Intelligent tunnelling information system

S. W. Hong Korea Institute of Construction Technology G. J. Bae Korea Institute of Construction Technology C. Y. Kim Korea Institute of Construction Technology Y. S. Seo Korea Institute of Construction Technology K. H. Lee Korea Institute of Construction Technology C. H. Park Korea Institute of Construction Technology

ABSTRACT: In the paper the authors describe the development of an integrated expert system prototype for site investigation, design and construction in tunneling. Geostructure Research Group in Korea Institute of Construction Technology (KICT) has developed the system during the past 2 years. The system mainly consists of 6 modules (1) visualisation module for ground data and structure, (2) analysis module for ground stability, (3) prediction module for ground surface settlement, (4) selection module for tunnel reinforcement, (5) prediction module for face failure, and (6) Virtual Reality (VR) for visualisation of tunnel construction. 1 INTRODUCTION In general, experience plays the most important role in construction of geotechnical engineering structures such as a tunnel. The reason for this is due to the difficulty of gathering sufficient information on the geology and evaluating the true mechanical behaviour of the ground at the design stage. And geological information is generally difficult to quantify because of imprecise and incomplete data observation. With insufficient information on the geology and the mechanical behaviour of the ground the tunnel design is carried out. The concept of 3-D visualisation system certainly gives a new aspect in the field of geotechnical engineering. The analysis and interpretation with attention to geotechnical or geological data using 3D visualisation techniques are valid to understand geotechnical site characteristics. Furthermore, the adequate analysis prevents construction delay and additional cost and leads to minimise geological and geotechnical uncertainties. As 3-D visualisation system has become popular in recent years, the authors have started to investigate how to systemise the experiences and decision making procedures of experts in tunnel design and c o n s t r u c t i o n b y u s i n g t h e 3-D visualisation technique, artificial intelligence (AI) and database (D/B) system since 1999. A national research laboratory (NRL) was established for developing the Intelligent Tunneling Information System (ITIS) in September 1999. As a result of 2 years studies, the prototype system consisting of 6 modules was develo ped. The modules are (1) visualisation module for ground data and structure, (2) analysis module for ground stability, (3) prediction module for ground surface settlement, (4) selection module for tunnel reinforcement, (5) prediction module for face failure, and (6) Virtual Reality (VR) for visualisation of tunnel construction. In the paper, the authors introduce the overview of ITIS such as the features and the functionality for analysis and interpretation. 2 SYSTEM OUTLINE Geotechnical site investigation, tunnel design and construction data generally exist in forms of manually completed documents , project reports, regional geotechnical maps and geotechnical relational database, etc. And the analysis of these data that is located against some form of map or plan is generally in two dimensions. ITIS is an integrated D/B management, analysis and visualisation system

for geotechnical projects especially in tunneling. It provides the environment that makes much easier to create 3-D subsurface model and analysis results and creates presentation-quality. In developing ITIS the important considerations are easy-operating environment and PC-Windows-based for civil engineers who have antiquated computing capabilities. ITIS was developed using C++, Open GL and Delphi for language, and GEOMania, GDK (GEOMania Development Kit) and GEOMania/3D for program and runs on Pentium PC operating under Windows 98/NT operating system. Using the graphical user interface, the user can operate the system with the mouse and the keyboard. The system consists of input, output, and analysis/interpretation modules. There are a lot of

input data available such as basic geological information acquired from borehole, field investigation and geophysical exploration, in situ test, laboratory test, pre-investigated reports and tunnel initialisation information, including start and endpoints. These data are managed and analyzed in a lump to make a 3-D model and several types of analysis with adapting mathematical and scientific algorithms . Within the system it is possible to zoom, move, rotate, cut, select a gradient and custom colouring, and copy/paste directly to other Windows compatible software. The user can take cross sections in all directions and create subsurface profiles in the cross sections. All of cross sections created are easily compared with each other.

Figure 1. ITIS graphical user interface with 3-D windows.

