參數(shù)自適應擴展卡爾曼濾波理論及其在隧道變形預測中的應用
發(fā)布時間：2018-09-10 11:10
【摘要】：隨著我國隧道工程快速全面的發(fā)展,單體規(guī)模更加龐大、技術規(guī)格要求更高、地質(zhì)環(huán)境更加惡劣的隧道建設工程在不斷增多,這些工程在采用新奧法進行設計與建設時,從施工控制、理論校驗的角度出發(fā),隧道現(xiàn)場監(jiān)控量測就顯得更加的意義重大。然而,由于監(jiān)測手段以及后期數(shù)據(jù)處理分析方法的不完善、反饋的信息質(zhì)量不高等問題,使得現(xiàn)場監(jiān)控量測作業(yè)在校驗設計及指導施工方面遠未發(fā)揮其應有的作用。針對上述問題,本文作者提出了基于參數(shù)自適應擴展卡爾曼濾波理論的隧道變形預測的研究課題,并基于大量的現(xiàn)場監(jiān)測數(shù)據(jù)和地質(zhì)資料,在已有研究成果基礎上進行了深入系統(tǒng)地研究。首先,對現(xiàn)有隧道施工監(jiān)控量測數(shù)據(jù)分析方法進行了歸納總結,重點分析了回歸分析法、時間序列分析法、灰色預測法以及卡爾曼濾波法的原理和計算方法,并比較了各方法在隧道變形預測分析中的優(yōu)勢與不足。其次,基于比較分析結果,確立了以卡爾曼濾波為主要方法的研究思路�？紤]到隧道圍巖的變形與發(fā)展,受較多難以定量化的不明因素影響,其變形與發(fā)展具有非線性的特點,而卡爾曼濾波是一種處理線性問題的有效方法,所以聯(lián)想到了化非線性為線性方法,即擴展卡爾曼濾波法。由于該濾波方法需要已知的精確模型參數(shù)及噪聲統(tǒng)計,而實際應用中往往難以滿足上述前提,因此提出了自適應的方法。鑒于目前較多的自適應方法集中于以噪聲統(tǒng)計為基礎的方差自適應研究,而較少從模型偏差角度來解決參數(shù)自適應的方法,本文創(chuàng)造性地通過殘差特性因子進行濾波的斂散性評判,基于斂散趨勢進行相應的模型參數(shù)縮放調(diào)節(jié),實現(xiàn)了模型參數(shù)自適應,防止濾波發(fā)散失效。最后,運用提出的參數(shù)自適應擴展卡爾曼濾波方法,建立了隧道拱頂下沉及周邊位移的預測模型,并結合大量的現(xiàn)場量測數(shù)據(jù),進行隧道洞內(nèi)變形預測分析處理。將參數(shù)自適應擴展卡爾曼濾波和標準擴展卡爾曼濾波的預測結果與觀測值進行對比,分析了數(shù)自適應擴展卡爾曼濾波對隧道洞內(nèi)變形的預測性能與優(yōu)勢,探討了其適用性。本文所做研究工作,立足于學科前沿,采用先進的數(shù)學方法和手段對隧道洞內(nèi)變形預測進行了研究,具有較高的理論和應用價值,為隧道監(jiān)控量測數(shù)據(jù)的處理提供了一種新的有效分析手段。
[Abstract]:With the rapid and comprehensive development of tunnel engineering in our country, the scale of individual is more huge, the technical specifications are higher, and the geological environment is worse, the number of tunnel construction projects is increasing. When these projects are designed and constructed by the new Austrian method, from the point of view of construction control and theoretical checking, the field monitoring and measurement of tunnel becomes more and more important. However, due to the imperfection of monitoring means and the method of data processing and analysis, and the low quality of feedback information, the field monitoring and measurement work is far from playing its due role in checking design and guiding construction. Based on a large number of field monitoring data and geological data, this paper makes a thorough and systematic study on the existing research results. Firstly, the existing data analysis methods of tunnel construction monitoring and measurement are summarized, with emphasis on regression analysis, time series analysis and gray. The principle and calculation method of color prediction method and Kalman filter method are compared, and the advantages and disadvantages of each method in tunnel deformation prediction and analysis are compared. Secondly, based on the comparative analysis results, the main research method of Kalman filter is established. Because of the primitive effect, its deformation and development are nonlinear, and Kalman filter is an effective method to deal with linear problems, extended Kalman filter (EKF) is proposed to transform nonlinear into linear method. In view of the fact that more adaptive methods focus on variance adaptation based on noise statistics and less on model bias to solve parameter adaptation, this paper creatively evaluates the convergence and divergence of filtering by residual characteristic factor, which is based on convergence and divergence trend. The parameters of the model are scaled and adjusted accordingly, so that the parameters of the model can be adapted adaptively and the filtering divergence can be prevented. Finally, the prediction model of tunnel vault settlement and surrounding displacement is established by using the proposed parameter adaptive extended Kalman filter method. Combined with a large number of field measurement data, the deformation prediction and analysis of the tunnel are carried out. By comparing the predicted results with the observed values, the performance and advantages of the digital adaptive extended Kalman filter for predicting the deformation in tunnel are analyzed, and its applicability is discussed. The prediction of deformation in tunnel is studied, which has higher theory and application value, and provides a new effective analysis method for the data processing of tunnel monitoring and measurement.
