A novel optimised self-learning method for compressive strength prediction of high performance concrete

Publication Type:
Journal Article
Citation:
Construction and Building Materials, 2018, 184 pp. 229 - 247
Issue Date:
2018-09-30
Metrics:
Full metadata record
Files in This Item:
© 2018 Elsevier Ltd Concrete strength (CS) is one of the most important performance parameters that are crucial in the design of concrete structure. The reliable prediction of strength can reduce the cost and time in design and avoid the waste of materials caused by a large number of mixture trials. In this study, a novel predictive model is put forward to predict the CS of high performance concrete (HPC) using support vector machine (SVM) approach, which has benefits of nonlinear mapping, high robustness and great generalisation capacity. In the proposed model, the input variables include the contents of water, cement, blast furnace slag, fly ash, super plasticiser, coarse and fine aggregates and curing age, which produces the CS of HPC as the output. In order to improve the model performance, a type of enhanced cat swarm optimisation (ECSO) is adopted to optimise the key parameters of SVM. Finally, the model is trained and evaluated using a total of 1761 data records, which are collected from existing literatures. The results indicate that the proposed SVM-based model exhibits better recognition ability and higher prediction accuracy than other commonly used models, and it can be considered as an effective method to predict the CS property of HPC in infrastructure practice.
Please use this identifier to cite or link to this item: