Load Identification For A Composite Laminated Shell Using Radial Base Function Neural Network And Genetic Algorithm

The smart structure is one self-sensing and self-adaptive structure that can apperceive the environmental changes, and it can aim at this kind of changes to make the suitable response. The development of the smart structure has initiated many research areas, and the load identification of the smart structure is one of them. Firstly, this article proposes a method that combines the finite element and the radial base function (Radial Base Function, concisely RBF), and models the electric charges response of the plate/shell structure withstanding the concentra- ted loads of different size and position through the finite element analytical method, then takes the electric charges value to carry on the normalization and takes them as the training samples of the neural network to identify the load size and the position. The RBF neural network has a stronger mapping function of the input and output, and it is mainly applied in the pattern recognition and the function approximation. Its prominent characteristic is that the neuron′s number of hidden layer does not need to be determined before the application and the optimized neuron′s number of hidden layer may be produced automatically under the training algorithm study that proposed in this article. Moreover, this article also proposes a method that combines Genetic Algorithm and the finite element, and calculates the electric charges response of the plate/shell structure withstanding the concentrated load through the finite element analytical method to design the fitness function. Finally,it completes the identification of the load size and the position through the Genetic Algorithm again. The Genetic Algorithm is one kind of global optimization and self-adaptive probability search algorithms which is developed by profiting from biological natural selection and the genetic evolution mechanism, and it produces a new generation by exerting the selection,cross and mutation and so on a series of genetic operations to the current generation,and gradually causes the generation to evolve to the condition which contains or approaches the optimal solution. When analyzing the structure by the finite element method, this article uses enhanced assumed strain(EAS) to develop a solid shell element formulation for analysis of under piezoelastic couplings. This element not only can be used as solid element but also can be used to model thin curved shell structures. Even for a thin plate/shell with very small thickness to length ratio, the predictions of this element are satisfactory.