Estimation of parameters for constitutive modelling of recycled aggregate concrete using artificial neural networks

Purpose This study aims to improve the performance of the existing constitutive model of recycled aggregate concrete (RAC) by estimating the parameters used in the model using artificial neural networks (ANN). Design/methodology/approach The use of RAC as structural concrete is gaining importance as it contributes towards several sustainable development goals (SDGs) defined by the United Nations. Several parameters are required to model the complete behaviour of concrete under multiaxial loading. The existing RAC model proposed by the authors requires four parameters to define the complete stress–strain curve. Concrete compressive strength ( <m:math xmlns:m="http://www.w3.org/1998/Math/MathML"><m:mrow><m:msubsup><m:mi>f</m:mi><m:mi>c</m:mi><m:mo>/</m:mo></m:msubsup></m:mrow></m:math> ) and Modulus of elasticity are the most important properties of concrete used in the design of concrete structures, whereas α and β are the other two parameters to control the damage growth rate and to capture the behaviour of RAC at respective peak stress levels. ANNs were trained to estimate these parameters by using the data available in the literature. Proposed ANN models were first validated and then used in the constitutive model to estimate these four parameters. Findings Proposed ANN models accurately estimated the parameters needed to improve the predicting capabilities of existing RAC constitutive models. The overall performance of the constitutive model in terms of peak stresses improved with the use of parameters predicted by ANN models. Originality/value ANN has been used for the estimation of parameters that directly influence the behaviour of RAC under multiaxial states of stress instead of conventional regression techniques.