【學位授予單位】：長沙理工大學
【學位級別】：碩士
【學位授予年份】：2014
【分類號】：U456.31
本文編號：2234274
[Abstract]:With the rapid and comprehensive development of tunnel engineering in our country, the scale of individual is more huge, the technical specifications are higher, and the geological environment is worse, the number of tunnel construction projects is increasing. When these projects are designed and constructed by the new Austrian method, from the point of view of construction control and theoretical checking, the field monitoring and measurement of tunnel becomes more and more important. However, due to the imperfection of monitoring means and the method of data processing and analysis, and the low quality of feedback information, the field monitoring and measurement work is far from playing its due role in checking design and guiding construction. Based on a large number of field monitoring data and geological data, this paper makes a thorough and systematic study on the existing research results. Firstly, the existing data analysis methods of tunnel construction monitoring and measurement are summarized, with emphasis on regression analysis, time series analysis and gray. The principle and calculation method of color prediction method and Kalman filter method are compared, and the advantages and disadvantages of each method in tunnel deformation prediction and analysis are compared. Secondly, based on the comparative analysis results, the main research method of Kalman filter is established. Because of the primitive effect, its deformation and development are nonlinear, and Kalman filter is an effective method to deal with linear problems, extended Kalman filter (EKF) is proposed to transform nonlinear into linear method. In view of the fact that more adaptive methods focus on variance adaptation based on noise statistics and less on model bias to solve parameter adaptation, this paper creatively evaluates the convergence and divergence of filtering by residual characteristic factor, which is based on convergence and divergence trend. The parameters of the model are scaled and adjusted accordingly, so that the parameters of the model can be adapted adaptively and the filtering divergence can be prevented. Finally, the prediction model of tunnel vault settlement and surrounding displacement is established by using the proposed parameter adaptive extended Kalman filter method. Combined with a large number of field measurement data, the deformation prediction and analysis of the tunnel are carried out. By comparing the predicted results with the observed values, the performance and advantages of the digital adaptive extended Kalman filter for predicting the deformation in tunnel are analyzed, and its applicability is discussed. The prediction of deformation in tunnel is studied, which has higher theory and application value, and provides a new effective analysis method for the data processing of tunnel monitoring and measurement.
【學位授予單位】：長沙理工大學
【學位級別】：碩士
【學位授予年份】：2014
【分類號】：U456.31
【參考文獻】
相關期刊論文 前2條
1 張忠亭;洞室圍巖收斂變形灰色理論預測[J];四川水力發(fā)電;1995年04期
2 陶志平;周德培;;蠕動型滑坡隧道的變形規(guī)律及災害預測[J];西南交通大學學報;2007年02期
，本文編號：2234274
本文鏈接：http://sikaile.net/kejilunwen/jiaotonggongchenglunwen/2234274.html
教材專著
