الفهرس 
Chapter 1: IntroductionOnly 14 pages are availabe for public view
 |  | from | 96 |  |  |
| | | from | 96 | | |
Abstract
With cutting edge technology of scientific applications especially, wide-ranging seismic surveys or earthquake monitoring mainly depends on Analog to Digital Converters (ADCs). In particular, the Egyptian National Seismic Network (ENSN) has more than 100 stations distributed in the whole area of Egypt. The main objectives of the ENSN are recording, monitoring, and analysing the natural and artificial seismic events. The basic components in each of the ENSN stations are the sensing element (seismometer) and the digitizer. A seismometer is an instrument that responds to ground motions, which may be caused by earthquakes, volcanic eruptions or explosions. The output of such device is an analog signal and is converted to digital signal using the digitizer.
Motivation for prototype implementation of this digitizer is to provide the first Egyptian seismic digitizer to be competitive with the international markets. Moreover, enhance the performance of the digitizer by increasing the signal to noise ratio with reducing the computational complexity. Also, increase the compression ratio of transmitted data will be additive value for this product. In return this will save a huge budget which was reserved for the used bandwidth, maintenance and upgrading digitizers used in ENSN. We suggest adding a picking module to the implemented digitizer to detect earthquakes. This is called triggered digitizer, which allows it to release an alarm. Once an earthquake is detected which can support the earthquake early warning system (EEWS).
In this thesis, we design and implement a digitizer for acquiring the seismic signal from three sensor components. The implemented digitizer consists of several blocks, i.e., the power source, the front-end circuit, analog to digital converter (ADC), GPS receiver, and microprocessor. Three finite impulse response (FIR) filters had been used to decimate the sampling rate of the input seismic data according to user needs. The incoming data is converted into standard seismological MiniSEED format for easy data archiving and streaming. The data is streamed between seismic station and the main centre using SeedLink protocol over TCP/IP. This protocol ensures data transmission without any losses as long as the data still exist in the ring buffer. In addition, deep learning techniques are proposed for seismic data compression and picking.
The compression techniques will lead to an efficient use of the bandwidth assigned for the communication link between the seismic stations and the main centre. In this thesis, two convolutional auto encoders (CAEs) are proposed for seismic data compression. The two algorithms are mainly based on the convolutional neural network (CNN), which has the capability to compress the seismic data into feature representations, thereby allowing the decoder to perfectly reconstruct the input seismic data. The results show that the first model is efficient at low compression ratios (CRs), while the second model improves the signal-to-noise ratio (SNR) from about 3 dB to 12 dB compared to the other benchmark algorithms at moderate and high CRs.
The implemented digitizer is a trigger device, where a deep learning module is implemented to pick the first arrival time of the event. We use the Capsule Neural Network for Seismic Phase Classification and Picking (CapsPhase) method as a picker module. CapsPhase achieves an accuracy of 94.77% in comparison of the STA/LTA, the MODWT, and the spectro-ratio which achieve an accuracy of 76%, 83%, and 87%, respectively. The system has a friendly user interface, which can monitor the seismic waveform in real time, realize seismic data receiving, process and adjust the parameters of acquisition unit. Finally, the prototype achieves a reasonable performance when tested in a station within the Egyptian National Seismic Network (ENSN) compared to the calibrated digitizers in different stations.