Using the report functionality, the user can get most of the requested information. The

extensive pre-prepared report formats are available. Furthermore, the query options can be used to generate lists of data that exist within the database. These can be exported as a pre-prepared files. Figure 1 shows the IT IS graphical user interface with 3-D windows. 3 MODULES OF ITIS 3.1 Visualisation module for ground data and structure (Hong 2001a, b) The module allows engineers and scientists to input, edit and manage data, and create 3-D models easily. Once logged on to the system, the user meets the Graphical User Interface (GUI). The GUI comprises two units. One enables the user to use all the functions such as file management, digital map editing, data input and analysis, drawing, setting of options, and help. The other is used for creating 2-D and 3-D models with selecting projects. It is possible to zoom, measure scaled distances, drag, select layers, and copy/paste directly to other compatible programs. This module treats most of the data observed and measured in site investigation. - Three-dimensional geomorphologic data (x, y coordinates and elevation)

Borehole location and depths o f samples and stratigraphic samples taken in borings Physical property and other attribute data measured at outcrops In Situ and laboratory test results Tunnel initialisation information, including start and endpoints

The user manages systematically all the data that the user wants to include in the project by using Object - oriented database system (OODB). Figure 2 shows the input and output windows for borehole information. Details of boreholes collected and observed are entered into a borehole database. It consists of project, company and borehole classes. Typical data contents are described in Table 1. With using the boring log function, borehole log is created from data and is automatically updated when the data is modified. A standard heading and contents of columns could be selected to form a suitable log format. Figure 3 shows the windows of cross-section and 3D model. This module allows the user to visualise two-dimensional surfaces, fence diagrams and threedimensional solid models. In 2-D surface the user creates surface models and contours for geology, rock mass classification, mechanical properties of

Figure 2. Input and output windows for borehole information.

Table 1 Typical data contents for borehole classes. Class Project Company Borehole (general) Borehole (technical) Data content ID and name, input-date, type, start and endpoint, period, description ID and name, address, president, field ID and name, input-data, location, coordinates, elevation, groundwater level, depth, direction, angle, driller, inspector, boring and machine type, etc. Sample number, method, depth, type, result of SPT, TCR/RQD, type and spacing of discontinuity, descriptions, etc.

introduced the procedure of making 3-D model in detail. 3.2 Analysis module for ground stability Rock masses usually contain several kinds of discontinuities such as bedding planes, joints, faults and other mechanical defects that show the mechanical characteristics of low shear strength and high permeability. One of the most important characteristics of discontinuities is their orientation relative to geotechnical structure or excavation face in tunnel. At the design stage of tunnel the location and orientations of actual discontinuities cannot be measured at the design site because the rock face is usually not accessible. In this reason it is necessary to measure discontinuity information from other rock faces and boreholes. The module is the tool to assist building a statistical model that represents the discontinuity orientation characteristics of the rock mass. It uses the stereographic projection method that simplifies graphical solutions for problems involving the relative orientations of planes in space and allows analyzing and visualizing discontinuous plane data measured in field or from borehole. The module has

several stratigraphic 2-D surface models. They also support the same features as 2-D surface models. All of the created cross sections are easily compared with each other. This function is useful to discuss or verify complex geologic site conditions and to solve a problem of disagreement between staffs concerned with construction projects. For creating 3-D model borehole and 2-D surface information are used and analyzed with adapting mathematical and scientific algorithms in a lump. The process of generating 3-D model from the relative data is quick and simple. Hong et al. (2001b)

Figure 3. Creation of cross-section and 3-D model.

