本文關(guān)鍵詞：基于粗糙集和RBF神經(jīng)網(wǎng)絡(luò)的地鐵施工安全風(fēng)險評估，由筆耕文化傳播整理發(fā)布。

基于粗糙集和RBF神經(jīng)網(wǎng)絡(luò)的地鐵施工安全風(fēng)險評估

作者：1”的相關(guān)文章'>陳帆1  2”的相關(guān)文章'>謝洪濤2 

單位：1. 湖南科技大學(xué)土木工程學(xué)院,湖南湘潭 411201;
2. 昆明理工大學(xué)建筑工程學(xué)院,昆明 650093

關(guān)鍵詞：安全管理工程 地鐵施工 安全風(fēng)險 RBF神經(jīng)網(wǎng)絡(luò) 粗糙集 

分類號：X948

出版年·卷·期（頁碼）：2013·13·第4期（232-235）

摘要：

為了解決當(dāng)前我國地鐵施工過程中所面臨的安全風(fēng)險評估問題,提出了一種基于粗糙集和RBF神經(jīng)網(wǎng)絡(luò)的地鐵施工安全風(fēng)險評估模型。在分析地鐵施工安全風(fēng)險評估指標(biāo)的基礎(chǔ)上,從地質(zhì)條件、施工技術(shù)、施工管理等方面選擇了33個變量指標(biāo)。針對傳統(tǒng)神經(jīng)網(wǎng)絡(luò)收斂速度慢、容錯性差、結(jié)果并不唯一的缺點,利用粗糙集理論的屬性約簡方法使RBF神經(jīng)網(wǎng)絡(luò)的輸入數(shù)據(jù)減少且不相關(guān),并利用長沙、武漢、杭州、昆明、北京、上海、廣州、重慶的28個地鐵工程項目的問卷調(diào)查數(shù)據(jù)實現(xiàn)模型的訓(xùn)練及檢測。結(jié)果表明,采用粗糙集方法約簡屬性能使RBF網(wǎng)絡(luò)的輸入數(shù)據(jù)從33個減少至15個,用經(jīng)過粗糙集約簡后的樣本集作為神經(jīng)網(wǎng)絡(luò)的訓(xùn)練樣本集可以有效地簡化神經(jīng)網(wǎng)絡(luò)的結(jié)構(gòu),減少訓(xùn)練步數(shù)與訓(xùn)練時間,并提高網(wǎng)絡(luò)的學(xué)習(xí)速度和判斷準(zhǔn)確率。同時,經(jīng)過粗糙集約簡后的神經(jīng)網(wǎng)絡(luò)的收斂速度和計算精度能夠滿足地鐵施工安全風(fēng)險評估的需要。通過粗糙集與RBF神經(jīng)網(wǎng)絡(luò)相結(jié)合所構(gòu)建的耦合模型可以識別地鐵施工過程的安全狀態(tài),進而有針對性地完善地鐵施工的相關(guān)安全技術(shù)。

The present article is inclined to bring forward an urban subway construction risk assessment based on the rough set and RBF neural network. As is known, it has become urgent to promote the safer and more securable subway construction environment with the current surge of subway construction tide in China. In this article, we would like to propose a subway construction risk assessment model based on the analysis of subway construction safety risk assessment model including indicators, which we have chosen of 33 variables in the light of geological conditions, the construction technology, management and so on. In order to overcome the drawbacks of the traditional artificial neural network, namely, slow convergence, poor fault tolerance and inconsistent results, we prefer to choose the rough set method, intending to reduce the table attribute of the sampling decisions and decrease the amount of irrelevant index inputs of the neural network data. The rough set method also allows us to do training and testing on how the given RBF neural network model can be used together with the rough set method by means of the actual subway construction project data. Practically speaking, we have already gained the above said survey data from the 28 subway construction practice from the field project experience in eight cities, that is, from Wuhan, Changsha, Hangzhou, Kunming, Beijing, Shanghai, Guangzhou and Chongqing. The results of our analysis of the data quoted show that the data input with the RBF neural network can actually be reduced from 33 to 15 by using the rough set method. After reducing the sampling data through the rough set method, the sampling sets can be used as the training sets of the RBF neural network. The RBF neural network structure can be effectively simplified with the training steps and training time decreased, and the network training and learning speed and efficiency can be greatly improved. And, as a result, the convergence speed and the calculation accuracy can be easily made to meet the requirements of the subway construction safety risk assessment through the attribute reduction. Therefore, it can be seen that this present coupling of the rough set method and RBF neural network system can be surely taken as the safety state guaranteeing means along with the relevant safety technology for the urban subway construction.

