Evolutionary Optimization of Neural Networks for Estimating Mechanical Properties of Lightweight Aggregate Concretes

Leonardo Goliatt¹; Michele Farage¹; Jonata Jefferson Andrade¹; Lucas Neves Silva¹; Rogerio Santos¹

¹ UFJF

doi:10.20906/CPS/CILAMCE2017-0194

Resumo

In this paper, neural networks with parameters adjusted by a Particle Swarm Optimization algorithm is used to predict mechanical properties of lightweight aggregate concretes. Unlike the approaches found in the literature, the proposed procedure estimates simultaneously the compressive strength and elasticity modulus. These properties were modeled as a function of four variables: water/cement fraction, lightweight aggregate volume, cement quantity and lightweight aggregate density. A Particle Swarm Optimization algorithm performs the model selection and automatically tunes the number of neurons in the hidden layer and the activation function. The results are compared with a model selection based on exhaustive search on the parameter space. The proposed approach arises as an alternative and competitive tool for estimating simultaneously the mechanical properties of lightweight aggregate concretes.

Palavras-chave: Extreme Learning Machines; lightweight aggregate concretes; Particle Swarm Optimization