Capable of surveying long sections in a short time, ground-penetrating radar can detect trackbed anomalies earlier than is possible with visual inspection methods. Maintenance can therefore be prioritised to reduce unnecessary cost and increase effectiveness

Gerard Gallagher and Prof Quentin Leiper work for Carillion plc. Maxwell Clark and Prof Michael Forde are at the School of Civil & Environmental Engineering at the University of Edinburgh. The authors would like to acknowledge the financial support of Carillion plc, GT Rail Maintenance Ltd and the University of Edinburgh. They also appreciate the support and input from many colleagues at Carillion and the University of Edinburgh.

On Britain's railway network, owned and operated by Railtrack, commercial pressures are driving efforts to improve the quality of track maintenance and reduce the length of possessions for this work. There is certainly potential for more efficient and more cost-effective methods of identifying trackbed anomalies. Current methods relying on visual inspection and the excavation of trial holes to assess ballast condition tend to be time-consuming, condition-driven and subjective in their findings.

Laboratory testing and trials using a full-scale section of track have suggested that the use of ground-penetrating radar (GPR) can provide objective and measurable information about the trackbed anomalies and the ballast/formation interface. From a GPR survey, the engineer will be able to prioritise the maintenance programme with confidence, only carrying out work where it is essential, reducing unnecessary cost and increasing the effectiveness of the programme.

Ballast characteristics

Laboratory experiments may be used as a model to examine the characteristics of ballast before GPR is used on site. One objective of the laboratory work was to investigate whether the condition of railway ballast can be classified by the dielectric constant er or by the velocity of propagation of electromagnetic waves. The experiments also aimed to assess the suitability of 500MHz and 900MHz antennae to provide adequate penetration and to produce clear images of acceptable resolution.

The laboratory experiments were undertaken in a brick tank 1, which enabled the depth of the ballast to be varied quickly and easily. Clean (unused) ballast was used, in addition to spent ballast taken from a railway line and considered to be at the end of its life span. Fig 1 shows the tank and its dimensions.

To simulate site conditions as accurately as possible, the ballast was compacted in layers as it was placed into the tank. The antenna acting as a transmitter/receiver was dragged along the top surface of the ballast layer. To reduce jarring of the antenna caused by the slight differential settlement in the surface layer, downward pressure was used when necessary. This procedure was repeated for the 500MHz and 900MHz antennae.

From the experiments undertaken, the dielectric constant for the different types of ballast was found using the equation er = (2ct/d)2, where:

c = Velocity of light = 3 x 108 m/s

er = Dielectric constant of a material

t = Time taken for electromagnetic wave to travel distance, d (s)

d = Depth of material layer (m).

The simplified equation for calculation of the dielectric constant was used because ballast was assumed to be a low loss medium 2. From the results shown in Table I, it can be seen that the dielectric constant for spent ballast is much higher than that for clean ballast.

Trackbed test rig

This laboratory-derived data was then used for an experiment using a full scale test rig (Fig 2). The mechanical properties of ballast are dictated by a combination of the physical properties of the individual ballast material and its in-situ physical state 3. Ballast by its nature is changeable with the possibility of numerous variations over small lengths of track.

Work to date has been driven by the need to calibrate the GPR equipment on the test rig. This was to allow the initial research to be carried out under realistic conditions in a controlled environment, without being restricted by possession time. The test rig was designed as a full-size section of permanent way built to current Railtrack specifications, including timber and concrete sleepers. Common trackbed anomalies can be incorporated and surveyed accurately, and the outdoor location of the rig allows weathering in the manner of working permanent way, including exposure to sunshine.

The aims of the initial experimental work included the identification of the location of ballast/formation interface, and the evaluation of the suitability of GPR to classify the degree of trackbed deterioration in terms of dielectric constants. It was also hoped to clarify what effects rail and sleeper reinforcement would have on the radar plot.

Plot results

The GPR antenna was moved continuously along the surface of the track parallel to the centre line as well as along the length of each crib. Taken along the centre line, Fig 3 shows the section of the radar plot between the clean/mixed ballast interface between Sleepers 10 to 12. The centres of the reinforced concrete sleepers are represented by the double line, whilst the centre of the crib is represented by the single line marker.

From Fig 3, it can be determined that the radar does not penetrate through the reinforced concrete sleeper. This is a result of almost all of the emitted radio waves being reflected by the reinforcing bars within the sleeper. In addition, a reflection time step can be identified between the mixed and the clean ballast. This is also the case between the spent and the mixed ballast. GPR can therefore detect clean, spent or moderately deteriorated ballast.

The test rig can be used to accurately determine dielectric constants for certain types of ballast under different moisture contents. There are strong reflections from the formation/ballast interface. From the theory and application of radar, the depth of the formation/ballast interface can be calculated using the equation:


d = Depth (m)

er = Dielectric constant

c = Speed of light in air (3 x 108 m/s)

t = Time (s)

Using Crib 11 as an example and er = 3·5 as the dielectric constant for clean ballast with a 5% moisture content 4, the depth to the formation was calculated as 0·441m. The actual as-built depth to the formation was 0·445m, resulting in an error of less than 1%.

Fig 4 shows a section of a radar plot along the length of the crib of Bed 9. Here, the metal rails do have an effect on the radar plot. The reflection of the radio waves in air can however be readily identified and removed from the plot, if required, using appropriate processing software. Rails also stop the penetration of the radio waves when the antenna is within 50mm from the side of the rail. Fig 4 shows clearly that the formation cannot be detected when the antenna is adjacent to the rail.

From the laboratory work, we concluded that the dielectric constant for clean ballast was 3·0, with that for wet clean ballast 3·5. The ideal antenna to use on ballast was found to be the 500MHz unit, and GPR proved itself to be a suitable technique to discriminate between and characterise railway trackbed ballast.

The initial testing of GPR on the outdoor trackbed test rig was very positive. This technique can detect the ballast/formation interface and is not adversely affected by the metal rail (provided the antenna is kept more than 50mm from the rail), nor by the sleeper reinforcement. GPR can be used to identify anomalies, and determine the degree of trackbed deterioration using the established values of ballast dielectric constant.

Since this work was completed, the researchers have developed a multi-antenna configuration which can quantitively classify the degree of ballast deterioration and depth, independent of changing moisture and geological conditions.

We would therefore conclude that GPR is suitable for application as a non-destructive testing technique within the rail industry.


1. Colla C, Burnside C, Clark M and Forde M. Comparison of Laboratory and Simulated Data for Radar Image Interpretation. NDT&E International, Vol 31, No 6, pp439-444, 1998.

2. Padaratz I and Forde M. A Theoretical Evaluation of Impulse Radar Wave Propagation Through Concrete. Journal of Non-Destructive Testing & Evaluation, 12, pp9-32, 1995.

3. Selig E and Waters J. Track Geotechnology and Substructure Management. Thomas Telford, 1994.

4. Clark M, Gillespie R, Kemp T, McCann D and Forde M. Electromagnetic Properties of Railway Ballast. Proc 1st Int Conf: Railway Engineering-98, Engineering Technics Press, Edinburgh, pp21-27, 1998.

  • CAPTION: An experimental section of track with differing ballast conditions was built to verify the laboratory trials with ground-penetrating radar
  • CAPTION: Fig 1. Dimensions of the brick tank used for laboratory experiments with variable ballast depths
  • CAPTION: Fig 2. Simplified general arrangement of the prototype test section
  • CAPTION: Fig 3. GPR plot along the test rig centreline
  • CAPTION: Fig 4. GPR plot along the crib of Bed 9