Embedded Machine Learning System Design for Post-Earthquake Structural Health Assessment Deprem Sonrası Yapı Sağlığı Değerlendirmesi için Gömülü Makine Öğrenmesi Sistemi Tasarımı

Tusun M. E., Metin E. M., Tigrel F., Ünsalan C.

32nd IEEE Conference on Signal Processing and Communications Applications, SIU 2024, Mersin, Türkiye, 15 - 18 Mayıs 2024, (Tam Metin Bildiri)

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Doi Numarası: 10.1109/siu61531.2024.10600859
  • Basıldığı Şehir: Mersin
  • Basıldığı Ülke: Türkiye
  • Anahtar Kelimeler: earthquake, embedded system, frequency response, structural health monitoring
  • İstanbul Medipol Üniversitesi Adresli: Hayır

Özet

Many buildings are damaged in earthquakes that occur in our country and these structures collapse in subsequent severe earthquakes, causing thousands of people to lose their lives. The structures that do not collapse continue to pose a great risk. Many tests such as cross-sectioning of columns are carried out for the detection of damaged structures. However, these tests take a long time and are quite costly. They also do not provide continuous monitoring of the structure. Although accelerometers are used to continuously monitor the condition of the structure, accelerometers are inadequate at the low frequency of earthquake waves. In this study, a sensor with strain gauge technique optimized for earthquake waves and structure vibrations and a system where the collected data is automatically evaluated on a low-power microcontroller is developed. Fourier transform is applied to the vibration data measured by the sensor to obtain the frequency response of the structure. There is a shift in the frequency response of the damaged structure. By detecting this shift, the structure is classified as intact, slightly damaged and very damaged. This classification is done with a deep learning model running on a microcontroller. The results show that the developed sensor is more successful than an accelerometer in assessing the health of the structure and the assessment process can be performed on the microcontroller using Fourier transform and machine learning.