Prediction of Splitting Tensile Strength of Self-Compacting Recycled Aggregate Concrete Using Novel Deep Learning Methods

The composition of self-compacting concrete (SCC) contains 60–70% coarse and fine aggregates, which are replaced by construction waste, such as recycled aggregates (RA). However, the complexity its structure requires a time-consuming mixed design. Currently, many researchers studying prediction properties using soft computing techniques, will eventually reduce environmental degradation other material waste. There have been very limited contradicting studies regarding different ANN algorithms. This paper aimed to predict 28-day splitting tensile strength SCC with RA artificial neural network technique comparing following algorithms: Levenberg–Marquardt (LM), Bayesian regularization (BR), Scaled Conjugate Gradient Backpropagation (SCGB). algorithms, so total 381 samples were collected from various published journals. input variables cement, admixture, water, superplasticizer; data randomly divided into three sets—training (60%), validation (10%), testing (30%)—with 10 neurons in hidden layer. models evaluated mean squared error (MSE) correlation coefficient (R). results indicated that all optimal accuracy; still, BR gave best performance (R = 0.91 MSE 0.2087) compared LM SCG. was model for predicting TS at 28 days RA. sensitivity analysis cement (30.07%) variable contributed most RA, water (2.39%) least.