Research On Strong-motion Data Processing Technology Oriented To Earthquake Early Warning

In a specific area, earthquake events are rare, abrupt, unpredictable and devastating. With the strong-motion network increasing its density and real-time transmission capacity, earthquake early warning and alarm system based on strong-motion observation become an important mean of earthquake disaster reduction and grow rapidly in recent years. My research focus on data processing technology for earthquake early warning system based on strong-motion data. This paper mainly completed the following research work:1. This dissertation studied the characters of noise in earthquake early warning and alarm system. Used spectrum analysis to determine the self-noise of strong-motion seismograph and compared with direct observation. The results show that in ordinary circumstances a reliable instrument self-noise can be determined using spectrum analysis. This dissertation studied the causes of baseline drift in strong-motion records. Results show that in addition to tilt, hysteresis loss is an important cause of baseline drift.2. Environmental noise could cause severe influence to earthquake early warning and alarm system and result in false alarm. The anti-false-trigger algorithm for earthquake early warning system in single station is studied. After extraction of multiple features of the first 3 seconds P wave, relationship between features and signal source are trained using decision tree which then could be used to distinguish between seismic event and other events. EMI interference, quarry blasting, train vibration and heavy machine vibration records verified the trained decision tree. For a station equipped with multi sensors, a method utilize the correlation between sensors to further eliminate interference was proposed.3. The method of noise reduction for earthquake early warning and alarm was studied. In single measure point situation, compared the effects of IIR low pass filter and wavelet methods. The wavelet analysis has unique merits, but the computation quantity is bigger than the filtering method, this paper chooses the filtering method to reduce noise. In high speed railway, which have dual sensor configuration in each earthquake observing station, independent component analysis method achieved better effect than the filtering method in removing vibration from high speed train.4. This dissertation proposed a ―synthesis weight‖ method comprehensively utilizes azimuth, phase arrival time, faults location and earthquake activity information to improve the accuracy of earthquake locating. And proposed a multi-objective optimization method to improve its computation speed. Compared three optimization methods and the results show DEMO method is best in speed and accuracy. This dissertation used many parameters in artificial neuron network to estimate earthquake magnitude and used massive strong-motion records to assess its error. The results show it could satisfy the need of earthquake early warning.5. This dissertation studied the realization technology of the earthquake early warning and alarm system. Presented system architecture, system development process, key technologies and system verification technology. The developed systems have been used in several high-speed rail lines.