道橋相關(guān)外文翻譯-- 隧道施工風(fēng)險(xiǎn)敏感型決策支持系統(tǒng)

1、<p><b>　　中文4560字</b></p><p>　　隧道施工風(fēng)險(xiǎn)敏感型決策支持系統(tǒng)</p><p>　　Veerasak Likhitruangsilp① 美國(guó)土木工程師學(xué)會(huì)準(zhǔn)會(huì)員</p><p>　　Photios G.Ioanou② 美國(guó)土木工程師學(xué)會(huì)會(huì)員</p><p><b>　　

2、摘要：</b></p><p>　　全面的和現(xiàn)實(shí)的隧道計(jì)劃必須爭(zhēng)取最優(yōu)決策，最大限度地減少時(shí)間和成本，同時(shí)解決的重要因素，如地質(zhì)不確定性和可變性，隧道生產(chǎn)力的不確定性，以及承包商的風(fēng)險(xiǎn)敏感性。本文提出了一種計(jì)算機(jī)化的決策支持系統(tǒng)，集成了所有重要的隧道風(fēng)險(xiǎn)。它由三個(gè)相互關(guān)聯(lián)的模型：地質(zhì)概率預(yù)測(cè)模型，概率隧道造價(jià)估算，風(fēng)險(xiǎn)敏感型動(dòng)態(tài)決策模型。概率地質(zhì)預(yù)測(cè)模型使用所有可用的地質(zhì)資料，從地面級(jí)躍遷的概率中去表

3、征地質(zhì)不確定性和可變性沿隧道輪廓。概率隧道成本估算模型評(píng)估隧穿時(shí)間和成本，表現(xiàn)為在利用蒙特卡羅模擬實(shí)際施工操作下，把不同的開挖和支護(hù)方法應(yīng)用到不同的地面條件。這兩個(gè)模型提供了對(duì)風(fēng)險(xiǎn)敏感的動(dòng)態(tài)決策模型這一系統(tǒng)核心的主要輸入，作為可用項(xiàng)目信息和承包商的風(fēng)險(xiǎn)敏感度職能，以此來(lái)確定最優(yōu)開挖與支護(hù)順序和相應(yīng)的風(fēng)險(xiǎn)調(diào)整后的隧道項(xiàng)目成本。該系統(tǒng)的一個(gè)實(shí)際公路隧道工程中的應(yīng)用說(shuō)明了該方法的兩個(gè)建模能力的量化，并納入風(fēng)險(xiǎn)，及其具有的對(duì)做出最佳決策的承包商

4、的風(fēng)險(xiǎn)敏感性程度的有效性職能。</p><p>　?、?，博士，講師，土木工程系，朱拉隆功大學(xué)，曼谷10330，泰國(guó)，fcevlk@eng.chula.ac.th</p><p>　?、冢┦浚淌冢聊九c環(huán)境工程系，密歇根大學(xué)安阿伯，MI48109-2125，USA，photios@umich.edu</p><p><b>　　交通工程巖土工程</

5、b></p><p><b>　　引言</b></p><p>　　隧道為現(xiàn)代化交通運(yùn)輸體系的重要選擇，因?yàn)樗麄兪俏ㄒ坏倪\(yùn)輸系統(tǒng)，可以解決困難的地形，有限的地表空間和運(yùn)輸需求增加的問(wèn)題。與此同時(shí)，隧道是昂貴的地下結(jié)構(gòu)，其中的各種風(fēng)險(xiǎn)在項(xiàng)目交付過(guò)程中的每個(gè)階段都遇到過(guò)。全面和現(xiàn)實(shí)的隧道計(jì)劃必須爭(zhēng)取最優(yōu)決策，最大限度地減少時(shí)間并且同時(shí)解決重要的隧道風(fēng)險(xiǎn)成本。為此，風(fēng)

6、險(xiǎn)敏感型決策支持系統(tǒng)已發(fā)展到量化所有重要的隧道風(fēng)險(xiǎn)，并確定最佳的隧道計(jì)劃和項(xiàng)目的風(fēng)險(xiǎn)調(diào)整后的成本（Likhitruangsilp2003）</p><p><b>　　隧道風(fēng)險(xiǎn)</b></p><p>　　在隧道工程中最重要的決定之一是沿隧道縱斷面優(yōu)化序列確定開挖方法和支持系統(tǒng)。這些決定都是由四個(gè)主要因素來(lái)表征：地質(zhì)的不確定性，地質(zhì)變化性，隧道生產(chǎn)力不確定性，風(fēng)險(xiǎn)敏感