3.3 Prediction module for ground surface settlement (Kim 2001a) Some tunnel and underground excavation inevitably disturbs the ground and the original stress field, which in turn causes ground movement leading to surface settlement. The ground movements can be large enough to disrupt the function of nearby structures and utilities. Particularly, in urban areas, the freedom of choice of alignment and tunnel depth is rather limited. There are several empirical and semi-empirical formulae available (Murayama 1971, Sagaseta 1998) for predicting ground surface settlement, and most o f t h es e d o not simultaneously take into consideration all the relevant factors, resulting in inaccurate predictions. The module is for the prediction of ground surface settlements incorporating artificial neural networks (ANN). In this module, ANN is incorporated with '113' of monitored field results to predict surface settlement for a tunnel site. To achieve this, a standard format (a protocol) for a database of monitored field data is first proposed and then used for sorting out a variety of monitored data sets available in KICT. Using the capabilities of pattern recognition and memorisation of the ANN, the attempt is made to capture the rich physical characteristics smeared in the database and at the same time filter inherent noise in the monitored data. This module uses a concept of relative strength of effects (RSE) (Yang 1997) in sensitivity analysis for various major factors affecting the surface settlement in tunneling. There are some examples that the RSE rationally enables us to recognise the most significant factors of all the contributing factors. Two verification examples are undertaken with the trained ANN using the database created in this study. It is shown from the examples that the ANN has adequately recognised the characteristics of the monitored data sets retaining a generality for further prediction. The capabilities of this system in making accurate predictions are entirely dependent on the quality and the quantity of data used in training ANNs. If the data is deficient or the training is inadequate, the proposed neural network based prediction should be treated with caution. Therefore, the collection and analysis of monitored data should be carefully carried out for guaranteed predictions. A number of investigations are currently being undertaken in the authors' groups to apply the methodology to various critical engineering problems, especially on tunneling.

Figure 4. An example of pole and contour plots using equal area projection several functions such as statistical contouring of orientation clustering, pole plot and plane plot, etc. Using these functions the slope stability analysis is possible. Figure 4 is an example of pole and contour plots using equal area projection. In tunnels excavated in jointed rock masses, the most common failures are wedge failure falling from the roof and sliding out of the sidewalls. These wedges or tetrahedral-shaped blocks of rock, which are defined by the intersection of discontinuities with free face, may fall or slide from the surface. The module has a function of a 3-D stability analysis and visualisation for tunnel in rock mass using key block theory. It uses average dip and dip direction data of discontinuities that are determined by the former tool. It is also able to identify potential wedges that can slide or fall from the roof or both sides of wall. Calculation of the safety factor for these wedges is possible. With this module we can quickly create a 3-D model, calculate a safety factor and build up reinforcement during excavation . Figure 5 shows an example of key block analysis.

Figure 5. An example of key block analysis using the module.

Figure 8. Safety evaluation of tunnel face. Figure 6. 2-D settlement prediction with trough. The module was developed to predict the safety of tunnel and choose proper tunnel reinforcement system using fuzzy quantification theory and fuzzy inference rule based on tunnel information database. The system has two main parts. In first part, tunnel information input items are decided based on the tunnel face mapping information that can be easily obtained in-situ site. Tunnel information database consists of major face mapping items and tunnel safety effect factors such as adjacent building and overburden above tunnel, but in this system, only face mapping data expressed 4 levels was used. Main input data was made of uni-axial compressive strength, RQD, joint space, joint condition, joint direction and water inflow that are almost RMR classification factors. In second one, using fuzzy quantification theory, fuzzy membership function is composed and tunnel safety level is inferred through this membership function. The comparison result between the predicted reinforcement system level and measured ones was very similar. In-situ data were obtained in three tunnel sites including subway tunnel under Han River. This system will be very helpful to make the most of in-situ data and suggest proper applicability of tunnel reinforcement system escaping from the depend ence of some experienced experts for the absent of guide. The module is used with information of ground surface settlement together. And the D/B part is concluded in ITIS system. Figure 8 shows an example of the safety evaluation of tunnel face. 3.5 prediction module for face failure

Figure 7. 3-D settlement prediction with predicted damage class of adjacent building. The module is connected to the ITIS visualisation system and shows the predicted area as contours with trough according to a stage of excavation. Figure 6 is a window that shows 2-D settlement prediction with trough according to a stage of excavation in Seoul subway. The module also gives a prediction result of 3-D settlement with adjacent building continuously during excavation (Fig. 7). 3.4 Selection module for tunnel reinforcement (Kim 2001b) In tunnel construction it is important to predict or define the cause of the abnormal phenomenon and then to carry out the countermeasure against the major cause. For building up this module a lot of types of tunnel failure were surveyed and the detail causes were studied after many cases of tunnel failure were collected.

