Influence of the ANN Hyperparameters on the Forecast Accuracy of RAC’s Compressive Strength

Abstract

The artificial neural networks (ANNs)-based model has been used to predict the compressive strength of concrete, assisting in creating recycled aggregate concrete mixtures and reducing the environmental impact of the construction industry. Thus, the present study examines the effects of the training algorithm, topology, and activation function on the predictive accuracy of ANN when determining the compressive strength of recycled aggregate concrete. An experimental database of compressive strength with 721 samples was defined considering the literature. The database was used to train, validate, and test the ANN-based models. Altogether, 240 ANNs were trained, defined by combining three training algorithms, two activation functions, and topologies with a hidden layer containing 1–40 neurons. The ANN with a single hidden layer including 28 neurons, trained with the Levenberg–Marquardt algorithm and the hyperbolic tangent function, achieved the best level of accuracy, with a coefficient of determination equal to 0.909 and a mean absolute percentage error equal to 6.81%. Furthermore, the results show that it is crucial to avoid the use of overly complex models. Excessive neurons can lead to exceptional performance during training but poor predictive ability during testing.

Prediction of the Compressive Strength of Vibrocentrifuged Concrete Using Machine Learning Methods

“…Discussing the results obtained and comparing them with the results previously obtained by other authors, it should be noted that the chosen question is unconditionally novel. The fact is that predicting the properties of concrete was previously known and described in works [6,11,15,17,18,20,21,[23][24][25][26][27][28]30,33,36,40,41,55]. The same study touches upon the topic of predicting the properties of special concretes, that is, variatropic concretes obtained using vibrocentrifuge technology.…”

“…In general, the introduction of machine learning methods allows to save costs on the production of many experimental samples and the procedure for testing them, as well as significantly speeds up the process of obtaining future properties of a concrete composite [25]. Recently, there is an increased interest in prediction of concrete properties using artificial neural networks (ANN) [26][27][28]. For example, in studies [29], researchers note the potential of a machine learning method to determine the structural behavior of reinforced concrete beams, since the forecast data using ANN are consistent with the results obtained using a traditional calculation method developed in accordance with current design codes.…”

Enhancing the Design of Experiments on the Fatigue Life Characterisation of Fibre-Reinforced Plastics by Incorporating Artificial Neural Networks

“…In this case, a number of hidden layers up to 5 is envisaged. The phenomenon of overfitting is also discussed, for example, by Almeida et al [52] who, however, noticed it with a much higher number of hidden layers, more than 30. The ideal number of hidden layers for the problem at hand is determined in this study.…”

The Influence of Materials on the Mechanical Properties of Ultra-High-Performance Concrete (UHPC): A Literature Review

“…Other studies have also investigated the optimal design of the mixture based on advanced statistical analyses, for instance, considering the interaction of the different raw materials on the mechanical properties of the UHPC [49,50]. Machine learning methods can be used to correlate mechanical qualities or durability with material proportions to achieve the optimal concrete mix, particularly for UHPC [51][52][53][54][55]. In practice, this allows a comprehensive understanding of the effect of various materials on the mix.…”