7、性。</p><p><b>　　地質(zhì)不確定性</b></p><p>　　隧道項(xiàng)目的選擇方法主要取決于隧道的預(yù)期的地質(zhì)條件，這是重要的巖體性質(zhì)如巖石類型和不連續(xù)狀態(tài)的集合。無(wú)論采取地下勘探的數(shù)量和程度，隧道地質(zhì)在開始之前不能稱為完美施工。雖然有多個(gè)隧道已采取減輕地質(zhì)不確定性的做法（例如，觀測(cè)方法），但他們不能完全消除這種隧道建設(shè)規(guī)劃的不確定性。</p>

8、<p><b>　　地質(zhì)變化性</b></p><p>　　大部分隧道穿越各種地質(zhì)條件，其中的位置和程度是不可能事先定義的。對(duì)于大多數(shù)具有明顯的地質(zhì)變化的隧道工程，掘進(jìn)方法的選定，必須適合于所有預(yù)期的地質(zhì)條件且不嚴(yán)重中斷開挖進(jìn)度。這些方法包括適應(yīng)隧道開挖方法的修改（例如，臺(tái)階和多個(gè)漂移），圓長(zhǎng)，鉆模式，并詳細(xì)介紹了支護(hù)。因此，隧道工程決策在本質(zhì)上是動(dòng)態(tài)的。</p>

9、<p><b>　　隧道生產(chǎn)力不確定性</b></p><p>　　隧道工程決策的另一個(gè)風(fēng)險(xiǎn)由隧道施工過(guò)程中生產(chǎn)效率的的不確定性導(dǎo)致。這種不確定性來(lái)源于施工設(shè)備的性能變化，工作輸出的變化，及意外事件等建設(shè)工程中的事故。這種不確定性是存在的，即使地質(zhì)條件已知的情況下。因此，其對(duì)隧道工程決策的影響必須明確解決。</p><p><b>　　風(fēng)險(xiǎn)敏感性&

10、lt;/b></p><p>　　福利和涉及的風(fēng)險(xiǎn)（例如，隧道的決策）的決策成本，個(gè)別估值往往是非線性的，因?yàn)檫@些決定不是基于預(yù)期的貨幣價(jià)值的最大化。換句話說(shuō)，在不確定性情況下決策時(shí)，決策者對(duì)風(fēng)險(xiǎn)是典型的敏感的，無(wú)論是風(fēng)險(xiǎn)規(guī)避或風(fēng)險(xiǎn)偏好。一個(gè)人的風(fēng)險(xiǎn)敏感度（風(fēng)險(xiǎn)偏好）是由幾個(gè)因素影響的，特別是個(gè)人目前的凈資產(chǎn)狀況。通常情況下，當(dāng)一個(gè)人的凈資產(chǎn)增長(zhǎng)，對(duì)相同的風(fēng)險(xiǎn)會(huì)有較低的風(fēng)險(xiǎn)規(guī)避行為。</p>

11、<p>　　承包商的風(fēng)險(xiǎn)規(guī)避和風(fēng)險(xiǎn)暴露程度可以對(duì)建設(shè)決策和風(fēng)險(xiǎn)溢價(jià)或嵌入在用來(lái)承擔(dān)這項(xiàng)工作的承包商價(jià)格中的應(yīng)急必要量，其有主要影響。更厭惡風(fēng)險(xiǎn)的承包商采用了一種更加保守的計(jì)劃，其中包括在他的出價(jià)中更高的津貼比少規(guī)避風(fēng)險(xiǎn)的承包商給的高（伊奧努1988）。因此，有必要將風(fēng)險(xiǎn)敏感度納入隧道的決策。</p><p>　　經(jīng)考慮上述所有因素，隧道的決策可以被認(rèn)為是一項(xiàng)風(fēng)險(xiǎn)敏感的動(dòng)態(tài)概率決策的過(guò)程，它可以是結(jié)構(gòu)的風(fēng)

12、險(xiǎn)敏感型決策支持系統(tǒng)（2003 Likhitruangsilp）。</p><p>　　風(fēng)險(xiǎn)敏感型決策支持系統(tǒng)</p><p>　　風(fēng)險(xiǎn)敏感型決策支持系統(tǒng)由三個(gè)相互關(guān)聯(lián)的模型組成：地質(zhì)概率預(yù)測(cè)模型，概率隧道造價(jià)估算，風(fēng)險(xiǎn)敏感型動(dòng)態(tài)決策模型。</p><p><b>　　地質(zhì)概率預(yù)測(cè)模型</b></p><p>　　概率地

