IT-based-tunnelling-risk-management-system-(IT-TURISK)-–-Development-and-implementation_2006_Tunnelling-and-Underground-Space-Technology.pdf

it based tunnelling risk management system it turisk development and implementation chungsik yoo a young woo jeonb byoung suk choic a department of civil and environmental engineering sungkyunkwan university 300 chun chun dong jan an gu suwon kyong gi do 440 746 republic of korea b esco engineers received in revised form 19 march 2005 accepted 25 may 2005 available online 10 august 2005 abstract this paper presents the development and implementation of an information technology it based tunnelling risk assessment system it turisk it turisk has been developed in a geographic information system gis environment with a capability of performing fi rst order assessment of tunnelling induced third party impact on surrounding environment within the framework of artifi cial intelligence ai technique the main emphasis in this study was placed in developing a gis framework that can be used in tunnelling risk assessment the developed system permits users to identify potential risk areas by performing computationally intensive tunnelling induced third party impact assessment that takes full advantage of the gis and the ai technologies the devel oped system was successfully applied to a seoul metro subway extension project this paper describes the concept and details of the it turisk development and implementation 2005 elsevier ltd all rights reserved keywords tunnelling information technology geographic information system artifi cial intelligence ground movement groundwater risk assessment 1 introduction in response to rapid growth in urban development there has been a pressing need for construction of new tunnels for transportation systems and underground utilities tunnelling has become an attractive alternative in creating underground spaces for transportation and utility networks an ever increasing awareness of the environment has also favored tunnelling over other con struction alternatives urban tunnelling however inev itably causes a variety of third party impact and therefore a substantial portion of the cost of a tunnelling project in the urban environments is being devoted to minimize the third party impact arising from the tunnel ling activity important information on the third party impact associated with a proposed tunnelling includes inter alia the magnitude and extent of tunnelling induced ground movements and their eff ects on adjacent build ings and utilities tunnel construction beneath ground water table will inevitably cause water infl ow into the tunnel and the consequent drawdown of groundwater level in the surrounding aquifer the related ground subsidence occurring as a result of the reduction in pore water pressures in permeable soil rock layers can dam age nearby structures utilities on account of the in creased public awareness on the impact of tunnelling onsurroundingenvironments thecontrolofthe groundwater has become an essential part of the plan ning design and construction of a tunnelling project 0886 7798 see front matter 2005 elsevier ltd all rights reserved doi 10 1016 j tust 2005 05 002 corresponding author tel 82 31 290 7518 fax 82 31 290 7549 e mail address csyoo skku edu c yoo tunnelling and underground space technology 21 2006 190 202 tunnelling and underground space technology incorporating trenchless technology research in response to increased public awareness on the tun nelling induced third party impact several attempts have been made to develop and implement computer ized settlement risk assessment systems for urban tun nellingworks recentlydevelopedtunnellingrisk assessment systems that fall into this category include turisk yoo and kim 2003 and gms netzel and kallberg 1999 turisk is a web based fi rst order settlement risk analysis system which was implemented to a daegu metro subway line 2 construction site in korea the main advantage of turisk is its ability to make a complete tunnelling induced settlement risk analysis including building utility damage assessment based on the currently available conventional methods of analysis in a web environment gms is a gis based settlement risk analysis system exclusively used in the north south metro line construction in amsterdam the settlement prediction and damage assessment pro cedures implemented in this system are rather simple and conservative in developing a better and effi cient platform that can be used within the framework of tunnelling risk assess ment an attempt has been made by the fi rst author to introduce information technology it such as gis and ai into the previously developed system turisk the ultimate goal of this eff ort is aimed at providing users with fully interactive gis and ai interfaces within the context of tunnelling risk assessment such an it based system allows added levels of complexity that can not be accounted for in the conventional approach more importantly the gis platform allows users to modify and update input parameters when new data be come available during construction through fi eld moni toring to further increase the accuracy of the assessment the computationally intensive gis structure adopted in it turisk allows one to perform the tunnelling risk assessment in a more effi cient way than the conventional approach this paper presents the details of develop ment of it turisk and its implementation in an ur ban tunnelling project 2 it based risk assessment 2 1 geographic information system gis the advent of gis has opened a window of new tech nological possibilities for geospatial data storage and analysis these possibilities have shown much potential in various fi elds of civil engineering where large amount of geographical data are needed stored and manipu lated the geographical data are stored in a gis system as a geodatabase in which relationships between individ ual geographic features can be permanently stored the geodatabase has proven to be a robust structure for cat