深基坑稳定性多参数风险评估方法研究

【作者】 杨军；

【导师】 吴国勇；

【作者基本信息】 江西理工大学 ， 测绘工程（专业学位）， 2015， 硕士

【摘要】 随着城市对地下空间的需求逐步提高,深基坑工程越来越受到广泛关注。如何利用深基坑有限监测数据对深基坑的健康状况进行评估,成为重要的研究课题。传统的深基坑风险评估主要是根据规范要求制定监测参数的限值,然后将监测参数的原始数据与限值进行比对,评估基坑的稳定性;在预测方面,采取某一数学模型对某一连续监测参数数据进行预测,然后与规范限值进行比对。这种单参数的深基坑稳定性评估方法已经不能全方位的反映深基坑的健康状况,而且对于复杂的深基坑,由于没有考虑到相关参数的变化情况,该方法的评估结果和预测结果不够合理、科学。针对以上问题,本论文拟提出以下三个解决方案:1.提出灰色BP神经网络模型,运用该模型对深基坑监测参数进行预测,为多参数评估方法提供基础数据。通过案例对比分析可以得到该组合模型预测精度优于灰色模型和BP神经网络模型,预测精度较高,预测结果可以很好的应用于深基坑稳定性单参数和多参数风险评估中。2.提出深基坑稳定性的多参数评估方法,通过对深基坑支护体系和周边环境的力学变形进行分析并归纳总结监测数据的变化规律,提出包括围护结构侧移、地表沉降、支撑轴力、立柱桩隆沉、地墙竖向位移等五个参数之间的关系,用该关系作为判断是否深基坑稳定的标准,通过实例验证表明多参数评估方法能够发现单参数评判方法不能发现的基坑风险,该方法能更好的反映深基坑支护体系和周围环境的应力变化情况,对单参数的基坑风险评估方法是一种很好的补充,可以给基坑施工人员提供一种更全面的基坑稳定性风险评估方法。3.提出将数据融合技术运用到深基坑稳定性多参数评估中,提高多参数评估的准确性。传统的深基坑监测数据处理方式是对某一监测参数数据进行分析预测,对于其他监测参数不能综合处理、融合分析,这种预测方法考虑因素太单一。以BP神经网络为核心的数据融合技术能够对深基坑各监测参数数据进行综合处理,拟合出一条最逼近实际情况的线性或者非线性的函数。通过实例研究表明该数据融合方法较之单个参数的预测方法,精度更高,预测结果更合理,可以给单参数评估方法和多参数评估方法提供更科学的基础数据。深基坑稳定性多参数评估方法是通过研究深基坑力学特性和数据统计经验得出的,该方法能够很全面的反映深基坑健康状况,是一种很好的评估方法。在多参数评估中,各参数的预测数据必须精度高,而且必须科学、合理,这样才能发挥多参数评估方法的优势,本论文提出的灰色BP神经网络模型和数据融合技术可以有效解决这个问题,可以给多参数评估方法提供更科学合理的数据。更多还原

【Abstract】 As the demand for underground space of the city gradually improve, deep foundation pit engineering more and more attention. How to evaluate the use of limited data monitoring of deep foundation pit in the health status of deep foundation pit, become the important research topic. The deep foundation pit of traditional risk assessment is mainly according to the specification requirement to develop monitoring parameters of the limit value, the original data and monitoring parameters and the limit value for comparison, evaluation of the stability of foundation pit; in the forecast, to take a mathematical model to predict a continuous monitoring data, and then the values were compared with the specification limits. The evaluation method of the deep foundation pit stability of this single parameter cannot have a full range of reflection of deep foundation pit’s health, but also for the deep foundation pit in complex, because there is no change in consideration of relevant parameters, the method of the evaluation results and the prediction results are not reasonable and scientific. To solve the above problem, this paper proposes the following three solutions:1. Grey BP neural network model is proposed, using this model to predict the deep foundation pit monitoring parameters, methods to provide the basic data for the multi parameter evaluation. Through analysis and comparison of cases can be obtained for the combined model prediction precision is better than that of grey model and BP neural network model, the forecast precision is high, the prediction results can be well applied in the deep foundation pit stability of single parameter and multi parameter in risk assessment.2. Proposed the multi parameters of deep foundation pit stability evaluation methods, through mechanical deformation of deep foundation pit support system and surrounding environment are analyzed and the change law of summary of the monitoring data, put forward the relationship between the five parameters including the retaining structure, the ground surface settlement, lateral displacement of supporting axial force of column pile, uplift and subsidence of the vertical displacement of the wall, etc. using this relationship, as to determine whether the stability of deep excavation standards, through the examples verify the evaluation method show that the multi parameters can be found in foundation pit risk not single parameter evaluation method of discovery, this method can better reflect the deep foundation pit supporting system and the surrounding environment and stress change pit risk evaluation method is a kind of single parameter good supplement, can give people the foundation pit construction to provide risk the stability of foundation pit of a more comprehensive assessment method.3.Put the data fusion technology is applied to the deep foundation pit stability of multi parameter evaluation, evaluation to improve the accuracy of multi parameter. The traditional deep foundation pit monitoring data processing method is to analyze and forecast a monitoring parameter data, for the analysis of other monitoring parameters can not be integrated processing, fusion, this forecasting method consideration factor is too single. To carry out comprehensive treatment of deep foundation pit of the monitoring parameter data with BP neural network as the core of the data fusion technology can, fitting out a most approximation of the actual situation of the linear or nonlinear function. Prediction method is superior to that of the single parameter method of the data fusion showed through the case study, higher accuracy, the prediction results more reasonable, can method and multi parameters to a single parameter assessment method provides basic data more scientific.Deep foundation pit stability evaluation method of multi parameters is derived through the study of deep foundation pit of mechanics characteristic and experience of data statistics, this method can comprehensively reflect the deep foundation pit of health, is a good evaluation method. In the multi parameter evaluation, prediction data of each parameter must be of high precision, and it must be scientific and reasonable, so as to play the advantages of multi parameter evaluation method, this paper puts forward the grey BP neural network model and data fusion technology can effectively solve this problem, can provide a more scientific and reasonable data to multi parameter evaluation method.更多还原