13、質(zhì)預(yù)報(bào)模型使用所有可用的地質(zhì)資料，以地面類過(guò)渡概率的形式來(lái)描述沿隧道輪廓的地質(zhì)不確定性和可變性。該模型是基于離散狀態(tài)的，是重要的地質(zhì)參數(shù)的連續(xù)空間馬爾可夫過(guò)程（例如，巖石斷裂）。這些地質(zhì)馬爾可夫模型是從區(qū)域數(shù)據(jù)（如地質(zhì)圖）中創(chuàng)建和使用直接評(píng)估或貝葉斯更新（伊奧努1984）通過(guò)特定位置的數(shù)據(jù)（例如，鉆孔測(cè)試）來(lái)更新。</p><p>　　該模型已被編程在MATLAB中。其輸入包括隧道的長(zhǎng)度，每一個(gè)階段的程度（例如，

14、圓長(zhǎng)度），地質(zhì)參數(shù)和它們的狀態(tài)，和地面類的定義?；谠撦斎?，模型計(jì)算出地質(zhì)參數(shù)和地面類在沿隧道的不同位置后的狀態(tài)的概率。這兩種狀態(tài)概率隨后由應(yīng)用復(fù)合地基級(jí)躍遷（2003年Likhitruangsilp）的概念，以確定隧道地質(zhì)地類轉(zhuǎn)移概率矩陣。由此產(chǎn)生的轉(zhuǎn)移概率矩陣成為風(fēng)險(xiǎn)敏感的動(dòng)態(tài)決策模型的輸入。</p><p><b>　　概率隧道成本模型</b></p><p>

15、　　概率隧道造價(jià)估算模型執(zhí)行性價(jià)比隨機(jī)評(píng)價(jià)，包括隧穿時(shí)間和開挖與支護(hù)方法，與不同的地面類（隧道方案）的不同組合，。該模型包括成本估算子模型和概率調(diào)度子模型。</p><p>　　成本估算子模型，是在計(jì)算機(jī)中的電子表格中創(chuàng)建的，其進(jìn)行組織隧道成本項(xiàng)目，進(jìn)行量化起飛的計(jì)算，并計(jì)算固定費(fèi)用和與每個(gè)選項(xiàng)相關(guān)的可變成本。除了正常掘進(jìn)成本，還考慮施工過(guò)程中選擇了錯(cuò)誤的開挖方法的風(fēng)險(xiǎn)。其輸入包括一個(gè)工作分解結(jié)構(gòu)（WBS），是專

16、門為隧道工程設(shè)計(jì)的，輸入開挖方法和支持系統(tǒng)規(guī)格，所有隧道作業(yè)的船員組成，以及材料，設(shè)備和勞動(dòng)力成本數(shù)據(jù)。成本估算子模型的最終輸出是由固定的成本和不同的選擇可變成本混合而來(lái)。可變成本提供概率調(diào)度子模型的輸入，包括每小時(shí)的成本（$/小時(shí)）和每通道長(zhǎng)度（元/平方米）材料單位成本。</p><p>　　概率調(diào)度子模型通過(guò)ProbSched，一種概率調(diào)度模擬程序（伊奧努和1998年馬丁內(nèi)斯）來(lái)實(shí)現(xiàn)。其對(duì)隧道不同方案的成本和

17、時(shí)間性能進(jìn)行評(píng)估。除了成本估算子模型的輸入，它需要為不同的隧道備選方案業(yè)務(wù)提供優(yōu)先級(jí)的網(wǎng)絡(luò);網(wǎng)絡(luò)活動(dòng)的時(shí)間方程中的活動(dòng);時(shí)間方程的參數(shù)，無(wú)論是確定性地定義或主觀評(píng)估;和用于計(jì)算單元的隧道成本（元/平方米）的公式。 概率調(diào)度網(wǎng)絡(luò)使用蒙特卡羅模擬分析。最終成果包括對(duì)不同隧道替代品單位成本的分布，這對(duì)于風(fēng)險(xiǎn)敏感的動(dòng)態(tài)決策模型提供了投入。該模型的詳細(xì)描述可以在<<Likhitruangsilp和伊奧努（2003）>>中找

18、到。</p><p>　　風(fēng)險(xiǎn)敏感型動(dòng)態(tài)決策模型</p><p>　　風(fēng)險(xiǎn)敏感型動(dòng)態(tài)決策模型，提出系統(tǒng)的核心，被配制成風(fēng)險(xiǎn)敏感的隨機(jī)動(dòng)態(tài)規(guī)劃模型。它的輸入包括地面類轉(zhuǎn)移概率矩陣，是針對(duì)每個(gè)階段的隧道的，是通過(guò)由概率隧道成本估算模型模擬出不同方案的隧道單位成本分布所確定的。該模型還需要決策者的（例如，承包商）風(fēng)險(xiǎn)厭惡系數(shù)（γ），它是用于風(fēng)險(xiǎn)偏好的決策者的程度進(jìn)行編碼的指數(shù)效用函數(shù)的參數(shù)。正γ

