Force Finding Of Suspended-Domes Using Back Propagation (Bp) Algorithm

Force finding is a key step when designing a suspended-dome. To give a general simple method for force finding, this report accomplished research on the application of a BP network to force finding of a suspended-dome under a certain tension process. First, this paper gives three typical states of the construction process, and then, it selects the internal force in the hoop cables at the initial state as the input data and the initial strain of the active element at the zero state as the output data. Then, a three-layer BP network was developed for force finding of the suspended-dome model. Second, this paper presents a method for restricting the training set range, in which 1000 sets of patterns were generated by the finite element software ANSYS. Then, we randomly selected 19 groups of training data from 1000 sets of samples to train the BP network and predicted the results of force finding. Last, this paper combines GA with BP to predict the results of force finding. The results show that the BP network can solve the force-finding problem accurately and effectively when the training samples are sufficient. The prediction stability of the BP network and the prediction precision can be significantly enhanced after the initial weights and the thresholds are optimized by a GA.