The text ONLY of the Paper is given below; but it will probably be more useful to read the .pdf file that includes illustrations and can be found in the Library. The text is given below simply so that Google and other "crawl engines" can find the text.
Failure in NATM (SCL) tunnelling
Cracking of shotcrete.
A.Paul Takacs Independent Consultant
Robin B. Clay Dar Al Handasah, Turkey
Shotcrete lining is the main supporting element in NATM tunnelling, often used together with rock bolts and steel ribs, forming a composite structure to support the ground. It sometimes happens that large deformations of the lining are recorded by the deformation monitoring system, with little or no evidence of distress being noticeable in the shotcrete. In other cases, it sometimes happens that conspicuous cracks occur in the shotcrete, but with no significant matching deformation. Such cracks, which prima facie may suggest that a catastrophic failure of the entire supporting structure may be imminent, are frequently found adjacent to steel ribs, at the joint between top heading and bench, and at other similar locations. This Paper gives some possible reasons for these two situations in the abstract, investigates possible reasons for the formation of cracks and possible mechanisms involved, and explores why cracks are not noticed sooner. 1
In a previous Paper about a tunnel break-through in difficult geological conditions  we gave one instance of the cracking of shotcrete. In another Paper on case histories of tunnelling incidents  we also mentioned the appearance of cracks as a precursor to an event. Based on these observations we may offer the axiom that “no serious damage occurs without previous cracking”. However, the reverse is not axiomatic - cracking does not always precede serious damage.
1.1 Construction joints
Almost all engineering structures are assembled or constructed in stages that are separated by joints - which may be considered as a special type of crack. This observation allow us to state that “no tunnel is without cracks”. The materials from which they are made are themselves inherently prone to cracking . The main difference between a joint and a crack is that joints are deliberate and are considered both in the design stage and during the construction stage of a project, and they are designed and constructed intentionally, to satisfy technical or constructional needs rather than structural requirements, while a crack, by contrast, is aleatory and unintentional, and so is not taken into consideration in either stage of the process.
1.2 Cracks and stress
Masonry arch structures are jointed (radially) and yet their carrying capacity is not diminished by the presence of these joints, because the joints are subjected notionally solely to compression. From simple observation, discontinuities subject to compression tend to close and only those subject to tension tend to open and extend. It is obvious that, so long as the structural element is under compression, its structural integrity is not affected by discontinuities perpendicular to the compression. In the case of the other two known types of crack-inducing stress, i.e., bending and shear, the problem is not so simple. An arch may suffer bending, but this simply means that the compression on one side is less than the compression on the other side; there is no tension. The side with the higher compression will compress more, resulting in greater deformation of the structure. When a discontinuity appears due to either bending or shear , the discontinuity acts as a hinge and the structure become multi-hinged, the only noticeable effect is that the deformations are larger, but the carrying capacity is not significantly affected, the bending effects being transformed into compression. The most critical cracks to be considered are those inclined to the direction of the load, because they allow relative displacement of the two sides, opposed only by the shear resistance between the two faces.
1.3 Lining / Ground interaction
Almost all underground structures are constructed in the form of an arch into surrounding ground which only accidentally may be without discontinuities (fissures, joints etc.). This ground has its own carrying capacity (which in some design approaches is intentionally ignored), and this acts together with the applied structure (the support) to give stability to the underground opening. Instances where there have been problems due to a lack of interaction between the two components of the arch is well presented in the technical literature ,,,,and for this to occur, it is necessary that there is no direct contact between the two elements, or that the contact is not uniform. Relaxation of the rock, or of the ground in general, will sooner or later allow contact to occur. From this moment on, the two components will interact in such way that, in time, a combined rock/support arch will be formed. However during this process the support may give signs of weakening, in early stages when the contact is not uniform and causes non-uniform loading. Such a phenomenon has been described thus: “At both sides of the top heading, next to the footings, longitudinal continuous shear cracks were formed with an overlapping of the edges of about 30 cm. Due to this overlap the lining spalled from the rock resulting in considerable voids and further deterioration and loosening. some of the anchors were broken, some were deformed according to the tangential movements.” If such signs do not appear, at least for NATM, then this shows that the rock itself is self-supporting, and the applied support is superfluous.
