【作者】 彭朝勇；

【导师】 杨建思； 薛兵；

【作者基本信息】 中国地震局地球物理研究所 ， 固体地球物理学， 2013， 博士

【摘要】 地震预警是近二十年来新发展起来的减轻地震损失、降低地震次生灾害、减少人员伤亡的有效手段。其技术核心是从各个环节缩短数据处理所需时间,并结合地震P波传播速度快于破坏性S波和面波传播速度,以及电磁波传播速度远远大于地震波波速的原理,使得在破坏性地震发生后,破坏性地震波到来前,提供数秒至数十秒的预警时间,以便于采取相应应急处理措施。由于地震预警系统对于信息的高度时效性要求,需要从各个环节进行突破,从而缩短整个系统数据处理时间,增加预警时间,减少“盲区”范围。在前端用于实时数据获取的仪器方面,也有许多地方值得改进,如减少实时数据传输网络延迟、缩短数据的传输间隔、为现有绝大部分强震观测台站提供网络数据传输支持、引入数字校正技术、解决时间“跳秒”现象和个别地方GPS天线架设不便的问题等。本文对上述各个环节进行了探讨和研究,并将研究结果应用到了最新研发的一款强震观测仪器内部Linux系统上,形成了一套基于单台仪器的“现地”地震预警系统,相关主要工作概括如下：1、在一体化设计的强震仪中引入了数字校正技术。通过采用重力法测试15个不同的位置数据计算加速度计的灵敏度和安装方位角偏差,并将计算结果应用到了嵌入式软件系统中进行实时数据校正,有效提升了实时数据的精度和准确性；在前人工作的基础上,提出了一种新的加速度计温度补偿算法。该算法采用恒温MEMS加速度计输出数据的低频段(10-4Hz)对力平衡加速度计进行实时温度补偿,并在算法实现时采用了MEMS恒温点迁移的方式动态调整恒温MEMS温度,可以拓展仪器的应用范围和降低系统功耗。测试结果表明,经过补偿后的力平衡加速度计的零点漂移水平降至恒温MEMS加速度计的零点漂移水平,大约降低了一到二个量级,水平向的补偿结果好于垂直向。另外,通过低通滤波器的使用,力平衡加速度计的低频段噪声亦有了改善,实现了温度补偿的目标；2、提出了一种采用GPS与NTP联合对系统进行授时服务的算法。在GPS信息存在的情况下,整个系统的授时精度可以达到±10μs；在单独使用NTP授时的情况下,系统的授时精度可达到±2ms。系统经过长期的运行测试,结果表明,使用这种算法可以有效解决在单独使用GPS授时服务时出现的“跳秒”现象和个别地方GPS天线架设不便的问题,而且授时精度也能够满足强震观测台站对时间服务的要求；3、研究出了新型的实时嵌入式Linux内核数据处理系统,对实时数据流缓存机制和组织管理结构进行了全新设计,将传统以秒为单位获取实时数据流的方式修改为根据用户指定数据帧长度提供内核实时数据流访问服务,有效提升了预警时间,最多可增加0.9s；通过将原来在单独的数据处理芯片上实现的实时FIR滤波和抽取运算移植到ARM9CPU上,与其它的系统功能一起共享CPU,并采用优化后的嵌入式ARM汇编来实现,有效降低了整机系统的成本和负载；4、研究出了一种新型的网络实时数据服务方式,即快速数据服务,可以按照100ms量级的数据包格式进行实时数据传输,并设计出了一种新型的实时数据解压缩算法用以解决由于以更低的延时进行数据传输带来的更大的传输数据量的问题。通过与现有世界地震行业应用最广泛的Steim2压缩算法相比,本文设计的压缩算法在各种类型的32位数据压缩方面都具有一定的优势,尤其是在编码长度超过20位后,本文的压缩算法不会出现Steim2算法中发生的数据扩展现象,仍然具有一定的压缩比;5、依托汶川主震及其余震的强震动观测数据,统计分析了特征周期τ。与震级的关系、P波前3s位移幅值Pd与PGV的关系以及τc×Pd与震级的关系。统计结果表明,两种预警参数不仅可以单独应用,也可以联合起来进行破坏性地震和非破坏性地震判别,并指出在实际应用时,考虑到我国的建筑物情况,可以适当降低τ。×Pd阈值,以便于更好地应用到我国的地震预警系统中。最后将这些统计结果应用到了一体化强震仪中,研发出了一套实时地震预警数据处理系统,可以直接用于“现地”地震预警。更多还原