This module was built to predict the pattern and magnitude of tunnel collapses using Neural Network

collapse using the data obtained from 20 sites in Korea. Input data for predicting tunnel collapse pattern are 16 factors such as condition of shear zone, discontinuity conditions near by shear zone, geometry of joint with tunnel, water inflow, face condition, excavation method, the effect of configuration in earth, rock mass structure, rock mass classification, rock type, overburden, tunnel depth, tunnel geometry, RQD, joint condition and joint spacing. Figure 9 shows an example of learning procedure of the module. It takes about several times depending on the learning type. The system is proven to be capable of the effective estimation of tunnel collapse types and magnitudes. However the effluent data of tunnel collapse should be accumulated in the system to infer the results with a reasonable accuracy. Figure 10 shows the inference result of tunnel collapse type and magnitude.

Figure 9. Learning procedure of the module. system. The module uses tunnel face mapping and

Figure 10. The result of inference that shows the predicted collapse type and magnitude. other geological information that are stored in the 3.6 VR for visualisation of tunnel construction D/B of ITIS. VR technology is providing engineers and geologist with new and interactive ways to model The inference-learning module uses the backthe design and construction of tunnel. The virtual propagation algorithm that is based on the environment, in which the visualisation of the design generalised delta rule by Rumelhart et al. (1986). It and construction progress is effected, shows to be is the most popular learning procedure for multiideal for simulating geological features, the layer neural networks among more than 50 neural predicted results for ground surface settlement and network models. face failure. Neural Network for the module is consisted of two main modules; inference module for tunnel collapse pattern and inference-learning module for tunnel The system prototype has been tested on two tunnel construction sites. Figures 11-13 shows examples of the prototype systems outputs. Figures

11 and 12 shows the examples of the tunnel that is constructed in the mountain area and Figure 13 in the subway in the Seoul.

4 CONCLUSIONS The authors have developed Intelligent Tunnelling Information System (called ITIS) Prototype that is an integrated expert system including visualisation, evaluation and prediction technology of tunnel survey, design, construction and reinforcement system based on D/B, Artificial Intelligence, 3-D visualisation and Virtual Reality technologies. It will be benefit for engineers who do not have sufficient knowledge or experience in tunnel construction. The system will be extended continuously with several additional modules as a National Research Laboratory (NRL) Project in Korea. ACKNOWLEDGEMENT

Figure 11. VR screen shots of the portal in tunnel This research has been performed as a part of the National Research Laboratory project sponsored by K M O S T (K o r e a M i n i s t r y o f S c i e n c e & Technology). The authors greatly appreciate their financial support. REFERENCES Hong, S.W., Bae, G.J., Seo, Y.S., Kim, C.Y., L e e , S . W . & L e e , K . H ., 2 0 0 1a. The development of geotechnical interpretation system for tunneling . AITES-ITA 2001, 715721.. Hong, S.W., Bae, G.J., Seo, Y.S., Kim, C.Y., L e e , S . W . & L e e , K . H ., 2 0 0 1b , 3-D simulation analysis for visualizing geotechnical site investigation, Proc. 15th Int. Conf. Soil Mech. and Geotech. Engng., Istanbul, 27-31 Aug. 2001. Kim, C.Y., Bae, G.J., Hong, S.W., Park, C.H., Moon, H.K. & Shin, H.S., 2001a, Neural network based prediction of ground surface settlements due to tunnelling, Computers and Geotechnics, 28, 517-547. Kim, C.Y., Park, C.H., Bae, G.J., Hong, S.W., Oh, M.Y., 2001b, Countermeasure Expert System for Tunneling Failure, AITES-ITA 2001, 409-416 Murayama, S. & Matsuoka, H., 1971, Earth press on tunnels in sandy ground, Transactions, JSCE, 3, 78. Rumelhart, D.E., MacClelland, J.L. & PDP Research Group, 1986, Parallel distributed processing, Cambridge, MA: Exploitation in the MIT Press. Sagaseta, C., 1998, Surface settlements due to deformation of a tunnel in an elastic half

Figure 12. VR screen shots of wedge failure.

Figure 13. VR screen shots of the predicted results for ground surface settlement. The system will be extended to the main tool for visualisation and also linked up with the D/B and other modules in ITIS.

plane-discussion, Geotechnique, 48, 709713. Yang, Y. and Zhang, Q., 1 9 9 7 , A hierarchical analysis for rock engineering using artificial neural networks, Rock Mech. Rock Engng, 30, 207-222.