Design and optimisation of laser scanning for tunnels geometry inspection

Highlights

• This work contributes to cost effective tunnel geometry inspections using TLS.

• We optimise the scanning parameters, control network design and georeferencing.

• The size of the scanner shift is adopted based on the incidence angle criteria.

• The survey control uses the positional uncertainty and reliability concept.

• We propose arbitrary georeferencing in long tunnel scanning applications.

Abstract

The use of terrestrial laser scanning technology in engineering surveys is gaining an increasing interest due to the very high spatial density of the acquired data. Recent improvements regarding the speed, accuracy, software algorithms and the fall in price have introduced a high potential for large scale applications of this technology in highly demanding engineering environments such as tunnels. Railway tunnels, in particular those of a long length, create challenges for surveyors due to their elongation to obtain satisfactory geometry of the scanned data. The purpose of this paper is to give an optimal solution for surveying tunnel geometry using laser scanning technology to reliably inspect railway tunnels and create “as-built” documentation.

The proposed methodology provides optimisation of scanning parameters, scans registration, the georeferencing approach and the survey control network design. The maximal size of the scanner shifting along the tunnel alignment is primarily conditioned by factors including the incidence angle of the laser beam and the point density distribution. The authors introduce the so-called arbitrary georeferencing approach in long tunnel scanning that controls the point cloud geometric distortions to the required limits and contributes to time and material resources savings. Optimal design of the survey control network ensures the required positional accuracy and the reliability of the measurements, together with a cost effective approach to tunnels surveying.

The proposed methodology is followed by the empirical results of the modelling and profiling of 12 tunnels in a single track railway. The lengths of these tunnels are from 60 m to 1260 m, with a total length of 3.5 km. Due to the specific geometry of the case study tunnels, the maximal favourable laser incidence angle is 78° with a distance of 13 m and consequently the optimal size of the scanner shifting along the tunnel alignment is 26 m. The survey control network is designed with the condition that the optimal reliability factors are within the required limits for engineering networks. A priori estimation of the control network positional uncertainty and a posteriori adjustment results shows that the achieved positional accuracy of the control points is approximately five times better than the requested absolute accuracy of the tunnel model: σ_m = 2 cm. On the largest tunnel example it is shown that the arbitrary georeferencing approach assures that the optimal registration error size is within the requested limits.

Introduction

The technology of terrestrial laser scanning (TLS) allows the possibility of non-contact, direct and automated 3D measurements. The measurement process produces a dense point cloud and textured surface. Basically, TLS uses the laser ranging principle with angular laser beam deflections during the scanning process. The 3D position of the point on the object surface is defined as the x, y and z coordinates relative to the origin in the scanner electro-optical centre. An additional non-geometric point attribute is the intensity value of the reflected laser beam as a fourth dimension of a registered point in the point cloud data.

Today TLS has become a relevant measurement technique in the field of engineering surveying. Recent TLS improvements regarding it’s performance, range and accuracy have opened up new challenges within engineering. While the first line of terrestrial laser scanners had several drawbacks and were not appropriate for surveyors, the latest generations of the TLS have taken their needs into account, e.g. the feasibility for precise and reliable levelling, centring, target pointing and orienting (Schulz, 2007). Target acquisition is the procedure of an algorithmic determination of the target centre based on high resolution scanning in a separate procedure from object scanning. Typical standard scanner targets are: flat targets (black/white, circular, retro reflective, etc.) and 3D shaped targets (sphere, cylinder). The control and tie points are usually marked with targets.

The real geometry of tunnels (“as built”) may be different from their design, and tunnel geometry constantly changes with time. Understanding these changes is crucial for the cost, safety and functionality of these tunnels. These changes may reflect: (1) deformation (convergence, squeezing) produced by the reorganization of stresses brought about by excavation, during the excavation phase (e.g. Kontogianni et al., 2004), or relatively shortly after the excavation, even after an apparent stabilization (Kontogianni and Stiros, 2005) and these changes refer either to the bare rock surface or the initial lining; (2) episodic deformation produced by earthquakes (Kontogianni and Stiros, 2003) or excavations of adjacent tunnels, open pits, loading of the ground, etc. mostly during the functional period of tunnels (fully lined tunnels); and (3) gradual deformation in old tunnels, mostly related to the decay of their lining, which are mostly limited to visual inspection.

Different contact (probe and protractor, finger probes, tape extensometers, train section profilers, etc.) and non-contact (theodolit, total stations, photogrammetry, optical triangulation, railway gauging train) instruments and methods have been used over the years to acquire the geometrical data of structures such as tunnels. This type of data can be used for the estimation of clearances, the checking of alignments in lift guide rails and ducts, the monitoring of changes and deformation analyzes, the compilation of inventories and “as-built” drawings, the determination of the volumes of excavation or lining materials, an indication of structural failure, information collection for refurbishment, checking the driving of tunnels and monitoring progress of projects (Clarke and Lindsey, 1992).

Under regulations in ISO 1101 or its equivalent ANSI Y14.5M, all types of geometric tolerance (form, profile, orientation, location and run-out) can be found by using TLS and analysed geometrically. These standards were originally developed for the manufacturing industry but have been adopted for the assignment of the geometric tolerances of concrete and steel structures in tunnel construction (Lam, 2006).

Recently TLS technology has begun to be recognised for its potential in the underground environment, especially in the tunnelling industry. The active tunnelling environment is more demanding and challenging than those where TLS has traditionally been used. Tunnels require robust systems that are effective in dusty and damp conditions and that can perform independently of underground lighting. In the past, the use of TLS in tunnels was impractical due to the long scanning times (Fekete et al., 2010). The new generation of TLS are characterised by very high point density, a very short scanning time frame, and a higher accuracy of 3D data. These performances and the decreasing trend of the price of TLS, allows significant applications of TLS technology in tunnel engineering projects. However, TLS is an optical method, and hence cables, machinery, the instrument itself can produce non-measurable areas on the tunnel surface.

Alternatively tunnels with elongated geometries particularly those of a single railway track are a very specific unfavourable case to provide geodetic measurements of a sufficient accuracy and reliability. When registering a sequence of scans along the tunnel, the errors of global registration may be rather large. This happens due to the error propagation from the first scan to the subsequent scans (Reshetyuk, 2009, Bornaz et al., 2003). A similar situation is found in alignment and breakthrough errors in tunnelling traversing. For example, if one new variable is the function of different variables which are not error free, than the new variable will also be influenced by errors. Errors propagate with the sum of variances, i.e. the variance of the new variable equals the sum of variances of the other two (Stiros, 2009). The longer the tunnel, the larger becomes the error budget of the global point cloud registration, which contributes to the pure geometrical quality of the complete tunnel 3D model. For such reasons, high precision scanning application in tunnels requires an adequate control geodetic network, with two main aims: to calibrate and check the scan data and to give a reference coordinate frame to the point cloud.

This work contributes to cost effective tunnels geometry inspection using TLS. It aims at the optimisation of the scan process, the scan registration, georeferencing and reliable survey control. The most important time limiting factor in tunnel scanning tasks is the maximal distance between adjacent scanner positions. In respect to tunnel geometry the maximal favourable distance between adjacent scanner positions has to be adopted based on factors such as the incidence angle of the laser beam and scanning resolution. An adequate registration and georeferencing method has to be conducted, especially in the case of long tunnels. The authors propose the so-called arbitrary georeferencing approach for long tunnels and a methodology of survey control network design.