【作者】 孙晏一；

【导师】 刘寒冰；

【作者基本信息】 吉林大学 ， 道路与铁道工程， 2012， 博士

【摘要】 道路桥梁等大型土木工程结构物的健康监测系统是目前世界范围内研究的热点问题之一，而各种监测系统中最重要的部分则是对监测数据的分析处理。特别对于无线传感器网络而言，监测数据的传输效率往往是整个系统成败的关键。在监测数据无线传输之前，对监测数据进行甄别可以有效减小数据传输量，提高传输效率，保证监测预警系统的实时性和有效性。我国从20世纪90年代中后期开始研究自动化的结构健康监测平台，虽然取得了一定的效果，但是对于监测数据自适应甄别处理的研究还不是很理想。为了在监测过程中能够有效的对数据进行自适应甄别处理，避免传输信道因海量监测数据而发生传输滞后现象，影响预警平台的实时性和有效性。本文结合国家高技术发展计划—“863”计划：“大范围季节冰冻区道路灾害参数监测与辨识预警系统研究”，并根据课题中有关数据甄别的研究要求，在阅读大量相关文献的基础上，通过将传统与现代数据分析处理技术相结合的方法实现了监测数据的自适应甄别处理过程：1、对边坡、路基和桥梁结构的监测过程进行了分析，对各种结构的监测参数特点进行了分析。明确了边坡、路基、桥梁的主要监测参数，确定了整个监测过程中实时监测数据甄别的基本框架和流程。2、提出了一种基于时间序列算法（ARMA）的数据甄别模型，实现了边坡位移、路基沉降实时监测数据的自适应甄别处理。通过室内试验和实际工程两方面验证了该方法的可行性和有效性。3、形成了一种基于灰色关联度理论的桥梁动力响应监测数据甄别方法，此方法能够剔除外界温度对监测数据的影响。使用该方法能够实现桥梁结构动力响应数据的自适应甄别处理。通过在长春市硅谷立交桥监测中的应用，验证了该方法的可行性和有效性。在监测数据的自适应甄别处理上，首先将监测传感器采集到的原始监测数据通过有线的方式存储到基站计算机的数据库中。然后在基站计算机中，通过编制的自适应甄别程序对数据进行甄别处理。最后，对于需要发送到预警平台的监测数据，使用3G无线网络进行发送，对于甄别后的无效数据则保存在基站计算机中一段时间后，自动清除，以保证基站计算机有足够的空间来实现监测数据的存储和甄别过程。在对边坡、路基监测数据进行甄别处理之前，首先要对监测传感器采集到的原始监测数据进行预处理，即保证监测数据的连续性和准确性。由于边坡路基是长时间处于稳定状态的结构，而且拉绳式位移计和精密单点位移计均是深埋入其中的，因此监测数据相对准确，受到外界的影响较小。因此，认为两种结构的监测数据受到噪声的影响不大。故使用不复杂的去噪方法对原始监测数据进行预处理。同时，由于太阳能供电可能导致电压不稳等客观原因，一般会产生监测传感器的漏采现象：即数据采集不是等时间间距的。为了不改变监测数据的性质，需要对漏采数据进行补齐。针对边坡路基数据本文使用拉格朗日插值的方法对其进行补齐。在对原始监测数据进行了以上预处理过程之后，需对数据进行平稳化处理。一般情况下，监测到的数据均是非平稳的时序，这时，可以使用差分的办法对数据序列进行差分处理，并对差分后的数据序列进行自相关系数检验以便保证数据序列已经近似为平稳时序，可以用来进行ARMA建模。在监测数据的ARMA甄别模型建立过程中，作者通过借鉴其参数的求解方法推导了ARMA数据甄别模型的计算公式，并对模型的求解方法进行了详细阐述。最后，对推导出的计算公式进行编程，并在室内试验过程中对此方法进行了验证，在取得了很好的甄别效果后，将其应用于长春地铁西客站和国道“102”线的监测工程当中，取得了良好的效果。在桥梁的监测数据甄别上，首先通过FFT变换将传感器采集的加速度响应计算为桥梁的实时频率数据，并通过二元回归分析剔除外界温湿度对其实时频率数据灰色关联度的影响。然后继续使用灰色关联度算法，将计算出的桥梁前4阶频率与桥梁初始状态的频率进行比较。以监测数据作为比较序列，以桥梁结构初始状态的频率作为原始序列。对于比较后，灰色关联度大于阈值的频率作为安全数据不对其进行无线传输发送；对于比较后，灰色关联度小于阈值的频率值作为不安全数据，将其通过3G无线网络发送到预警平台进行判断。将以上过程进行编程以实现桥梁监测数据的自适应甄别处理，并将其应用于长春市硅谷立交桥的监测工程中，取得了很好的监测效果。更多还原

