Neural network: An alternative to pile driving formulas

Abstract

Artificial neural networks are capable of learning complex nonlinear relationships from a large amount of accumulated data, and similar to human brains, are noise and fault tolerant. This unique capacity suggests that neural networks would be very useful in certain geotechnical engineering applications. A back-propagation network is set up and trained to predict the pile bearing capacity from dynamic testing data. The trained network produces better results than a pile driving formula approach. The effects of various network parameters on the network results are examined in detail. The general understanding developed is potentially useful for the application of neural networks in other geotechnical engineering problems. © 1995.