19、意味著決策者是風(fēng)險(xiǎn)厭惡者，而負(fù)數(shù)γ意味著決策者是風(fēng)險(xiǎn)偏好者。</p><p>　　這種被編程在MATLAB中的風(fēng)險(xiǎn)敏感型動(dòng)態(tài)決策模型，進(jìn)行決策和風(fēng)險(xiǎn)分析來(lái)確定最佳的隧道策略和項(xiàng)目的風(fēng)險(xiǎn)調(diào)整隧道費(fèi)，這兩者都是可用信息和決策者風(fēng)險(xiǎn)敏感程度的功能。</p><p><b>　　應(yīng)用</b></p><p>　　懸湖隧道，在科羅拉多州的高速公路隧道項(xiàng)目

20、，被用來(lái)證明了該系統(tǒng)的應(yīng)用。這種巖石隧道項(xiàng)目涉及一對(duì)雙車道公路隧道的建設(shè)：東行及西行隧道。在這里，我們專注于由多個(gè)漂移和爆破方法開挖西行隧道的一部分。</p><p>　　基于幾個(gè)巖體分類系統(tǒng)，地質(zhì)條件可分為三個(gè)地面類：GC1（最好），GC2（中），以及GC3（最差）。三種開挖方法（EM1，EM2，EM3）和初始支持系統(tǒng)（SS1，SS2，SS3）被設(shè)計(jì)來(lái)對(duì)應(yīng)于三個(gè)地類。例如，EM2和SS2是六個(gè)標(biāo)題（漂移）和巖石

21、加固系統(tǒng)，包括銷釘，spiles和噴漿，其中最經(jīng)濟(jì)和結(jié)構(gòu)充分挖掘的，如圖1。地類分類和開挖及支護(hù)方法的規(guī)格說(shuō)明可以在Scotese和阿克曼（1992），和埃塞克斯郡等（1993）找到。因此，有九個(gè)可能的隧道方案（即3開挖與支護(hù)方法×3地面班）。例如，替代（EM2，GC3）代表使用EM2為特定的圓的決定，和爆破后當(dāng)時(shí)的地面類是GC3（即，結(jié)構(gòu)不夠）。</p><p><b>　　地質(zhì)概率預(yù)測(cè)模型

22、</b></p><p>　　懸湖隧道的地質(zhì)概率預(yù)測(cè)模型是基于三個(gè)重要的地質(zhì)參數(shù)被開發(fā)的：巖石質(zhì)量指標(biāo)（RQD），斷裂頻率，風(fēng)化及蝕變。這些地質(zhì)參數(shù)狀態(tài)的組合被分為三個(gè)地類，分別對(duì)應(yīng)由埃塞克斯等人（1993）描述的分類。</p><p>　　每個(gè)地質(zhì)馬爾可夫模型的參數(shù)，通過(guò)分析從鉆孔（利茲，希爾和朱厄特，公司1981）的日志數(shù)據(jù)被估計(jì)。后驗(yàn)狀態(tài)概率的觀測(cè)點(diǎn)進(jìn)行編碼的主觀根據(jù)是地

23、質(zhì)專家不同的評(píng)估，包括利茲，希爾和朱厄特公司（1981），和Scotese和阿克曼（1992）。這些概率被用來(lái)確定在沿隧道3.7米（12英尺）的時(shí)間間隔后的狀態(tài)的概率的非觀測(cè)點(diǎn)。任何兩個(gè)階段之間的地面類轉(zhuǎn)移概率矩陣然后會(huì)基于復(fù)合地基級(jí)躍遷的概念而確定。模型輸出的一個(gè)例子是位置746.1米（2,448英尺），749.8米（2,460英尺）之間的接地類轉(zhuǎn)移概率矩陣：</p><p>　　例如，由于隧道地質(zhì)在地面1級(jí)7

24、46.1米的位置，這將使過(guò)渡到749.8米的位置的地面1級(jí)（保持不變），地級(jí)2，和地面3級(jí)，的概率分別為44.52，46.79，和8.96％，（例如，上面的矩陣的第一行）。</p><p><b>　　概率隧道造價(jià)估算</b></p><p>　　根據(jù)本項(xiàng)目采用的施工資源獲得的信息，成本估算子模型對(duì)設(shè)備，勞動(dòng)力，以及每個(gè)替代設(shè)備的成本進(jìn)行組織和計(jì)算。這些費(fèi)用則分為固定

