Assistant Professor, Faculty of Civil Engineering, Babol Noshirvani University of Technology, Iran
Tuned liquid column gas damper is a new type of energy absorber that can mitigate the vibrations of structures if their frequency and mass parameters are well tuned. Since this damper has recently been introduced and its behaviour in certain structures such as offshore oil platforms and wind turbines has already been tested, a suitable and accurate method is required to identify these optimal parameters. Therefore, considering the complexity of loads exerted on wind turbines in seas (wave and wind loads), in present study attempts are made to use a new artificial neural network approach to obtain optimal tuned liquid column–gas damper (TLCGD) parameters for mitigation of wind turbine vibrations. First fixed offshore wind turbines at various depths are designed in the MATLAB coding environment. After obtaining the stiffness, damping and mass matrices of the structures, the program enters the Simulink, and the wind turbine structure along with the TLCGD is exposed to different wave-wind load combinations within reasonable range of damper parameters. The neural network training is launched based on available statistical data of the offshore wind turbine with different heights as well as different frequency and mass ratios of the damper. According to this method, the percentage of errors found in the neural network outputs was negligible compared to the actual results obtained from the analysis in Simulink (even for inputs that stood outside the training range of the neural network). The mean error percentage, the standard deviation and the effective value of the neural network with actual values are below 10% for all three types of the structure. Finally, the method presented in this study can be used to obtain optimal parameters of the TLCGD for all kinds of offshore wind turbines at different depths of the sea, which leads to the optimal design of this damper to reduce the vibrations of wind turbines under wave and wind load pressures.