Suction caissons are frequently used for the anchorage of large compliant o
ffshore structures. The uplift capacity of the suction caissons is a critic
al issue in these applications, and reliable methods: of predicting the cap
acity are required in order to produce effective designs. In this paper a b
ack-propagation neural network model is developed to predict the uplift cap
acity of suction foundations. A database containing the results from a numb
er of model and centrifuge tests is used. The results of this study indicat
e that the neural network model serves as a reliable and simple predictive
tool for the uplift capacity of suction caissons. As more data becomes avai
lable, the model itself can be improved to make more accurate capacity pred
iction for a wider range of load and site conditions, The neural network pr
edictions are also compared with finite element based predictions. (C) 2001
Elsevier Science Ltd. All rights reserved.