aloging and storing data the gis technology is well suited for tunnelling risk assessment which manages voluminous geographical information as demonstrated by netzel and kallberg 1999 the gis has been proven to be an eff ective tool for storage retrieval and display of data but its analytical capabilities are only occasionally tapped the gis how ever can be used as an excellent analysis tool when interfaced with modules that calculate desired quanti ties a recent excellent example is the application of gis in the landslide hazard assessment as demonstrated by xie et al 2003 in which a gis grid based 3d deter ministic slope stability analysis model was developed the concept of gis based risk hazard assessment can also be well fi tted in tunnelling works for example provided with a digital site map and proper mathemat ical functions the ground movements at specifi ed loca tionscanbecomputed stored retrieved and displayed the results can then be used for damage assessment of buildings utilities within the area of infl u ence furthermore by defi ning spatial distributions of other quantities of interest once computed as layers within a gis platform the spatial distributions of the impact of tunnelling can be easily visualized 2 2 artifi cial intelligence ai based tunnelling impact assessment experiences from a number of case histories have shown that the ai technique can be used to identify and determine certain dependencies between variables in various engineering fi elds among the various artifi cial intelligence techniques the artifi cial neural network ann is increasingly becoming a standard ai tech nique for solving civil engineering related problems artifi cial neural networks are mathematical models in spired by the biological structure and functioning of the brain the major advantage of ann is its ability to learn recall and generalize from training data by assigning or adjusting the connection weights wkj fig 1 once trained with proper data an ann can successfully describe relationships between variables that might otherwise be diffi cult when using mathemat ical terms a summary of geomechanical applications of anns can be found in the paper by atto okine 1998 theoretical background of the ann is beyond the scope of this paper and can be found in demuth and beale 1997 the risk assessment of a proposed tunnelling requires the prediction of magnitude and extent of third party impact associated with the tunnelling existing mathe matical expressions for the required assessment how ever havelimitationswhenappliedinpractical tunnelling situations on account of the limitations of the existing mathematical expressions that will be ad dressed later in this paper attempts were made to incor porate the ann technique in estimating required c yoo et al tunnelling and underground space technology 21 2006 190 202191 quantities for the tunnelling induced risk assessment first step toward developing an ann based tunnelling risk assessment procedure was to develop back propaga tion neural network modules that can be used in the gis environment a built in programming tool visual basic for application vba was used to write ann modules that are able to compute the required quantities such as tunnelling induced ground movements groundwater in fl ow rate and magnitude of drawdown the back prop agation anns adopted in this study essentially followed the formulations of eberhart and dobbins 1990 de tails of the training and implementation procedures for the anns are described in subsequent sections 3 strategy for third party impact assessment 3 1 general in view of tunnelling risk assessment it is of prime interest to identify areas being at risk in terms of the third part impact for a given tunnel design so that nec essary modifi cation to the original design can be made to meet preset requirements it is therefore essential for any tunnelling risk assessment system of this kind to be equipped with modules that can compute the quantities related to the third party impact in this study a simplifi ed but conservative strategy was adopted desired quantities were computed for pre selected transverse sections perpendicular to the tunnel drive spaced at a rather close interval of 5 m along the entire tunnel route either the available closed form solutions or the anns were used the computed quantities are thus relevant to an end of construction eoc condition this approach is justifi ed since in most of the urban tunnelling cases buildings are usually set back from the tunnel alignment to some degree thus the transverse settlement profi le tends to exert more damage to the buildings than the progressive longitudi nal profi le for cases where a building is located close to or directly above the tunnel centerline the reverse is true as the sequential tunnel construction process causes a temporary wave of disturbances as the tunnel face ad vances fig 2graphicallyillustratesthestrategy adopted in the third party impact assessment process the following sections describe the detailed strategies adopted in the third part impact assessment data error feed forward back propagation input layer output layer hidden layer wkjwkj fig 1 typical structure of ann fig 2 strategy for third party impact assessment 192c yoo et al tunnelling and underground space technology 21 2006 190 202 3 2 ground movements and building utility damage the error function approaches proposed by peck 1969 o reilly and new 1982 and extended by new and o reilly 1991 were adopted for the prediction of the magnitude and extent of ground movements associ ated with a proposed tunnelling for building and utility damage assessment the procedures given by mair et al 1996 bracegirdle et al 1996 respectively were used these methods are computationally simple can be easily implemented in a gis platform and are good fi rst order approximations the necessary equations were trans lated into arcgis modelling environment with the help of vba in the building damage assessment a building is treated as an idealized