It was observed above that shear-induced cracks are the most critical, although this is not a universal opinion. In one approach, when large deformations are expected, the support is longitudinally “slotted”, with the intention of eliminating the cracks by artificially and deliberately forming a much larger open joint. But what is the difference? What is the risk? If with formed slots the structure does not fail, would it have failed with natural cracks instead? However, it must be emphasised that all collapses (general loss of stability) occur, inevitably, by cracking of the structure. This is one of the paradoxes of modern tunnelling. 2 Relation between cracks and monitoring Generally speaking, there are no tunnels that do not have cracks in the shotcrete lining. It is the duty of the specialist to evaluate the risk of these cracks in respect of the global stability of structure. It is very easy to decide “for safety’s sake” to apply some additional support, the extent of which depends more on the temper of the decisionmaker than on the realities of site. It should not be forgotten that, regardless of quantity, any additional support means additional cost and additional time that may lead to delays in the programme. In most cases, cracks are discovered long after their formation, and any additional support is applied after the structure has already achieved stability by itself. Recent experience from the construction of some tunnels in the Near East has shown that the decision to cover cracks with a thin layer of shotcrete is in most cases the most efficient method of absolving us from the “risk obsession”. It may be added that there is little doubt that, in these tunnels, almost without exception, the cracks in the shotcrete lining were due to deficiencies in execution. This Paper attempts to find a possible relation between the occurrence of cracks and the results of monitoring. It should be mentioned that it is only by accident that a monitoring section is in the cracked area; in most cases the monitoring section is merely ‘in the vicinity’, i.e. from some metres to some tens of metres away from the cracked area.
3 Cracks in Tunnel 4
The cases described here are linked to only a few of the numerous cracks identified during construction of this pair of tunnels.
3.1 North Tube
On 30.06.1993 in Round 267 at km. 11+028, cracks were observed at the connection of the top heading with the bench. Strengthening consisted of the installation of some additional rock bolts in rounds 264 to 269. There was by chance a monitoring section in the middle of the area, at km. 11+024. The deformations monitored by this station are shown in Figure 1. The occurrence of cracks is not believed to be related to the cross adit constructed nearby, at chainage km. 11+078. Figure 1. Deformations at MS 11+024. 3.2 South Tube - March 1994 Cracks were observed on 22.03.1994 between Round 162 and Round 174, between km 21+175 and km 21+162. Strengthening was done by the installation of some additional rock bolts. The closest monitoring section to the affected area was at MS21+132. The deformations recorded at this station are shown in Figure 2.
Figure 2. Deformations at MS 21+132. p. 5 3.3 South Tube - April 1993
On 06.04.1993 cracks were observed in Rounds 268 to 269, between chainage km 21+029 to km 21+033. Strengthening was again done by the installation of additional rock bolts. In this case, two monitoring sections were in the vicinity, at MS21+058 and MS21+029. The deformations monitored at these stations are shown in Figure 3 and Figure 4 respectively.
Figure 3. Deformations at MS 21+058
Figure 4. Deformations at MS 21+029
Two weeks later, on the 18.04.1993 in the same tunnel, a collapse occur at chainage km. 21+963, 70m away from the above mentioned stretch. No link or relation is drawn between the two events. 3.4 South Tube - early May 1993 On 10.05.1993 on the right side between Round 271 and Round 275 and on the left side between Round 274 and Round 275, cracks were observed which also were then strengthened by additional bolts. This is a stretch in the vicinity of a cross adit between the two tubes. In a later construction stage during trimming and re-profiling for the final lining, it become obvious that the junction between the topheading and bench had been not formed correctly. The above mentioned monitoring sections MS21+058 and MS21+029 were once again used for risk evaluation.
3.5 South Tube - end May 1993
Two weeks after the initial discovery of cracks, on 26.05.1993, in Rounds 261 to 269 between km 21+043 and km 21+030, more cracks were observed, which were in turn consolidated by additional rock bolts. The same monitoring section mentioned above was used to justify the additional support measures.
4.1 Monitoring data
The figures are scanned from the original monitoring diagrams produced by the ‘convergence and levelling’ monitoring method. A large number of similar charts for other monitoring stations were generated during the construction of over 7 km of tunnels, and, being produced from original readings, generate a “jagged” effect rather than the “smoothed” theoretical lines more usually seen.
4.2 Discovery of cracks
Cracks are reported only after they have been identified during routine or special visual inspections. In general, the width of a crack is but a few millimetres and rare are the cracks which cause spalling of the shotcrete. Cracks or surface irregularities only become noticeable due to shadow or difference in colour of the shotcrete. However, the lighting of a tunnel under construction, away from the face, is not the best for such small details and routine inspections are done from a distances of several metres away from the surface. Generally speaking, cracks are discovered by accident, and are frequently already large before they are noticed (which may be the reason why they are noticed).
