A neural network for error prediction in a true triaxial apparatus with flexible boundaries

The true triaxial apparatus with flexible boundaries (TTAF) is a versatile tool for characterizing the stress-strain behaviour of soils. The evaluation of the deformations suffered by soil in a TTAF must account for the effects of the sample membrane and of the pressurizing cushion, the non-uniform spatial stiffness of the frame of the apparatus, and the electronic errors. In order to overcome the difficulty associated with such a complex set of factors of influence, a neural network (NN) approach was used to predict the errors that may affect the displacements. The current performance of the prediction system is reported.