Risk analysis during tunnel construction using Bayesian NetworksPorto Mctro case study.docx

1、Risk analysis during tunnel construction using Bayesian NetworksPorto Mctro case studyTunnelling and Underground Space Technology 27 (2012) 86100Contents lists available at SciVerse ScienceDirectTunnelling and Underground Space Technologyjo urn a l h o m ep ag e: www. e l s evi er .co m /l oca t e/t

2、u st Risk analysis during tunnel construction using Bayesian Networks: Porto Metro case studyRita L. Sousa, Herbert H. Einstein Dept. of Civil and Environmental Engineering, Massachusetts Institute of Technology, Cambridge, USAa r t i c l e i n f o Article history:Received 16 December 2010Received i

3、n revised form 15 July 2011Accepted 17 July 2011Available online 27 August 2011Keywords: Risk TunnelingBayesian Networksa b s t r a c t This paper presents a methodology to systematically assess and manage the risks associated with tunnel construction. The methodology consists of combining a geologi

4、c prediction model that allows one to pre- dict geology ahead of the tunnel construction, with a construction strategy decision model that allows one to choose amongst different construction strategies the one that leads to minimum risk. This model used tunnel boring machine performance data to rela

5、te to and predict geology. Both models are based on Bayesian Networks because of their ability to combine domain knowledge with data, encode dependen- cies among variables, and their ability to learn causal relationships. The combined geologic prediction construction strategy decision model was appl

6、ied to a case, the Porto Metro, in Portugal. The results of the geologic prediction model were in good agreement with the observed geology, and the results of the construction strategy decision support model were in good agreement with the construction methods used. Very signicant is the ability of

7、the model to predict changes in geology and consequently required changes in construction strategy. This risk assessment methodology provides a powerful tool with which planners and engineers can systematically assess and mitigate the inherent risks associated with tunnelconstruction. 2011 Elsevier

8、Ltd. All rights reserved.1. IntroductionThere is an intrinsic risk associated with tunnel construction because of the limited a priori knowledge of the existing subsur- face conditions. Although the majority of tunnel construction pro- jects have been completed safely there have been several inciden

9、ts in various tunneling projects that have resulted in delays, cost overruns, and in a few cases more signicant consequences such as injury and loss of life. It is therefore important to systematically assess and manage the risks associated with tunnel construction. A detailed database of accidents

10、that occurred during tunnel con- struction was created by Sousa (2010). The database contains204 cases all around the world with different construction meth- ods and different types of accidents. The accident cases were obtained from the technical literature, newspapers and correspon- dence with exp

11、erts in the tunneling domain.Knowledge representation systems (or knowledge based sys- tems) and decision analysis techniques were both developed to facilitate and improve the decision making process. Knowledge representation systems use various computational techniques of AI (articial intelligence)

12、 for representation of human knowledge Corresponding author. Address: 70 Massachusetts Ave., Room 1-342, Cam- bridge MA 02139, USA. Tel.: +1 617 253 3598; fax: +1 617 253 6044.E-mail address: einsteinmit.edu (H.H. Einstein).and inference. Decision analysis uses decision theory principles supplemente

13、d by judgment psychology (Henrion, 1991). Both emerged from research done in the 1940s regarding development of techniques for problem solving and decision making. John von Neumann and Oscar Morgensten, who introduced game theory inGames and Economic Behavior (1944), had a tremendous impact on resea

14、rch in decision theory.Although the two elds have common roots, since then they have taken different paths. More recently there has been a resur- gence of interest by many AI researchers in the application of prob- ability theory, decision theory and analysis to several problems in AI, resulting in

15、the development of Bayesian Networks and inu- ence diagrams, an extension of Bayesian Networks designed to include decision variables and utilities. The 1960s saw the emer- gence of decision analysis with the use of subjective expected util- ity and Bayesian statistics. Howard Raiffa, Robert Schlaif

16、er, and John Pratt at Harvard, and Ronald Howard at Stanford emerged as leaders in these areas. For instance Raiffa and Schlaifers Applied Statistical Decision Theory (1961) provided a detailed mathemati- cal treatment of decision analysis focusing primarily on Bayesian statistical models. Pratt et

17、al. (1964) developed basic decision anal- ysis. while Eskesen et al. (2004) and Hartford and Baecher (2004) provide good summaries on the different techniques (fault trees, decision trees, etc.) that can be used to assess and manage risk in tunneling.0886-7798/$ - see front matter 2011 Elsevier Ltd.