4.3 Seriousness of cracks
In a case of a general loss of stability the development of the cracking is exponential. At the beginning, just a few fine cracks appear which grow and multiply with time. When they become visually noticeable, it is almost too late for any practical action because their propagation and development is by now too rapid. In practice, there is no time even to evaluate the potential risk of visually-recorded cracks. On the other hand, some cracks were monitored immediately upon discovery by installing glass tell-tales, and these showed that in the great majority of cases the cracks had already stabilised or their development was very slow. However, there can be no guarantee that what is today stable will not tomorrow become active once more. A true assessment of the risk that a crack represents should focus in principal on the potential effect of further excavations in the vicinity.
One reason for monitoring deformations is the hope that a crack might be discovered by an interpretation of monitoring readings. As mentioned above, the normal width of a crack is of the order of just one millimetre. The distance between monitoring pins is of the order of metres and in ‘normal’ conditions of deformation this distance changes each day (at least in the first period of monitoring) with a velocity of many millimetres per day. At the same time, the deformation rate is not constant from one section to another. These factors combine to make it virtually impossible to predict if or when the shotcrete will fail, and should it fail, to predict either the amplitude and speed of this failure, and if it will then lead to a general instability of the structure, or will stabilise by itself.
5 Conclusions and recommendations
The ability to predict the risk of structural instability of underground construction by means of geotechnical monitoring has yet to be proved. It seems more likely that the concept itself, though attractive and possibly theoretically sound, is basically flawed and reliant upon a false understanding of the practical realities both of site practice and of Mother Nature herself as manifested in the always variable ground. Obviously, cracks are the most evident signs of a possible weakness of a structure. All of those cracks discussed above were discovered during visual inspections. By geotechnical monitoring the highest achievement was the ability to say that cracks would probably occur, in our particular geological conditions of the project, once differential deformations had reached fifty to sixty millimetres. Geotechnical monitoring is a costly activity, although according to some estimates less than 0.1% of the construction costs . However the cost is related to the monitoring method, the distribution of monitoring sections and the frequency of readings. In particular conditions this may reach up to 3-4% of the excavation cost. The practical possibilities and achievements have yet to be defined. Any geotechnical monitoring system should be a practical tool in the hands of site practitioners and the operating level of such systems must be brought to the average level of skill on site, rather than being just the preserve of a minority of highly-qualified professionals not often seen daily on work sites; otherwise it is no longer a practical operating tool for day-to-day decision-making. For this purpose, a concentrated effort is needed to produce an operating manual or code of good practice, or at least a set of guidelines for using geotechnical monitoring. Without such, questions such as this one, posed by de Mello  : “Has anybody seen real case-history PREDICTION characteristic curves with their successive on-job revisions as direct experience should dictate? Or are such elusive predictions kept in the waistcoat pocket until after tunnelling finishes and the time comes for publishing? Can experience be gained without a fixed target for aiming and progressively correcting?” will remain pure rhetoric.
 Clay R.B. & Takacs A.P., 1987: Anticipating the unexpected; Tunnelling ‘97, The Institution of Mining and Metallurgy, ISBN 1 870706 34 X
 Golser J.: Controversial Views on NATM; Rock Engineering 2/96 April
 Golser J.: The New Austrian Tunnelling Method (NATM) (Theoretical backgroundpractical experiences); Conference on Shotcrete for Ground Support ASCE/Easton 1976
 Golser J.: The New Austrian Tunnelling Method (NATM); (manuscript)
 Rabcewicz L.v.: The New Austrian Tunnelling Method; Water Power, November and December 1964 and January 1965
 Takacs A.P. & R.B.Clay, 1998: Lessons from a NATM (SCL) breakthrough in difficult geological conditions; 8th Congress of the International Association of Engineering Geology and Environment, September 1998, Balkema
 Takacs A.P.,1997: Tunele si metrouri pentru cai de comunicatie; Capitolul 3: Tunelul: Manifestare plenara a artei ingineresti, Subcapitolul: Monitorizarea deformatiilor; (unpublished)
 S.J. Boone et al, 1998: Building and utility damage assessments, risk, and construction settlement control; Proceedings of the World Tunnel Congress’98 on Tunnels and Metropolises vol.1; A.A.Balkema
 V.F. de Mello, 1996: Fallacies in routines of NATM (RSST) shotcrete supported tunnelling and promises thereof; Proceedings ITA’s Congress North American Tunnelling’96, A.A. Balkema
 J.E. Gordon: The New Science of Strong Materials, or, Why you don't fall through the floor Chapter 4, Penguin Books, Ltd., UK. ISBN 0 14 02.0920 4