System identification of flexibly mounted cylindrical pipe due to vortex induced vibration

Abstract

This paper presents a system identification for an elastically supported rigid cylinder in one degree of freedom direction undergoes vortex induced vibration (VIV). The cylinder is constrained to oscillate in perpendicular direction with respect to a free stream. The input-output data of the vibrating system are recorded by using accelerometers in capturing the cylinder movement in certain period of time. Both developed Recursive Least Square (RLS) and Adaptive Neuro Fuzzy Inference System (ANFIS) modus are utilized in order to model an offline dynamic system response of the cylinder pipe at a particular speed. The Auto-regressive external input (ARX) model is chosen to represent the model structure of the system and the performances of both modus are compared in term of Mean Squared Error (MSE). Both methods are capable in predicting the system response. However, the research revealed the superiority of ANFIS in modeling the system with the lowest mean squared error of 3.411×10 -10 as compared to RLS.