify a unique F.D. value that delineates areas of bond from areas of no-bond.

Average values for the F.D. for each area are given in Table 1 (a values in brackets indicates the percentage difference compared to the interface with tack coat for the same surfacing). It can be seen from these data that, for both surfacings, the average FD is lowest for the interfaces with a normal amount of tack coat and highest for the interfaces with the clay slurry. It is interesting to note that the difference between the bond conditions is greater for the Thinpave surfacing compared to the CGM surfacing.

  Thinpave CGM
with slurry swept tack coated tack coated swept with slurry
Fractal Dimension1.261(11.91%)1.116
1.126(0.0%)1.121(0.0%) 1.152(2.76%)1.201(7.13%)
Table 1: Average Fractal Dimension Values.

Figure 9 shows the correlation between the F.D. and the Leutner test values: there is a reasonable correlation between the two sets of results, although the scatter is quite high (no Leutner test was performed for already split specimens from locations n° 1 to 3 and 16 to 18). It should be noted that the Leutner test results are quite low indicating that, even in the best areas, the bond is poorer than has previously been obtained from laboratory testing [10]. This may have been due to the fact that an existing surface that was quite smooth had been used as the receiving layer minimising mechanical interlock with the new surfacing materials.

Fig 9: Correlation between Hammer test F.D. results and Leutner results.

Figure 10 presents a comparison between the F.D.s from the two phases of testing (the bars at the top of the graph indicate the percentage difference in F.D. between the two visits).

Fig 10: Comparison between F.D.s from the two testing phases.

It can be seen from this figure that the general trend is for the F.D. to decrease indicating that the bond has improved. It is interesting to note that the largest increase in bond is where the bond was initially poorest. As there was no trafficking this increase in bond is likely to be due to climatic conditions between the two testing phases (6 months had elapsed between the two sets of tests). The average pavement surface temperature at the time of the first set of tests was 8°C compared to 23°C at the time of the second set of tests. It is known that bitumen displays healing characteristics, particularly at elevated temperatures [12] which could help to explain the observed increase in bond.


  • This paper describes a method for the in-situ determination of bond beneath a surface using an impulse hammer approach.
  • The vertical acceleration of a pavement surface is measured adjacent to a point where an impulse is applied.
  • Six different combination of bond condition and wearing course material have been constructed in a full-scale field trial.
  • Differences between acceleration signals from bonded and debonded areas are qualitatively distinguishable.
  • A Fractal Dimension (F.D.), calculated from measured acceleration time history, has been used as a quantitative indicator of bond conditions.
  • Areas of good bond display a low F.D. value whereas areas of poor bond display a higher F.D..
  • The calculated F.D.s range from approximately 1.1 (well bonded area) to approximately 1.3 (de-bonded area).
  • An overall increase in average bond condition was measured between the two phases of testing.
  • There is a reasonable correlation between Leutner shear load and the calculated F.D.s.


  1. Kennedy C.K. and N.W. Lister, Experimental studies of slippage, The Performance of Rolled Asphalt Road Surfacing, ICE, London, pp 31-56, 1980.
  2. Peattie K.R., The incidence and investigation of slippage failures, The Performance of Rolled Asphalt Road Surfacing, ICE, London, pp 3-15, 1980.
  3. Brown S.F. and J.M. Brunton, The influence of bonding between bituminous layers, Highways and Transportation, vol. 31(5), 16-17, 1984.
  4. SETRA / DTC, Le décollement des couches de revêtement de chaussées, Information Note n° 25, SETRA, 1986.
  5. Lepert P., Poilane J.P. and Villard-Bats M., Evaluation of various field Measurement Techniques for the assessment of Pavement Interface Condition, Proceedings of the 7th International Conference on Asphalt Pavements, vol.3, 224-237, 1992.
  6. Pös J., M. Kruntcheva, A.C. Collop and N.H. Thom, Verbund von Asphaltschichten: Prüfung, Einflüsse, Bewertung - Vergleichende Betrachtungen Deutschland - Großbritannien"; Straße und Autobahn, 4/2002.
  7. Al Hakim B., Armitage R. and Thom N.H., Pavement assessment including bonding condition: case studies., Proceedings 5th International Conference on Bearing Capacity of Roads and Airfields, Trondheim, vol.1, 439-448, 1998.
  8. Al Hakim B., An Improved Backcalculation Method to Predict Flexible Pavement layers Moduli and Bonding Condition between Wearing course and Base course.,PhD Thesis, Liverpool John Moores University, 1997.
  9. Al Hakim B., Al Nageim H., Poutney D. and Lesley L., The Development of an improved Pavement Backcalculation, Proceedings of the 1st Conference on Rehabilitation and Development if Civil Engineering Infrastructure Systems, paper n° 72, Lebanon, 1997.
  10. Sangiorgi C., Collop A.C., Thom N.H., Laboratory Assessment of Bond Condition Using the Leutner Shear Test, 3rd International Conference Bituminous Mixtures and Pavement, Thessaloniki, GR, November 2002.
  11. Sangiorgi C., Collop A.C., Thom N.H., Non-Destructive evaluation of Bond between asphalt layers, Atti del XII Convegno Internazionale SIIV 2002, Parma, Italy, 30-31 Ottobre 2002.
  12. Uchida K., Kurokawa T., Himeno K., Nishizawa T., Healing Characteristics of Asphalt mixture under high Temperature Conditions, Proceedings 9th International Conference on Asphalt Pavements, Copenhagen, Denmark, August 2002.
STARTPublisher: DGfZPPrograming: NDT.net