Evaluation of fresh and hardened properties of self compacting concrete using machine learning techniques

Abstract

Self-compacting concrete (SCC) is a special concrete which can be placed and compacted under its own weight with little or no vibration effort. Cohesiveness should be maintained without segregation or bleeding. SCC is used to facilitate and ensure proper filling of the formwork and good structural performance in restricted areas and heavily reinforced structural members. The numerous advantages of using SCC include (i) reduction of the construction time (ii) labour cost (iii) elimination of vibration (iv) reduction of noise pollution (v) enhanced compactability in highly congested structural members (vi) good structural performance. It is well known that for evaluation of fresh and hardened properties of various SCC mixes, numerous experiments are to be performed. It is difficult to carry out experiments as it takes considerable amount of time and effort. Meta models (MMs), are a data-driven models that try to emulate the complex input/output behavior of underlying system, by using a limited set of computational expensive simulations. The envisaged scope and objectives of the present investigation include To compile various fresh and hardened properties for different SCC mixes with and without fibres To identify the key variables which will influence the fresh and hardened properties of various SCC mixes To develop models such as SVR, RVM, MARS, GPR and MPMR to predict various fresh and hardened properties of different SCC mixes To validate the developed models Various MMs such as support vector regression (SVR), relevance vector machine (RVM), multivariate adaptive regression splines (MARS), Gaussian process regression (GPR) and Minimax Probability Machine Regression (MPMR) have been proposed to employ for the development of models to predict the fresh and hardened properties for various SCC mixes. The fresh properties include slump flow time, L-box, slump, V-funnel time. The hardened properties include compressive strength, split tensile strength, modulus of elasticity and flexural strength. SVM employs