Buckling load prediction of laminated composite stiffened panels subjected to in-plane shear using artificial neural networks

• Use of neural network as analysis tool for predicting the buckling load of composite stiffened panels subjected to in-plane shear loading.
• The parameters D1/D2, (EA)s/(EA)p, (A11/A66)p and D1D2/D3 are identified as the parameters which affect the in-plane shear buckling load.
• Development of computationally efficient analysis procedure in the form of ANN with architecture 4-8-4-1 was found to give sufficiently accurate results.
• ANN based analysis tool is validated with FE results and the generalization capabilities of the tool are demonstrated to use in optimization tools with confidence.
• This neural network, architecture and weights can be used in the optimization routines for designing better laminated composite stiffened panels.

Stiffened panels are basic building blocks of weight sensitive structures. Design of laminated composite stiffened panels is more involved and requires the use of an optimization approach, which needs a computationally efficient analysis tool. This paper deals with the development of an analytical and computationally efficient analysis tool using artificial neural networks (ANN) for predicting the buckling load of laminated composite stiffened panels subjected to in-plane shear loading. The database for training and testing is prepared using finite element analysis. Studies are carried out by changing the panel orthotropy ratio, stiffener depth, pitch length (number of stiffeners). Using the database, key parameters are identified and a neural network is trained. The results shows that the trained neural network can predict the shear buckling load of laminated composite stiffened panels accurately and will be very useful in optimization applications where computational efficiency is paramount.