Vibration Control of a Pier Pile Using Deep Learning LSTM Network


Creative Commons License

Namlı B., Bayındır C.

ICAME 2021, Balıkesir, Türkiye, 1 - 03 Eylül 2021, sa.111, ss.1-6

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: Balıkesir
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.1-6
  • İstanbul Teknik Üniversitesi Adresli: Evet

Özet

Balancing the offshore structures in the ocean or sea against the forces created by seismic movements and waves is extremely important for the serviceability and the safety of the structure. In order to achieve this, various approaches are currently being considered [1-4]. Region-specific parameters should be recorded and used when needed to assist these approaches. In addition, the use of artificial intelligence methods, which have been in demand recently, to reduce the oscillation in buildings, allows more effective results in the application that can be done in this field [5-7]. Using the long-short-time memory (LSTM) algorithm [5], one of the deep learning methods, time series prediction can be performed. As a result of the prediction, better approaches can be developed for the future. This study shows that the vibration control of offshore platforms can be achieved against various types of loadings by the deep learning techniques, which is a branch of artificial intelligence. For this purpose, a long pile is analyzed by solving the equation of motion under forcing described by the Morison equation [8-9]. Thus, a realistic wave load is applied to analyze system behavior in a more realistic setting. The applied wave loads are predicted using the LSTM deep learning network and applied to the system as negative feedback. It is shown that a significant reduction in the vibration amplitudes can be achieved by this approach. Our findings and their possible applications are also discussed.