Parametric Modelling of Fibre-Concrete Interface and Prediction of Compressive Strength of Concrete With Waste Tyre Steel Fibres Using Artificial Neural Network Model

The utilisation of concrete fabric materials and predicting mechanical properties using the artificial neural network (ANN) model in structural engineering has increased in recent years. However, the current practice of predicting the compressive strength of concrete at 28 days using early compressive strength at 3 and 7 days is based on concrete without fibres. In addition, the application of fibres in concrete is based on random mixing into the matrix without focusing on the interfacial properties of the fibre and concrete. The study used Abaqus to develop a parametric model of the fibre-concrete interface. The model utilised different fibre lengths (30, 50 and 60 mm), and the interface properties were deduced from shrinkage strains and available shrinkage models in the literature. The ANN model utilised fibre length, fibre content, and compressive strength at 3- and 7-curing days as the primary inputs to predict the compressive strength of high-strength concrete at 28 days of curing. The model had a single hidden layer with seven (7) neurons, and the Levenberg-Marquardt training algorithm was used in the training. The results show that the prediction was close to the experimental results, with a mean squared error, coefficient of determination (R2) and a20-index of 0.0194, 0.961 and 0.963, respectively. For 60 mm fibre length, the finite element model of the interface showed that the maximum shear force of 649.6 N is generated when the interface is displaced/cracked by less than 0.15 mm, signifying the contribution of the fibre in the concrete matrix on carrying loads. All inputs in the ANN model significantly impact the predicted results. However, the compressive strength in 7 days had more influence.