GEAKDES: Gerçek Zamanlı Deprem Afet / Süreç Yönetimi İçin Yapay Zekâ Temelli Akıllı Karar Destek Sistemi

TÜBİTAK Projesi, 1001 - Bilimsel ve Teknolojik Araştırma Projelerini Destekleme Programı, 2022 - 2024

Proje Türü: TÜBİTAK Projesi
Destek Programı: 1001 - Bilimsel ve Teknolojik Araştırma Projelerini Destekleme Programı
Başlama Tarihi: Şubat 2022
Bitiş Tarihi: Şubat 2024

Earthquakes are natural disasters that frequently occur worldwide and have serious

consequences. Modern technology, especially in seismically active regions, allows for real-

time seismic measurements, enabling rapid intervention. Quick and accurate damage

assessment after earthquakes ensures effective management of emergency aid and rescue

operations. Earthquakes cause significant economic and human losses globally, particularly

posing a threat in seismically active areas. Building strengthening efforts and disaster

prevention plans can enhance communities' resilience to earthquakes.

Machine learning and artificial intelligence have significant applications in earthquake-related

areas. These technologies are used in earthquake damage prediction, seismic activity

forecasting, and building strengthening strategies.

The GEAKDES project provides an integrated disaster decision support system. Real-time

machine learning algorithms perform earthquake damage prediction using characteristic

features such as building, earthquake, and ground data. This information is combined with

satellite image analyses to achieve more accurate earthquake damage predictions.

Additionally, post-earthquake aid needs are identified, and a logistics network model is run to

determine aid routes.

The Cost-Sensitive Parallel ABC-ANN and Cost-Sensitive Parallel GA algorithms developed

within the project stand out for their high accuracy and fast training times in earthquake

damage prediction. Sentinel-2 and Sentinel-1 satellite images are used for post-earthquake

damage detection, with optical images identifying building collapses and SAR images

detecting ground changes. Integration of this information results in a 91% accuracy rate.

The use of open-source Sentinel-1 SAR satellite images, integrated with machine learning

methods, contributes to understanding earthquake-induced damage. GEAKDES models a

logistics network for earthquake region aid delivery planning using damage prediction

information. Routes are calculated using the MM-CSA approach, and aid distribution routes

are determined in pilot areas through the Substitute Product Strategy. The project aims to

share acquired knowledge and experiences for the benefit of humanity.