25、成本和可變成本?？勺兂杀颈挥米鞲怕收{(diào)度子模型的輸入，進(jìn)行隧道運(yùn)營(yíng)調(diào)度概率分析。輸出包括9個(gè)替代隧道單位成本分布，如圖2。</p><p>　　在一個(gè)特定的一輪施加挖掘方法中的爆破后單位成本取決于當(dāng)時(shí)的地面類。如果選定的方法是適當(dāng)?shù)娘@露地質(zhì)條件，這一決定將導(dǎo)致最低的單位成本在這輪[例如，（EM1，GC1），（EM2，GC2），（EM3，GC3）]的地質(zhì)條件。相反，如果所選擇的方法在結(jié)構(gòu)上不夠[例如，（EM3，GC1

26、）]實(shí)際地面條件，隧道的單位成本將比正確的決策案例更高。</p><p>　　風(fēng)險(xiǎn)敏感型動(dòng)態(tài)決策模型</p><p>　　基于以往機(jī)型的投入，隧道決定使用決策和風(fēng)險(xiǎn)分析，來(lái)確定風(fēng)險(xiǎn)調(diào)整后的成本和最優(yōu)的隧道策略，這兩者都是承包商的風(fēng)險(xiǎn)敏感度的功能。</p><p>　　圖3顯示了承包商的不同程度的風(fēng)險(xiǎn)敏感性所產(chǎn)生的隧道風(fēng)險(xiǎn)調(diào)整成本。如可以看到的，這個(gè)項(xiàng)目預(yù)期的隧道成本

27、（γ= 0）大約為30.3M。由于風(fēng)險(xiǎn)厭惡系數(shù)γ增大（即承包人變得更加規(guī)避風(fēng)險(xiǎn)），風(fēng)險(xiǎn)調(diào)整后的成本增加幾乎呈線性。與此相反，作為風(fēng)險(xiǎn)厭惡系數(shù)降低（即，一個(gè)承包商有更多的風(fēng)險(xiǎn)偏好），風(fēng)險(xiǎn)調(diào)整后的成本幾乎呈線性下降。</p><p>　　圖4顯示了對(duì)給定西隧道段最佳的隧道策略，其承包商風(fēng)險(xiǎn)規(guī)避系數(shù)γ=5。九條圖中的施工過(guò)程中對(duì)應(yīng)地類和開挖方法的九種可能的組合。例如，由于隧道地質(zhì)在位置40.2米（132英尺）時(shí)遇到的G

28、C1和EM1是在上一輪所用，為規(guī)避風(fēng)險(xiǎn)承包與γ=5的最優(yōu)策略是使用同樣的方法（即，第一欄）。然而，如果當(dāng)前的地質(zhì)條件和GC2 EM1被使用時(shí)，承包商在該隧道階段應(yīng)切換到EM2（即，在第四欄）。</p><p><b>　　結(jié)論</b></p><p>　　被建議的風(fēng)險(xiǎn)敏感型決策支持系統(tǒng)，既可以量化，并納入與隧道工程相關(guān)的所有重大風(fēng)險(xiǎn)是第一個(gè)系統(tǒng)。該系統(tǒng)可用于確定動(dòng)態(tài)優(yōu)

29、化隧道計(jì)劃和風(fēng)險(xiǎn)調(diào)整后的成本，來(lái)作為承包商的風(fēng)險(xiǎn)敏感性的功能。因此，它可以提供最佳的決定不僅規(guī)劃和估算隧道工程施工前還根據(jù)實(shí)際地質(zhì)條件和施工開挖過(guò)程中的支護(hù)方法，選擇最優(yōu)的開挖。</p><p>　　圖1。隧道斷面和支持系統(tǒng)類型3（SS3）</p><p>　　圖2。對(duì)于不同選擇的隧道的單位成本累積分布函數(shù)[注：a-（EM1，GC1），b-（EM1，GC2），c-（EM1，GC3），d-（

30、EM2，GC1），e（EM2， GC2），f-（EM2，GC3），g-（EM3，GC1），h-（EM3，GC2），i-（EM3，GC3）]</p><p>　　圖3。風(fēng)險(xiǎn)調(diào)整后的隧道成本和風(fēng)險(xiǎn)規(guī)避系數(shù)之間的關(guān)系（γ）</p><p>　　圖4。懸湖隧道工程隧道最優(yōu)政策 </p><p>　　（西段）為γ=5（承包商的規(guī)避風(fēng)險(xiǎn)）</p><p>

31、;<b>　　參考文獻(xiàn)</b></p><p>　　埃塞克斯郡，河，路易斯，D.，克萊因，S.和特拉帕尼，R.（1993）。 “懸湖隧道，格倫伍德峽谷，科羅拉多的巖土工程方面的問(wèn)題?！边^(guò)程?？焖倬蜻M(jìn)和隧道機(jī)密。，卷907-926</p><p>　　約安努，P.G.（1984）。 “地質(zhì)勘查的經(jīng)濟(jì)價(jià)值，地下施工風(fēng)險(xiǎn)降低策略?！辈┦空撐模聊九c環(huán)境工程，麻省理工學(xué)院，劍

