EXPERIMENTAL AND ARTIFICIAL NEURAL NETWORK BASED MODELLING APPROACHES FOR STRENGTH PREDICTION OF CONCRETE INCORPORATING RECLAIMED ASPHALT PAVEMENT AND RICE HUSK ASH

Abstract

Construction industry is exhausting natural resources thereby posing environmental problems. On the other hand, solid waste generation from agricultural and construction industries is growing at an upsetting rate that causes a heavy burden on landfill facilities. For these reasons, there is a pressing need for reusing and recycling waste materials for use in concrete production. This study assesses the potential use of rice husk ash (RHA) and reclaimed asphalt pavement (RAP) which are originating from agro-industry and road sector respectively as promising construction materials. To this end, the study characterizes these materials and investigates their effects on the fresh and hardened properties of concrete. An effort has been also made to develop an artificial neural network (ANN) model for predicting the 28th-day compressive and tensile splitting strengths of concrete containing RHA and RAP. The materials were characterized in terms of gradation, fineness modulus, water absorption, specific gravity, bulk density, void space between aggregate particles, chemical analysis, aggregate crushing and impact values. The wet and hardened concrete properties were assessed by partially replacing cement and virgin aggregates with RHA and RAP, up to 20% (by mass of cement) and 50% (by mass of aggregates), respectively. A total of 22 mixes were prepared and studied, twelve of which were devoted to studying the collective effects of RHA and RAP on the properties of concrete. ANN model was then developed in MATLAB version R2017a using the predetermined attributes of concrete constituent materials. The study results show that RHA and RAP decrease slump, compacting factor, density, absorption and sorptivity. RHA increases compressive and tensile splitting strength, whereas RAP decreases compressive and tensile splitting strength. Comparable strength, favourable absorption and sorptivity vi vii values were obtained when 15% RHA was combined with up to 20% RAP in the concrete mixes. The developed ANN model predicted the compressive and tensile splitting strength with prediction error values of 0.648 and 0.072 MPa respectively. The compressive strength (fc) was overpredicted on average by 0.123 MPa, whereas the model underpredicted the tensile splitting strength (fts) by 0.019 MPa. The predicted compressive and tensile splitting strength deviated on average by 2.088 and 2.905 % respectively from experimental results. The ANN model was successful in predicting the compressive and tensile splitting strength results accurately. On this basis, it is recommended to use RHA and RAP inclusive concrete for general concreting works. Further research could be undertaken to evaluate economic, social and environmental benefits of using RHA and RAP as concrete ingredients.