【Abstract】 At present, health monitoring system of road and bridge structures is one of the mostpopular research directions in the world. And in the health monitoring system, the mostimportant part is the analysis of monitoring data. In addition, especially for the wirelesssensor networks, the data transfer efficiency is the key to the success of the entire healthmonitoring system. Before the wireless transmission of monitoring data, the discriminationof monitoring data can effectively reduce the amount of data to be transported and also canimprove the efficiency of the transmission, so that it can ensure the real-time andeffectiveness of monitoring and warning system. In China, the study of automated structuralhealth monitoring platform is from the late1990s. Although we have obtained some goodachievements, for the adaptive discrimination processing of monitoring data is not to besatisfied.In order to solve the adaptive discrimination processing of monitoring data during themonitoring process and avoid the transmission channel lag phenomenon due to the mass ofmonitoring data which would affect the real-time and effectiveness of warning platform, wehave analyzed the monitoring process of slope, subgrade and bridges on the basis of readinglots of relevant literatures. Then, we study on the features of monitoring parameters of thethree structure above combined with the National High Technology Research andDevelopment Program ("863"Program) of China named “The research of wide range seasonfrozen road disasters parameter monitoring and identification of warning system” andaccording to the research topics in the data discrimination requirements.1. We analyze the monitoring process of slope, subgrade and bridge structure. Themonitoring parameters of these structure are studied. Then the frame of data discriminationwere determined.2. ARMA data discrimination algorithm based on modern time series model has beendecided, and the setting up and solving process is derived in detail and elaboration. Slopedisplacement, subgrade settlement of the real-time monitoring data adaptive discriminationprocess were established. Feasibility and effectiveness of this method are verified throughlaboratory tests and practical engineering.3. Dynamic response of the gray relational theory-based bridge monitoring data discrimination is established, and the discrimination model and solution procedure areanalyzed in detail. Adaptive discrimination of bridge frequency data is realized. In the modelbuilding process, the use of binary regression analysis of gray relational grade of thereal-time frequency data corrected to standard temperature, thus excluding the impact ofoutside temperature change. This method is applied to the Changchun City, Silicon Valleyoverpass monitoring projects, to verify the feasibility and effectiveness of discrimination onthe bridge dynamic response data.And on the adaptive discrimination processing of monitoring data, we first save themonitoring data into the database of base station computer. This stored procedure is finishedby the transmission line connected with monitoring sensors and base station computer.Secondly, we realize the adaptive discrimination processing of monitoring data byprogramming in the base station computer. Finally, after the data discrimination, for themonitoring data need to be send to the warning platform, we use3G Wireless Networks toconduct the transmission process. After data discrimination, the invalid monitoring data isstored in base station computer for some time, and then it would be automatically cleared toensure that the base station computer has enough space to store monitoring data andcomplete the process of data discrimination. On the algorithm of data adaptivediscrimination processing, after comparative study of a variety methods of datadiscrimination, we decide that the slope displacement data and subgrade settlement data areusing time series methods to data discrimination. However, the bridge frequency data areusing the gray grey correlation method to discriminate.Before the discrimination of displacement data of slope and subgrade, we must do somepretreatment to the original monitoring data collected through the monitoring sensors in filedso that we can ensure continuity and accuracy of monitoring data. Because slope andsubgrade structure is also in a stable state for a long time, and the rope type displacementmeter and precision single-point displacement meter are buried deep into the slope andsubgrade, respectively, the monitoring data of the two meters is relatively accurate and lesssubject to outside influence. On this basis, we believe that the two structural monitoring datais not much affected by noises. Thus, we use simple denoising method to preprocess the rawmonitoring data. At the same time, because solar power supply may lead to voltageinstability and for other objective reasons, it will generally have a monitoring sensor leakagephenomenology: the data acquisition is not the equal time interval. So for the nature ofmonitoring data could not be changed, we need to fill the data which is lacking. In this paper, we use lagrange interpolation method to fill the data. After the pretreatment process above,we need to test the stationarity of the monitoring data. Under normal circumstances, themonitoring data are non-steady time series. We usually use the differential approach to treatthe monitoring data sequence and then test the self-correlation coefficient in order to the datasequence has become stationary time series. Then we can use it for the ARMA modeling.During the ARMA modeling of monitoring data discrimination, the author use the solutionof how to solve the parameter to derive the formula of ARMA modeling and describe thesolution process in detail. Finally, we use the derived formula for programming. Then, thismethod is validated in the room test and obtains a good discriminate effect. After that, weapply the method to Changchun Metro West Railway Station and State Road “102” linemonitoring project and it is also has achieved good results.In discrimination of bridge monitoring data, we first calculate the bridge frequencythrough the FFT transform using the acceleration response data. Then, we use greycorrelation algorithm to compare the bridge frequency with the initial state of bridge. We usethe monitoring data as the comparison sequence and the frequency of the initial state of thebridge structure as the original sequence. After comparison, if gray relational degree of themonitoring is greater than the threshold frequency, we regard it as the safety data and won’tsend it wirelessly. On the opposite side, we send the data whose relational degree is less thanthe threshold frequency to the warning platform to judge. We use the method above forprogramming in order to realize the adaptive discrimination of bridge monitoring data andwe do the indoor experiments. After achieving better results in the laboratory test, we use itfor the bridge monitoring in Changchun City and have a good monitoring effect.更多还原