Optimization Of Stability Of Laminated Composite Shell Based On BP Neural Network

Composite cylindrical shell has been widely applied in engineering applications. The more complex and extreme the working condition is, the better it demonstrate the excellent properties of composite materials. Composite cylindrical shell structure is so common in the aerospace, marine environment, energy, transportation and other fields. Followed by a series of mechanical problems, stability analysis is an important part.A flexible composite cylindrical shell is taken as the research object. The characteristics of shell buckling stability are concluded from the buckling analysis. The first order buckling load which is obtained from the buckling analysis is described as the optimization index and the optimization model is constructed by BP neural network algorithm. The largest critical buckling load is calculated with this method. It mainly includes the following:The first order buckling load is different for different design parameters of composite cylindrical shell. In order to obtain the optimal structure, it takes the ply angles as the main design parameters, using orthogonal design method to obtain test input data. The buckling analysis of composite cylindrical shell is simulated with the large-scale finite element software ABAQUS, access to the critical buckling load of specific structural type, at the same time; it gains the experimental data samples.BP neural network algorithm is introduced, taking test data samples as training samples. Network structure contains the input layer, hidden layer, output layer, the transfer function, the weights and threshold. Neural network is trained by adjusting parameters of the neural network; the training will be over when it comes to certain accuracy. The parameters of the trained neural network were extracted to obtain the objective function of the first order buckling load. With the specific form of the objective function, appropriate mathematical method is used to seek optimal parameters of the structure.The dissertation used the finite element method and neural network to optimize the first order buckling load of composite cylindrical shell. This optimization design method will play a future role in promoting structure optimization of composite cylindrical shell.