You are on page 1of 6

Underground Space the 4th Dimension of Metropolises Bartk, Hrdina, Romancov & Zlmal (eds)

2007 Taylor & Francis Group, London, ISBN 978-0-415-40807-3

Expert system for D&B tunnel construction


C.W. Yu & J.C. Chern
Sinotech Engineering Consultants, Inc., Taipei, Taiwan

ABSTRACT: To improve the technology for drill and blast tunnel construction in Taiwan, an expert system
consisted of data bank, tool bank and decision making auxiliary system has been developed. Various tools
can be used efficiently in carrying out data collection, processing, analysis and evaluation work required in
the tunneling process. Using the successful tunnel case histories data learned from the past experiences, the
system can provide multi-expertise assistance on the decision of support system and excavation procedures to be
adopted in the construction site. The system can also provide rational estimation on the tunnel deformation under
the selected support system and construction procedure with the aid of an artificial neural network approach.
Application of the system to actual tunnel construction is given for demonstrating the capability of the system.

INTRODUCTION

The engineering practice for drilling and blasting


(D&B) tunnel construction in Taiwan adopts similar
approach used in the western countries. In the design
stage, support requirements and construction procedures were pre-determined by using empirical rock
classification method, mainly based on limited geological information and rock mass conditions obtained
from the preliminary investigations.
During construction, actual rock mass rating was
obtained at the tunnel face, and appropriate support
system and construction procedures were selected and
implemented. Monitoring instruments were installed
for observing the performance of rock masses surrounding the tunnel. Revision and remediation were
made on the support system or construction procedures, if necessary, to achieve a safe and economic
tunnel construction.
However, due to the poor rock condition often
encountered in Taiwan, excavation large span tunnel
through poor quality rock mass has resulted in numerous engineering difficulties in the past two decades.
These difficulties included uncontrolled excessive tunnel deformation, serious damage to the support system
and even tunnel cave-in.
Figure 1 shows an example of driving a large
span tunnel through a fault zone in northern Taiwan
by using the traditional rock classification method.
It required systematic pre-stressed tendons and remining and re-supporting the tunnel to overcome
the uncontrolled excessive deformation. Controversies over the adequacy of tunnel support design
and construction procedures based on experiences of

Figure 1. Systematic pre-stressed tendons, re-mining and


re-supporting had to be used to control large deformation of
tunnel through a dry fault zone in the construction of the
Second Freeway in northern Taiwan.

different geological environments have become issues


for the local tunneling community. This also calls for
the development of a more rational system to deal
with the processes in D&B tunnel construction. This
paper presents the system and its application in a tunnel
construction project.

EXPERT SYSTEM FOR TUNNEL


CONSTRUCTION

The tunnel expert system compiles expert knowledge


and tools for facilitating the data collection and processing, data analysis and evaluation, and decision

799

(2)

(3)

(4)

(5)
Figure 2. Working processes of the expert system
developed.
Table 1. Tool bank of the expert system.
Tools

Main functions

TUN_MAP

Geological Information Collecting and


Processing
Key Block Analysis
Construction Decision-making Assistance
Deformation Prediction
Construction Simulating
Tunnel Safety Evaluation
Construction Status Demonstrating

TUN_KBL
TUN_AUX
TUN_DEF
TUN_CAD
TUN_SAF
TUN_STA

(6)

(7)

making in the tunnel construction processes. Figure 2


shows the working processes of the system.
The working processes contain 3 major blocks, i.e.,
(1) construction data collection at the site, (2) tool bank
for various data processing, analysis, simulation and
evaluation, and (3) decision making for support design
and construction procedures recommendation, safety
evaluation and construction status demonstration.
The functions of each module in the tool bank
are outlined in Table 1. The working process will be
described further in the following sections.

The application for some of the systems will be


demonstrated by the results of case study presented.
4

The tool bank of the expert system comprises of 7


subsystems as shown inTable 1, and the main functions
of the subsystems are:
(1) Geological Information Collecting and Processing System (TUN_MAP) utilizes digital image
taken on the tunnel face to retrieve the geological
information for statistical analysis of joint systems

DECISION-MAKING MODULES

Decision-making modules include the recommendation of optimum tunnel support requirement and
construction procedures, tunnel deformation estimation and safety evaluation during tunnel construction
stages. It also provides the status of tunnel construction for better control of construction progress and
tunneling conditions by the tunnel engineer.
4.1