32、橋，麻省處。</p><p>　　約安努，P.G. （1988）。 “地質(zhì)勘探和減少地下工程風(fēng)險(xiǎn)。”中國(guó)建設(shè)工程與管理，ASCE的，114（4）,532-547</p><p>　　約安努，P.G.和馬丁內(nèi)斯，J.C.（1998）。 “使用基于狀態(tài)的概率決策網(wǎng)絡(luò)工程調(diào)度?！边^(guò)程。 1998年冬季仿真會(huì)議，華盛頓特區(qū)，1287年至1295年</p><p>　　利茲，

33、山和朱厄特公司（1981）。 “初步地質(zhì)調(diào)查和實(shí)驗(yàn)室檢測(cè)：懸湖隧道，格倫伍德峽谷，科羅拉多州”的公路系，科羅拉多州準(zhǔn)備報(bào)告。</p><p>　　Likhitruangsilp，V.（2003）。 “以風(fēng)險(xiǎn)為基礎(chǔ)的動(dòng)態(tài)決策支持系統(tǒng)的隧道施工。”博士論文，土木與環(huán)境工程，密歇根大學(xué)，密歇根州安阿伯系。</p><p>　　Likhitruangsilp，五，和約安努，P.G. （2003）。

34、 “使用離散事件仿真隨機(jī)的隧道傳輸性能的評(píng)估?！边^(guò)程。 2003結(jié)構(gòu)研究大會(huì)，3月19-21,2003，夏威夷州檀香山。</p><p>　　Scotese，T.R.，以及阿克曼，J.L.（1992）。 “工程考慮爆破的懸湖隧道工程，格倫伍德峽谷，科羅拉多州?！边^(guò)程。國(guó)際協(xié)會(huì)炸藥的工程師，奧蘭多，佛羅里達(dá)州，387-402。</p><p>　　Risk-sensitive decisio

35、n support system for tunnel construction</p><p>　　Comprehensive and realistic tunneling plans must strive for optimal decisions that minimize time and cost while addressing important factors such as geologic

36、 uncertainty and variability,uncertainty in tunneling productivity , and the contractor’s risk sensitivity . This paper presents a computerized decision support system that incorporates all important tunneling risks. It

37、consists of three interrelated models: the probabilistic geologic prediction model , the probabilistic tunnel cost estimati</p><p>　　The probabilistic tunnel cost estimating model evaluates tunneling time an

38、d cost performances for applying different excavation and support methods to different prevailing ground conditions by using Monte Carlo simulation to actual tunneling operations . Both models provide the main input for

39、the risk-sensitive dynamic decision model, the core of the system, to determine the optimal excavation and support sequence and the corresponding risk-adjusted tunneling costs for the project as functions o</p>&l

40、t;p>　　The application of the system to an actual highway tunneling project illustrates both the modeling power of the approach to quantify and incorporate risk, and its effectiveness for making optimal decisions as fu

41、nctions of the contractor’s degree of risk sensitivity.</p><p>　　1, Ph.D. ,Lecturer,Department of Civil Engineering, Chulalongkorn University, Bangkok 10330, Thailand , fcevlk@eng.chula.ac.th</p><

42、p>　　2, Ph.D. , Professor, Department of Civil and Environmental Engineering, University of Michigan,Ann Arbor, MI 48109-2125, USA, photios@umich.edu</p><p>　　Geotechnical engineering for transportation pr

43、ojects</p><p>　　Introduction</p><p>　　Tunnels are vital options for modern transportation systems because they are the only transportation system that can solve problems of difficult terrains,li

44、mited surface space,and increased demand of transportation. At the same time, tunnels are expensive underground structures where a variety of risks are encountered in every phase of the project delivery process. Comprehe

45、nsive and realistic tunneling plans must strive for optimal decisions that minimize time and cost while addressing important</p><p>　　Tunneling risks </p><p>　　One of the most important decision

46、s in tunneling is to determine the optimal sequence of excavation methods and support system along the tunnel profile. These decisions are characterized by four primary factors: geologic uncertainty, geologic variability

47、, uncertainty in tunneling productivity,and risk sensitivity.</p><p>　　Geologic uncertainty</p><p>　　The selection of tunneling methods for a project depends primarily on the expected geologic c

48、onditions of the tunnel, which are the aggregation of states of important rock mass properties such as rock type and discontinuities. Regardless of the number and extent of subsurface exploration undertaken, the tunnel g

49、eology cannot be known perfectly before construction begins. Even though several tunneling practices(e.g.,the observational method) have been adopted to mitigate geologic uncertainty, the</p><p>　　Geologic v