deep beam with span l and height h deforming under a central point load bur land and wroth 1974 bending strain eb and diago nal shear strain ed developed in a building subjected to tunnelling induced ground movements are calculated using eqs 1 and 2 respectively these quantities are computed for sections perpendicular to the tunnel drive to evaluate possible damage due to the transverse ground movements after complete passage of the tun nel face d l l 12t 3ie 2tlhg eb 1 d l 1 hl2g 18ie ed 2 in eqs 1 and 2 d is a maximum defl ection of the building h and l are respectively the height and span of the building but limited by any point of infl ection or extent of settlement trough e and g are respectively young s modulus and shear modulus of the building i is the second moment of inertia of the equivalent beam and t is the furthest distance from the neutral axis to the edge of the beam i e h 2 in the sagging zone and h in the hogging zone for the utility damage assessment on the other hand either the transverse or longitudinal ground movement profi les based on the normal probability equation atte well et al 1986 are used depending on the utility orien tation relative to the tunnel alignment based the procedures described by bracegirdle et al 1996 and ex tended by yoo and kim 2003 the normal probability function given as eq 3 that estimates the magnitude and extent of the surface settle ment requires the volume of the settlement trough per unit distance of tunnel advance vs and the infl ection point i both of which are in essence primary variables of interest sv vs ffi ffi ffi ffi ffiffi 2p p i exp y2 2i2 3 where y is the cartesian coordinate of any point along the settlement trough with y 0 at the tunnel centerline the value of vsin eq 3 can either be directly input by the user or automatically estimated by the trained ann associated ground slopes and strains are then subse quently computed using the respective equations that can be found in yoo and kim 2003 the ann was trained using a set of training data obtained from a series of fi nite diff erence analyses using flac itasca 2000 on tunnelling situations frequently encountered in seoul metro tunnel construc tion including geology and support patterns the re sultswerethenusedtoestablishrelationships between the tunnelling condition and the characteristics of the transverse settlement trough i e vsand i using the ann the required input parameters include sup port pattern cover depth stratifi cation and ground type through which the tunnel is excavated three hid den layers were required to accurately reproduce target output values for the training data the performance of the trained ann is illustrated in fig 3 the ann successfully predicted the relationships between the training data and target output values although there exist some discrepancies between the predicted and target values these errors can be considered of insignifi cance from the practical tunnel engineering point of view details of the ann architectures em ployed and training procedures are available in yoo et al 2003 3 3 groundwater infl ow rate and drawdown information important to the impact of tunnelling on surrounding environment are those related to the groundwater i e the water infl ow rate q and the mag nitude of drawdown hd the rate of water infl ow q can be estimated using eq 4 proposed by lei 1999 q 2pksh ln h r ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi h r 2 1 q 4 where q is the water infl ow m3 m s ksis the hydraulic conductivity of ground around tunnel m s h is the dis tance from tunnel to equipotential m and r is the tun nel radius m although eq 4 works well for homogeneous ground conditions it does do not take into account any stratifi cation of the geology further more no mathematical expression is available for esti mating the magnitude of drawdown associated with a proposed tunnelling recognizing these limitations the ann was used as an alternative tool for estimating the relevant quantities related to the impact of a pro posed tunnelling on the groundwater similar to the development of the ann for the ground movement prediction the training data for c yoo et al tunnelling and underground space technology 21 2006 190 202193 the ann were obtained from a series of seepage anal ysis using flac on tunnelling situations frequently encounteredinseoulmetrotunnelconstruction including the stratifi cation of geology and the ground water level the results were subsequently used to establish relationships between a given tunnelling con dition as well as q and hd the required input param eters include location of groundwater table cover depth permeabilities of ground layers and tunnel diameter two hidden layers were required to achieve the desired accuracy of the ann the performance of the trained ann is shown in fig 4 in which good correlations are observed between the predicted and the target values it should be noted that the ann pre dicts q and hdfor a section perpendicular to the tun nel drive after a steady state fl ow condition has been established details of the ann architecture employed and training procedures are available elsewhere yoo et al 2003 4 development of it turiks 4 1 platform and system overview arcgis esri 2003 was used as a primary platform for it turisk arcgis is a desktop gis and mapping software which provides data visualization query anal ysis and integration capabilities along with the ability to create and edit geographic data arcgis consists a num ber of desktop applications including arcmap arccat alog and arctoolbox arcmap provides data display query and analysis arccatalog provides geographic and tabular data management creation and organiza tion arctoolbox provides basic data conversion a distinct feature of it turisk is its ability to cus tomize database for a given tunnel construction site and to carry out computationally intensive tunnelling in duced third party impact assessment it turisk con sists of two independent but fully interfaced modules 01020 30 40 targe