3 TOOL BANK

and production of developed geological map of


tunnel surface.
Key Block Analysis System (TUN_KBL) can
be used to perform key block analysis from the
actual recorded joint information or statistically
generated joint information.
Construction Decision Making Auxiliary System
(TUN_AUX) is a case history based artificial
intelligence system for making multi-expertise
recommendations on tunnel support system and
construction procedures.
Tunnel Deformation Prediction System (TUN_
DEF) is an numerical analysis data based artificial neural network system for providing a rational
estimation on the tunnel deformation under various geological and construction conditions.
Tunnel Construction Simulation System (TUN_
CAD) provides numerical analysis programs for
carrying out back analysis of monitoring data and
forward prediction analysis of tunnel performance
by simulating the construction processes in the
field.
Tunnel Safety Evaluation System (TUN_SAF)
utilizes the monitoring tunnel deformation for
evaluating the safety of tunnel construction by
using a case history based empirical tunnel safety
criterion.
Construction Status Demonstration System
(TUN_STA) uses the records collected during
tunnel excavation, including geological information, monitoring data and construction progress
records, for demonstrating the status of tunnel
construction.

Optimum support requirement

A Decision Making Auxiliary System (TUN_AUX)


is a case-based artificial intelligence system together
with the traditional empirical rock classification system. The working processes of system TUN_AUX are
shown in Figure 3.
In the field, the parameters, including the tunnel
span, strength of rock material, rock mass classification and overburden depth, collected are used for the

800

Figure 3. Flow chart of application of rock support decisionmaking auxiliary system.

Figure 4. Optimized chart for selecting a suitable excavation procedure according to tunnel span and the strength
stress ratio.

expert system. TUN_AUX can give a multi-expertise


recommendations on the major support elements such
as the length and spacing of rock bolts, thickness of
shotcrete, size and spacing of steel sets, etc.
4.2

Optimum construction procedures

The Decision Making Auxiliary System (TUN_AUX)


also provides a function for the recommendation of
optimum construction procedure. The excavation procedures suitable for various tunnel spans and ground
conditions were categorized into full face, central
diaphragm (CD) and side drift. Based on the expert
knowledge collected from experienced tunnel engineers and the results of construction simulation under
various tunneling conditions, the optimum construction procedure is suggested by the tunnel span and
strength/stress ratio at the construction site. The chart
for selecting a suitable construction procedure is
shown in Figure 4. Figure 5 shows the construction of a four-lane freeway tunnel under poor ground
and very shallow rock cover by using dual side-drift
method. Severe unstable condition was developed in a
3-lane adjacent tunnel with top heading and benching
procedure.
4.3 Tunnel deformation estimation
Tunnel Deformation Prediction System (TUN_DEF)
provides an estimation on the magnitude of tunnel
deformation expected under various geological and
construction conditions. The system was developed
based on large amount of 3-D numerical simulations
of the tunnel construction under various geological
conditions, excavation sequences and support measures. Tunnel deformations, including roof settlement,
horizontal convergency, as well as plastic zone thickness were given by a back propagation artificial neural
network from the data base. It is intended to provide
a rough estimate on the order of magnitude of tunnel
deformation expected as guidance for tunnel engineer.

Figure 5. Four-lane freeway tunnel driven by dual side-drift


excavation method adjacent to a shallow slope.

4.4 Tunnel safety evaluation


Empirical tunnel safety criterion based on case histories in Taiwan (Chern, et al. 1998) is used in the safety
evaluation of tunnel. In the criterion, the information
of the tunnel inward movement (), tunnel radius (R),
rock mass strength (UCS) and actual performance of
tunnel case histories were used to establish 3 warning
levels of tunnel safety and the necessary measures to
be taken.
Below warning level I, tunnel is considered to be in
stable condition, and no special measure is required.
Between warning levels II and III, tunnel is susceptible
to instability, and review of support system and construction procedures adopted should be made. Above
warning level III, tunnel is in unstable condition, and
remedial measures should be taken to stabilize the
tunnel.
Figure 6 shows the flow chart of tunnel safety
evaluation using TUN_DEF and TUN_SAF systems.

801

Table 2.
system.

Suggested support elements predicted by the

Support type
Design Type III
Design Type IV
Suggested
Type III
Suggested
Type IV

Figure 6. Flow chart of tunnel deformation prediction and


safety evaluation.

Figure 7. Geological profile of the flood diversion tunnel.

1.01.4 m 4 m
0.81.2 m 6 m
1.52.0 m 68 m

20 cm
G220
25 cm
G220
2025 cm H125

1.01.5 m 810 m

2025 cm H150

Use lattice girdle. Use H-beam.

Figure 9. The highly disturbed rock exposed at 2k+008.

Figure 8. Tunnel section showing the excavation span and


excavation steps.

