Prediction of parallel clay cracks using neural networks – a feasibility study

Cracking in drying clay soil is a common phenomenon especially in arid and semi-arid regions. Proper understanding and reliable prediction of the extent and nature of cracks in clay is vital for the design and construction of geo-infrastructures. While many models have been developed over the years to predict cracking, they are focused on a single crack rather than the whole network. This paper presents a feasibility study on a novel intelligent approach based on artificial neural network to predict the number of cracks in soil for a given combination of input parameters. Initial moisture content, specimen layer thickness and size of the specimen are used as inputs to the model. The output is the number of cracks. The collected database is used to train, validate and optimise the neural network models. The optimisation steps are discussed and analysed as the predicted number of cracks are compared to the experimental ones. A reasonable agreement was found between the experimental and predicted data. The results indicate that the model can be further improved to make more reliable predictions.