Computational & Technology Resources
an online resource for computational,
engineering & technology publications
PROCEEDINGS OF THE SIXTH INTERNATIONAL CONFERENCE ON THE APPLICATION OF ARTIFICIAL INTELLIGENCE TO CIVIL AND STRUCTURAL ENGINEERING
Edited by: B.H.V. Topping and B. Kumar
Neural Identification for Critical Flutter Load of a Polar-Orthotropic Annular Plate
Department of Mechanical Engineering, Kanagawa Institute of Technology, Atugi-shi, Kanagawa-ken, Japan
I. Takahashi, "Neural Identification for Critical Flutter Load of a Polar-Orthotropic Annular Plate", in B.H.V. Topping, B. Kumar, (Editors), "Proceedings of the Sixth International Conference on the Application of Artificial Intelligence to Civil and Structural Engineering", Civil-Comp Press, Stirlingshire, UK, Paper 17, 2001. doi:10.4203/ccp.74.17
Keywords: inverse problem, neural network, vibration, plate, critical flutter load.
The increasing interest in minimum weight designs for industrial fields such as in mechanical, aerospace and rocket engineering has generated substantial interest in the analysis of vibration and stability of plates. In particular, the inverse problems of continuous bodies are becoming important in design practice.
The identification technique for support conditions of continuous bodies is becoming important, with the increasing size and complexity of machines and vessels. Recently, Yasuda and Goto and Kamiya, et al. proposed the experimental identification technique for boundary conditions of the beam. Saito, et al. presented the identification of non-linear support systems by using transient response. Takahashi proposed the identification method for the axial force and boundary conditions of a beam using the neural networks. Takahashi[5,6] also studied the neural identification for the axial force or critical speed of a plate.
In this paper the possibility of using a Multilayer Perceptron Network trained with the Backpropagation Algorithm for detecting the critical flutter loads of tapered annular plates is studied. The considered model is a polar-orthotropic plate, using a transfer matrix method, to estimate the changes in various modal parameters, caused by the shape parameters and support condition of plates.
The input data to train a neural network is very important. Kudva et al. and Worden et al have presented the fault identification in a strucutral element using the neural network which was trained on the strain data. On the other hand, the natural frquency is global and more stably measurable than the strain. Therefore, we use it as the input data.The concept used in this study is the same as that discussed earlier[4,5,6].
The basic idea is to train a neural network with simulated patterns of the relative changes of natural frequencies and corresponding critical flutter load and one parameter of plates in order to recognize the behaviour of the structure. Subjecting this neural network to un-learning natural frequencies should imply information about the critical flutter load and one papameter of plates.
The transfer matrix method is applied to plates with linearly varying thickness , and the natural frequencies are calculated numerically, to provide information about the effect on them of varying thickness, rigidity, intermediate support and radius ratio. We will discuss the four effects of the taper ratio, rigidity, intermediate support and radius ratio on the critical flutter load of the tapered plate subjected to a constant follower force at the free edge.
purchase the full-text of this paper (price £20)