50、ariability</p><p>　　Most tunnels traverse a variety of geologic conditions, the locations and extents of which are impossible to define in advance with certainty. For most tunneling projects with significant

51、 geologic variability, the selected tunneling methods must be adaptable to all anticipated geologic conditions without seriously interrupting excavation progress. These adaptable tunneling methods encompass the modificat

52、ion of excavation methods(e.g.,heading and bench, and multiple drift), round length, drill pa</p><p>　　Uncertainty in tunneling productivity</p><p>　　Another risk in tunneling decisions results

53、from uncertainty in the productivity of tunneling processes. This uncertainty stems from the variation of construction equipment performance, the variation of worker outputs, and unexpected events such as accidents durin

54、g construction. This uncertainty exists even if geologic conditions are known. Thus, its impact on tunneling decisions must be addressed explicitly.</p><p>　　Risk sensitivity</p><p>　　Individual

55、 valuation of benefits and costs for decisions involving risk (e.g., tunneling decisions) is often nonlinear because these decisions are not based on the maximization of expected monetary value. In other words, when maki

56、ng decisions under uncertainty a decision maker is typically sensitive to risk, either risk averse or risk preferring. An individual’s risk sensitivity (risk preference) is influenced by several factors, especially that

57、person’s current net asset position. Typically, as</p><p>　　A contractor’s risk aversion and its degree of risk exposure can have a major influence on construction decisions and the necessary amount of risk

58、premium or contingency embedded in a contractor’s price in order to undertake the work. A more risk-averse contractor adopts a more conservative plan and includes a higher allowance as contingencies in his bid than a les

59、s risk-averse contractor does (Ioannou 1988). Thus, it is necessary to incorporate risk sensitivity into tunneling decisions.</p><p>　　By considering all above factors, tunneling decisions can be considered

60、a risk-sensitive dynamic probabilistic decision process, which can be structures by the risk-sensitive decision support system (Likhitruangsilp 2003).</p><p>　　Risk-sensitive decision support system</p>

61、;<p>　　The risk-sensitive decision support system consists of three interrelated models: the probabilistic geologic prediction model, the probabilistic tunnel cost estimating model, and the risk-sensitive dynamic

62、decision model.</p><p>　　Probabilistic geologic prediction model</p><p>　　The probabilistic geologic prediction model uses all available geologic information to characterize geologic uncertainty

63、 and variability along the tunnel profile in the probabilistic form of ground class transitions. The model is based on discrete-state, continuous-space Markov processes of important geologic parameters (e.g., rock fractu

64、re). These geologic Markov models are created from regional data (e.g., geologic maps) and updated by location-specific data (e.g., borehole tests) using direct</p><p>　　The model has been programmed in MATL

65、AB. Its input includes the length of tunnel, the extent of each stage (e.g., round length), geologic parameters and their states, and the definition of ground classes. Based on this input, the model calculates the poster

66、ior state probabilities of geologic parameters and ground classes at different locations along the tunnel. Both state probabilities are subsequently used to determine the ground class transition probability matrices of t

67、he tunnel geology by ap</p><p>　　Probabilistic tunnel cost model</p><p>　　The probabilistic tunnel cost estimating model performs stochastic evaluation of tunneling time and cost performance for

68、 different combinations of excavation and support methods with different ground classes (tunneling alternatives). The model includes the cost estimating submodel and the probabilistic scheduling submodel.</p><

69、p>　　The cost estimating submodel, created in a computer spreadsheet, organizes tunneling cost items, performs quantify takeoff computations, and calculates fixed costs and variable costs associated with each alternati

70、ve. In addition to normal tunneling costs,it also considers risks of selecting a wrong excavation method during construction. Its input includes a work breakdown structure (WBS) designed specifically for tunneling projec

71、ts; specifications of excavation methods and support systems; crew </p><p>　　The probabilistic scheduling submodel is implemented in ProbSched, a probabilistic scheduling simulation program (Ioannou and Mart

72、inez 1998).It evaluates tunneling time and cost performances for different alternatives.In addition to the input from the cost estimating submodel, it requires precedence networks of tunneling activities for different al

73、ternatives; time equations for activities in the activity networks; parameters of the time equations, either defined deterministically or assessed sub</p><p>　　Risk-sensitive dynamic decision model</p>

74、<p>　　The risk-sensitive dynamic decision model, the core of the proposed system, is formulated as a risk-sensitive stochastic dynamic programming model. Its input includes the ground class transition probability

75、matrix for each tunneling stage determined by the tunneling unit cost distributions for different alternatives simulated by the probabilistic tunnel cost estimating model. The model also requires the decision maker’s (e.