18、 All rights reserved. doi:10.1016/j.tust.2011.07.003Various commercial and research software for risk analysis dur- ing tunnel construction have been developed over the years, the most important of which is the DAT (Decision Aids for Tunneling), developed at MIT in collaboration with EPFL (Ecole Pol

19、ytechnique Fdrale de Lausanne). The DAT are based on an interactive pro- gram that uses probabilistic modeling of the construction process to analyze the effects of geotechnical uncertainties and construc- tion uncertainties on construction costs and time. (Dudt et al.,2000; Einstein, 2002) However,

20、 the majority of existing risk analy- sis systems, including the DAT, deal only with the effects of ran- dom (common) geological and construction uncertainties on time and cost of construction. There are other sources of risks, not considered in these systems, which are related to specic geo- techni

21、cal scenarios that can have substantial consequences on the tunnel process, even if their probability of occurrence is low.This paper attempts to address the issue of specic geotechnical risk by rst developing a methodology that allows one to identify major sources of geotechnical risks, even those

22、with low probabil- ity, in the context of a particular project and then performing a quantitative risk analysis to identify the optimal construction strategies, where optimal refers to minimum risk. For that pur- pose a decision support system framework for determining theoptimal (minimum risk) cons

23、truction method for a given tunnelFig. 1. Bayesian Network example.the relations between variables. In this example the arrow fromC to B2 means that C has a direct inuence on B2.Specically, a Bayesian Network is a compact and graphical rep- resentation of a joint distribution, based on some simplify

24、ing assumptions that some variables are conditionally independent of others. As a result the joint probability of a Bayesian Network over the variables U = X1, . , Xn, represented by the chain rule can be simplied from:nYalignment was developed. The decision support system consists of two models: a

25、geologic prediction model, and a construction strat- egy decision model. Both models are based on the Bayesian Net-PU itoPXi j x1 ; . . ; xi 1 work technique, and when combined allow one to determine thePU Qn PX x j parents X , where parents (X ) is theoptimal tunnel construction strategies. The dec

26、ision model con-iparent set ofi i i iXi.tains an updating component, by including information from theexcavated tunnel sections. This system was implemented in a real tunnel project, the Porto Metro in Portugal.2. Background on Bayesian NetworksBayesian Networks are graphical representations of know

27、ledge for reasoning under uncertainty. They can be used at any stage of a risk analysis, and may substitute both fault trees and event trees in logical tree analysis. While common cause or more general depen- dency phenomena pose signicant complications in classical fault tree analysis, this is not

28、the case with Bayesian Networks. They are in fact designed to facilitate the modeling of such dependen- cies. Because of what has been stated, Bayesian Networks provide a good tool for decision analysis, including prior analysis, posterior analysis and pre-posterior analysis. Furthermore, they can b

29、e ex- tended to inuence diagrams, including decision and utility nodes in order to explicitly model a decision problem.A Bayesian Network is a concise graphical representation of thejoint probability of the domain that is being represented by theIt is this property that makes Bayesian Networks a ver

30、y power- ful tool for representing domains under uncertainty, allowing one to store and compute the joint and marginal distributions more efciently.In order to obtain results from Bayesian Networks one does inference. This consists of computing answers to queries made to the Bayesian Network. The tw

31、o most common types of queries are: A priori probability distribution of a variablePA X . . X PX1 ; . . ; Xk ; A 1X1 Xkwhere A is the query-variable and X1 to Xk are the remaining variables of the network. This type of query can be used during the design phase of a tunnel for example to assess the proba- bility of failure under design conditions (geology, hydrology, etc.). Posterior distribution of variables given evidence(observations)PA; erandom variables, consisting of (Russel and Norvig, 2003):