Abstract:
Earthquakes can occur suddenly and unexpectedly, causing widespread damage and loss of life.
For example, the 2023 Syria-Turkey earthquake resulted in the death of over 50,000 people.
Primary Wave (P-wave) and Secondary Wave (S-wave) are two types of seismic waves generated
by earthquakes. P-wave is faster than S-wave, However, S-waves can do more damage than P-
waves, and P-waves are used to detect the earthquake. Early earthquake warnings can help reduce
these damages by giving people time to protect their lives. This leads us to propose a highly
accurate, cost-effective, and accessible earthquake early warning system.
The system utilizes a wireless sensor network to collect data from the ground using acceleration
sensor ADXL335 and pass it to the ESP32 microcontroller to detect and predict earthquakes using
a logistic regression machine learning model. The system also includes an alert mechanism using
buzzers capable of notifying users of a near earthquake and providing audible alerts on mobile
applications to increase speed and allow users sufficient time to take safety measures. Finally, the
system uses a shaking table that significantly simulates earthquake conditions for evaluation and
testing purposes.
After picking 6000 earthquake and not-earthquake acceleration samples for x, y, and z from the
shake table, these samples are divided and used to calculate the mean and standard deviation, each
100 samples together will give one vector of x, y, and z standard deviation values, and number of
samples after calculating stand deviation is 43 per second, these sample will be used later to train
and test the logistic regression model, that gives us a very high accuracy of 95.3%, which mean
predict the earthquake in 0.5-1 sec. In the WSN, we implement star topology of two nodes, the
100% confidence of an earthquake happening depends on both nodes' values of alert, and the time
arrival of alert to the mobile application depends on the internet latency, the total time needed from
detecting the earthquake to mobile alerting is 7.5-12.5sec. mobile application is designed in a way
that achieves easy and simple user interaction.