76、g., contractor’s) risk aversion coefficient (γ), which is the parameter o</p><p>　　The risk-sensitive dynamic decision model, programmed in MATLAB, performs decision and risk analysis to determine the optima

77、l tunneling policies and risk-adjusted tunneling costs of the project, both of which are functions of available information and the decision maker’s degree of risk sensitivity.</p><p>　　Application</p>

78、<p>　　The Hanging Lake Tunnel, a highway tunneling project in Colorado, is used to demonstrate the application of the proposed system. This rock tunneling project involved the construction of a pair of two-lane hi

79、ghway tunnels: the eastbound and the westbound tunnels. Here, we focus on the part of the westbound tunnel excavated by multiple-drift and blast methods.</p><p>　　Based on several rock mass classification sy

80、stems, the geologic conditions were classified into three ground classes: GC1 (best), GC2(medium), and GC3(worst). Three excavation methods (EM1,EM2,EM3) AND initial support systems (SS1,SS2,SS3) were designed correspond

81、ing to the three ground classes. For example, EM2 and SS2 are the most economical and structurally adequate excavation of six headings (drifts) and rock reinforcement systems consisting of dowels,spiles,and shotcrete,as

82、shown in Figure </p><p>　　Probabilistic geologic prediction model</p><p>　　The probabilistic geologic prediction model for the Hanging Lake Tunnel was developed based on three important geologic

83、 parameters: rock quality designation (RQD), fracture frequency, and weathering and alteration. The combination of these geologic parameter states are classified into three ground classes corresponding to the classificat

84、ion described by Essex et al.(1993).</p><p>　　The parameters for each geologic Markov model were estimated by analyzing data from the logs of boreholes (Leeds,Hill and Jewett, Inc. 1981).The posterior state

85、probabilities at the observation points were subjectively encoded based on a variety of assessments by geology experts, including Leeds,Hill and Jewett, Inc. (1981), and Scotese and Ackerman (1992). These probabilities

86、were used to determine the posterior state probabilities for non-observation points at intervals of 3.7 m (12 ft) alon</p><p>　　For example, given that the tunnel geology class 1 at location 746.1 m, the pr

87、obabilities that it will make a transition to ground class 1 (remain the same), ground class 2, and ground class 3 at location 749.8 m are 44.52, 46.79, and 8.96 percent, respectively (i.e., the first row of the above ma

88、trix).</p><p>　　Probabilistic tunnel cost estimating model</p><p>　　According to available information of the construction resources used in this project, the cost estimating submodel organized

89、and calculated the equipment, labor, and equipment costs for each alternative. These costs then categorized into fixed costs and variable costs. The variable costs were used as inputs for the probabilistic scheduling sub

90、model, which performed the probabilistic scheduling analysis of tunneling operations.The output included tunneling unit cost distributions for the nine al</p><p>　　The tunneling unit costs for applying an ex

91、cavation method in a particular round depend upon the prevailing ground class after blasting. If the selected method is appropriate for the revealed geologic conditions, this decision will lead to the lowest unit cost fo

92、r the geologic conditions in that round [e.g., (EM1,GC1),(EM2,GC2),(EM3,GC3)]. In contrast, if the selected method is structurally inadequate [e.g., (EM3,GC1)] for the actual ground conditions, the tunneling unit costs w

93、ill be higher tha</p><p>　　Risk-sensitive dynamic decision model</p><p>　　Based on the inputs from the previous models, the tunneling decision was solved by using decision and risk analysis to d

94、etermine risk-adjusted costs and optimal tunneling policies, both of which are functions of the contractor’s risk sensitivity.</p><p>　　Figure 3 shows the resulting risk-adjusted tunneling costs for differen

95、t degrees of the contractor’s risk sensitivity. As can be seen, the expected tunneling cost for this project (γ=0) is approximately ＄30.3M. As the risk aversion coefficient γ increases (i.e., a contractor becomes more r

96、isk adverse), the risk-adjusted cost increases almost linearly. In contrast, as the risk aversion coefficient decreases (i.e., a contractor is more risk preferring), the risk-adjusted cost decreases almost lin</p>

97、<p>　　Figure 4 shows the optimal tunneling policies for the west tunnel segment given that the contractor is risk averse with γ=5. Nine bars in the figure correspond to the nine possible combinations of ground clas

98、ses and excavation methods during construction. For example, given that the tunnel geology encountered at location 40.2 m (132 ft) is GC1 and EM1 was used in the previous round, the optimal policy for the risk-averse con

99、tractor with γ=5 is to use the same method (i.e., the first bar). Howev</p><p>　　Conclusions</p><p>　　The proposed risk-sensitive decision support system is the first system that can both quant

100、ify and incorporate all important risks associated with tunneling work. The system can be used to determine dynamic optimal tunneling plans and risk-adjusted costs as functions of a contractor’s risk sensitivity. Thus, i