Requirement of a prediction

A smooth tunnel construction process is highly desirable because the construction of large tunnels may cost several ten thousand Euros a day. But dangerous accidents and expensive delays can result if major voids, water-filled fractures, cataclastic or breccia zones are encountered unexpectedly during tunnelling. Faults and layer boundaries could indicate a different geology ahead. Proximity to buildings of basements and erratic boulders require special measures to be taken. Construction engineers want to anticipate those changes in rock or ground properties to adjust drilling parameters.

Tunnelling personnel has a need for a prediction method that is robust, not disturbing the tunnelling process, and yields results quickly and at moderate costs to avoid dangerous accidents and expensive delays.

tunnel accidentsWorn out scratching tool of a tunnelling machine after a geology change has not been recognised in time (left) and water flooding of a tunnel after a water lense has been drilled unexpectedly (right).

Prediction methods

How can we obtain the desired information of the ground conditions ahead of the tunnel? Boreholes from the surface are often available but they only provide point information and are expensive. Seismic measurements, georadar and potential methods are often impossible to acquire at the surface due to habitation, rivers or mountainous terrain. They also may not yield the required spatial resolution of about 1 m. Gallery boreholes from the tunnel face delay the tunnel drilling process. Radar and geoelectric measurements in the tunnel may have insufficient penetration.

Seismic measurements from the tunnel face or the tunnel wall to illuminate the ground ahead can satisfy the tunnel-specific requirements (Sattel et al, First Break, 1992; Kneib et al., First Break, 2000). The seismic method measures the amplitude of elastic waves as function of time. Usually an artificial source (explosives, vibrator, hammer) creates an elastic wave that propagates in the earth and the movement of particles (or its velocity or acceleration) is recorded. The resulting time series' are processed to derive parameters such as propagation velocity of various wave types, the location, shape and strength of structural boundaries, dynamic elastic modula, density, attenuation, porosity, etc.. Those properties are mapped as multidimensional elastic models of the ground. Seismic traveltime tomography derives a spatial distribution of seismic velocities from seismic arrival times. Refraction seismology studies headwaves which travel along high-velocity layers in order to map the refractor depth. Reflection seismology maps the spatial distribution of contrasts in elastic properties, like the acoustic impedance. As an intermediate result reflection seismology also determines an approximate velocity model. In that sense it is the most general seismic method.

Tunnelseismology applies seismology from a tunnel, i.e. sources and receivers are placed at or close to the tunnel wall or tunnel face and elastic disturbances propagating close to and around a tunnel are studied. The aim is to predict elastic ground properties ahead of and around the tunnel.

This tunnel seismic approach faces several challenges: Humidity, dirt, and dust put considerable strain on personnel and seismic acquisition hardware. Some hardware deployed in tunnelling machines may not be accessible for maintenance in years. The prediction distances of at least few tens of metres could be drilled within a day such that seismic predictions often must be available within hours. The acquisition geometry is restricted to source and receiver positions at or close to the tunnel. As a result reflection and scattering angles are small and the spatial resolution is not optimum while imaging obstacles ahead. Resolution is further degraded by seismic attenuation which is particularly strong at the required high frequencies. To reach the desired spatial resolution of about 1 m with compressional waves, we need  signal frequencies up to a few thousand Hertz. Attenuation is also a major factor that limits the prediction distance, together with the seismic background noise level. The noise due to heavy machinery is mostly considerable in tunnels. Source-generated noise such as tunnel waves may be difficult to separate from body waves of interest. Despite the challenging nature of the recorded data the processing and acquisition will have to be automated, at least partly. It would be too costly to employ a full-time specialist at the tunnel site to process and interpret the seismic data. If those tasks are not automated they may have to be performed by civil engineers.

Challenges to Tunnelseismology:

Humidity, dirt, and dust


Tough, nearly maintenance-free equipment

Time pressure


Quasi real-time signal processing

Restricted acquisition geometry


Need for high spatial resolution

Seismic attenuation


Prediction distances of several 10 metres

High noise level


Automation of acquisition and seismic processing

Last updated August 20, 2005. © 2005, Tunnelseis. All rights reserved.