Advance Rockbolt Shotcrete Steelper round length


thickness rib size

CASE STUDY

To illustrate the application of the system developed,


a case study on a flood diversion tunnel in northern
Taiwan is presented.
The diversion tunnel is 2,835 m long driving
through sedimentary rock formation of Miocene Age.
The geological profile is shown in Figure 7. The maximum excavation span is 14.1 m and was excavated in
3 stages as shown in Figure 8. The major difficulties
encountered in tunneling are associated with the fault
zones and the disturbed zones adjacent to a volcanic
intrusion.
The rock support system was designed according
to the traditional RMR rock classification method.

Five support types were proportioned according to the


RMR rating. The rock support requirements were also
examined by using the expert system TUN_AUX during construction. The results for type III and type IV
rock are shown inTable 2 along with the original design
of support system.
Most of the tunnel sections belong to the good quality rock with type II and type III support systems. The
type III support adopted worked quite well. However,
in a disturbed zone adjacent to the volcanic intrusion
near station 2k+008, the poor quality rock mass (Figure 9) with RMR values in the range of 28 to 34 showed
squeezing behavior under 140 m overburden stress.
The monitoring results during tunnel excavation are
shown in Figure 10.
The monitoring data for top heading excavation
showed potential unstable condition according to the
tunnel safety evaluating system TUN_SAF as shown
in Figure 11. This proved to be true by the tunnel
convergence occurred during subsequent benching.
Reviewing the support requirement and excavation
procedure, it was found that the expert system would
suggest a more conservative support system and excavation procedure. These can be shown by the longer
bolt length as suggested in Table 2, and excavation by
CD or side drift method recommended as shown in
Figure 12.
Before benching, various remedial measures were
further studies by using 3-D numerical simulations.

802

enhance the tunnel stability. The tunnel did not suffer


from severe unstable condition during benching, but
rather large tunnel convergence occurred with accompanying cracks of shotcrete in the tunnel crown area.
This is also predicted by the tunnel safety evaluation
system as shown in Figure 11.
6

Figure 10. Monitoring results during tunnel excavation in


disturbed zone near 2k+008.

Figure 11. Potential unstable condition according to


TUN_SAF evaluation.

Figure 12. Excavation by CD or side drift method as


recommended.

The predicted tunnel deformations after benching


down are shown in Figure 11. Due to limited option
available at the construction site, only additional
9-meter long systematic bolts were installed to

CONCLUSIONS

An expert system consisted of tools and expert knowledge for performing construction data collection and
processing, data evaluation and analysis, and decision
making assistance in the working process of D&B
tunneling was developed. The results of preliminary
application of the system in the construction of a
flood diversion tunnel showed the system can provide
efficient tools for carrying out various data collection, analysis and evaluation work in the tunneling
process. Rational suggestions on support requirement, construction procedure and safety evaluation
can also assist the field engineer in the decision making
process.
However, In a practical application sense, the system developed is considered to be a prototype due
to the lack of precedence and case histories with
sufficient accuracy covering wide range of tunnel
conditions. For further improvement of the system,
trial application and collection of additional case
information are still needed.
REFERENCES
Hoek, E. and Brown, E.T. 1980, Underground excavations in
rock, London: Institution of Mining and Metallurgy.
Dershowitz, W.S. and Einstein, H.H. 1984, Application of
Artificial Intelligence to Problems of Rock Mechanics,
25th W. S. Symp. Rock Mechanics, pp.483494.
Zhang, Q., Mo, Y. and Tian, S. 1988, An Expert System
for Classification of Rock Masses, Proc. 29th U.S. Symp.
Rock Mech., pp.283288.
Butler, A.G. and Franklin, J. A. 1990, Classex: Expert System
for Rock Mass Classification, Proc. ISRM Int. Symp. Rock
Mech., Mbabane, Swaziland, pp.7380.
Zhang, Q., Nie, X.Y. and Tian, W.T. 1995, A Case-Based
Reasoning for Tunnel Support, Proc. 8th Int. Cong. Rock
Mech., Tokyo, pp.907909.
Chern, J.C., Shiao, F.Y. and Yu, C.W. 1998, An empirical
safety criterion for tunnel construction. Proc., Reg. Symp.
Sedimentary Rock Engineering, Taipei, 325330.
Palassi, M. and Franklin, J.A. 1998,Tunnex:An Expert System
for Tunnelling through Rock. Tunnelling Association of
Canada (TAC) News, Vol. 16, No. 1, pp.1218.
Hoek, E. 2001. Big tunnel in bad rock. Journal of Geotechnical & Geoenvironment Engineering, Vol.127, 726740.
Liu, J.M. 2005. River Yuanshantzu flood-diversion project.
Proc., Int. Symp. Design, Construction and Operation of
Long Tunnel, Taipei: 923933.

803

You might also like