Boundary Condition Identification of Fluid-Filled Piping Systems Using Neural Networks

Abstract

An online identification approach for boundary condition identification of fluid-filled piping systems is developed. Considering the lateral vibration of the fluid-filled pipes, the method combines the traveling wave method and the BP (backpropagation) neural network to estimate the boundary parameters. The traveling wave method is used to generate the training samples that contain several lower natural frequencies as inputs and boundary parameters as outputs for the BP neural network. Thus, the relationship of the natural frequencies and the boundary conditions of the piping system is established in the BP neural network. When the experimentally measured natural frequencies are put into the BP neural network, the boundary conditions can be identified immediately. An experiment is presented to demonstrate the feasibility of the proposed method. In the experiment, the four lowest natural frequencies of the lateral vibration of a fluid-filled straight pipe are measured. The boundary mass and stiffness are identified. Results show that the approach is efficient and precise.