參考文獻：

[1] HOU Yanjuan(侯艷娟), ZHANG Dingli (張頂立), LI Pengfei(李鵬飛). Analysis and control measures of safety accidents in Beijing subway construction [J]. Journal of Beijing Jiaotong University(北京交通大學(xué)學(xué)報), 2009, 33(3): 52-59. [2] GHOSH S, JINTANAPAKANONT J. Identifying and assessing the critical risk factors in an underground rail project in Thailand: a factor analysis approach[J]. International Journal of Project Management, 2004, 22(8): 633-643. [3] PENG Junlong(彭軍龍), ZHANG Xuemin(張學(xué)民), YANG Junsheng(陽軍生). Displacement back analysis of deep foundation pit for metro based on genetic algorithm and neural network [J]. Rock and Soil Mechanics(巖土力學(xué)), 2007, 28(10): 71-75. [4] PAN Guorong(潘國榮). Deformation prediction of subway tunnel with neural network method [J]. Journal of Geodesy and Geodynamics(大地測量與地球動力學(xué)), 2007, 27(1): 35-40. [5] JU Q, YU Z B. Division-based rainfall-runoff simulations with BP neural networks and Xinanjiang model [J]. Neurocomputing, 2009, 72(13): 2873-2883. [6] JIANG F, SUI Y F, CAO C G. Some issues about outlier detection in rough set theory [J]. Expert Systems with Applications, 2009, 36(3): 4680-4687. [7] GUO Zhijun(郭志軍), HE Xin(何昕), WEI Zhonghui(魏仲慧). Classification algorithm based on neural network and rough set [J]. Application Research of Computers(計算機應(yīng)用研究), 2011, 28(3): 94-98. [8] AQ 8004—2007 Safety pre-evaluation standard of urban rail transit(城市軌道交通安全預(yù)評價細(xì)則)[S]. [9] AQ 8005—2007 Safety evaluation standard on project completion of urban rail transit(城市軌道交通安全驗收評價細(xì)則)[S]. [10] DING Lieyun(丁烈云), WU Xianguo(吳賢國), LUO Hanbin(駱漢賓). Research on standard for construction safety assessment of metro engineering [J]. China Civil Engineering Journal(土木工程學(xué)報), 2011, 44(11): 121-127. [11] LIU Renhui(劉仁輝), YU Bo(于渤), JIN Zhen(金真). Study on index system of safety risk evaluation for subway construction based on interval estimation[J]. Forecasting(預(yù)測), 2012, 31(2): 62-66. [12] XUE Dingyu(薛定宇), CHEN Yangquan(陳陽泉). Advanced applied mathematical problem solutions with MATLAB(高等應(yīng)用數(shù)學(xué)問題的MATLAB求解)[M]. 2nd ed. Beijing: Tsinghua University Press(清華大學(xué)出版社), 2008: 126-170.

服務(wù)于反饋：

【文章下載】【發(fā)表評論】【查看評論】【】

提示：您還未登錄，請登錄！

  本文關(guān)鍵詞：基于粗糙集和RBF神經(jīng)網(wǎng)絡(luò)的地鐵施工安全風(fēng)險評估，，由筆耕文化傳播整理發(fā)布。