【Abstract】 Earthquake early warning (EEW) is a new technology and new measure for seismic hazard mitigation in last twenty years and has the potential to reduce fatalities, casualties and costs. Its core technology is to shorten the data processing time in all aspects, and combine with the principles that the P-wave (traveling at about6-7km/s) is faster than S-and surface waves (traveling at about3.5km/s or less) and information (which travels at the speed of an electromagnetic signal-about300,000km/s) is much faster than seismic waves (which travel at speeds of the order of a few km/s) so that it can provide few seconds to tens of seconds of advanced warning time for impending ground motions, allowing for mitigation measures to be taken in the short time. In order to obtain high timeliness for earthquake information, we need to break from every aspect for shortening data processing time of the system, increasing lead-time and reducing the range of "blind zone". From the point of view of earthquake instruments for real-time data acquisition, there is plenty of room for improvement, such as decreasing network latency of real-time data transimission, shortening the data transmission interval, providing network support for the vast majority of existing strong-motion observation stations, introducing digital correction, resolving "jumping seconds" phenomenon and problems of inconveniently installing GPS antenna in some places, etc. In this paper, these aspects are studied and the results are applied to a real-time embedded Linux system in the newly developed strong-motion seismograph in which an onsite-warning EEWs based on a single instrument is developed. The contents are summarized as follows:1. Digital correction techniques have been introduced into the integrated strong-motion seismograph. Fifteen different position data measured by Gravimetric Method are used to calculate sensitivity and installation azimuth deviation of the accelerometer and the results are applied to the embedded software system for real-time data correction, effectively improving the precision and accuracy of real-time data. On the basis of previous work, a new temperature compensation algorithm for accelerometer has been proposed. This algorithm uses the low frequency band (10-4Hz) of the thermostatic MEMS accelerometer data to compensate the force-balance accelerometer in real-time and MEMS thermostat point migration is adopted to dynamically adjust the thermostatic MEMS temperature, which can expand application range of the instrument and reduce system power. The test results show that the zero-drift level of the compensated force-balanced accelerometer reduces to the level of the thermostatic MEMS accelerometer, approximately one or two orders of the results without compensation, and the compensation results of the two horizontal components are better than the vertical component’s. In addition, the low-band noise of the force-balanced accelerometer is also improved through the use of low-pass filter and the goal of temperature compensation is achieved.2. An algorithm combining GPS and NTP for system timing service has been proposed. The timing accuracy can reach±10μs when existing GPS information, while using only NTP the timing accuracy can reach±2ms. The results obtained from long-term test show that this algorithm can effectively resolve the "jump second" problem when using only GPS for timing service and the inconvenience settiing up GPS antenna in some places, while timing accuracy can meet the requirements of the strong motion observation stations on time service.3. A new real-time data processing system in embedded Linux kernel has been developed, in which the caching mechanism and the organization and management structure have been newly designed and the mode for accessing real-time data stream has been changed from by a packet per second into by the user-specified data frame length, effectively increasing lead time to a maximum of0.9s. By porting the real-time FIR filtering and decimation operations originally realized in a separate data processing chip to the ARM9CPU which is shared with other system functions, and realizing these operations in optimized embedded ARM assembly language, the cost of the whole system and the CPU load have been effectively reduced.4. A new method for real-time network data service, namely "rapid data service", has been developed. According to this method, the real-time data can be transferred by the order of100ms. In order to solve the problem of more amount of data transmitted which is caused by the low-latency data transmission, a new algorithm for real-time data compression and decompression has been proposed. By comparing with the Steim2compression algorithm which is most wided used and has the best compression ratio recognized by the world seismic industry now, the newly designed algorithm has advantages in various types of32bit data compression. Especially in the code length of more than20bits, the compression algorithm will not appear the data expansion phenomenon existed in the Steim2algorithm and still has a certain degree of compression ratio.5. Based on the strong motion data of Wenchuan earthquake and its aftershocks, we have statistically analyzed the relationship between characteristic period (τc) and magnitude, the relationship between the displacement amplitude of the first3s of P wave (Pd) and the peak ground velocity (PGV), the relationship between τc×Pd and magnitude. The results show that these two early warning parameters can not only be applied independently, but also be joined together to discriminate damaging earthquakes from non-damaging earthquakes. In practical applications, taking into account buildings in China, the threshold of τc×Pd should be appropriately reduced in order to be applied to China’s earthquake early warning systems better. Finally, these results have been applied to the integrated strong motion seismograph and a real-time EEW data processing system has been developed which can be directly used for "On-site" EEW.更多还原