Neural network modelling of creep in masonry

Stresses and deformations in concrete and masonry structures can be significantly altered due to creep. However, accurate prediction of creep is difficult due to its dependency on a large number of parameters (e.g. section geometry, relative humidity, stress level, age of loading). This paper introduces a new method based on artificial intelligence to model creep of masonry. Feedforward artificial neural networks (ANN) are investigated as a modelling technique for predicting creep. Experimental data for creep of structural masonry are used to develop the networks. Changes in network architecture are examined to produce prediction models. Fifteen networks are developed and analysed statistically. Creep models with accuracy in the range ± 15% are attainable using ANN.