Prediction of Modal Response of Towers Using Artificial Neural Networks

Yucel O. B., Demir B., Ates H. I., ALDEMİR A.

Electrical Transmission and Substation Structures Conference 2022: Innovating for Critical Global Infrastructure, Florida, United States Of America, 2 - 06 October 2022, pp.393-403 identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1061/9780784484463.033
  • City: Florida
  • Country: United States Of America
  • Page Numbers: pp.393-403
  • Middle East Technical University Affiliated: Yes


© 2022 American Society of Civil Engineers.Overhead transmission line towers may be subjected to a severe level of earthquake-related forces if the line lies in a seismically active region. However, the traditional practice in the industry generally overlooks the effect of seismic forces on the structural demand in parallel with universally valid design codes claiming that the base shear that emerges on an earthquake would generally be less than wind or wind + ice combined conditions. On the other side, in recent years, a rising number of scientific articles in relation to the possibility of severe consequences of an earthquake on the components of a transmission line (e.g., towers) have been published. Additionally, in late 2019, the Ministry of Environment and Urbanization of the Turkish Republic put a comprehensive earthquake code along with a new seismic map of the country into effect, also covering transmission line towers. Preliminary studies revealed that earthquake-related forces might be critical for specific components of towers, especially in case the line was located close to the fault zones. In this regard, modal analyses of towers have become a must for the industry of transmission line engineering in some countries. In addition, even the most common commercial software packages used in tower analysis and design do not have the option to perform the modal analysis. In order to determine the modal response of towers, the finite element model (FEM) of the tower or whole tower family must be generated or be transferred to an alternative FEM software, which is a cumbersome procedure. At this point, artificial neural networks (ANN) can be a practical solution. In brief, this study aims to construct and train an ANN architecture in order to determine the relationship between the structural parameters related to the mass and stiffness of the tower with first mode frequencies. In this way, required modal response data of the considered tower can be obtained rapidly with a high accuracy rate.