•  
  •  
 

CIB Conferences

Abstract

Accurate estimation of the compressive strength of rice husk ash (RHA) concrete is essential for the effective use of this sustainable material in structural applications. In this study, a multilayer perceptron (MLP) model is developed using an experimental database comprising a wide range of RHA replacement levels, mixture parameters, and curing periods. The predictive capability of the model is systematically evaluated through repeated random data partitioning, showing stable performance and good generalization across different training–testing scenarios. To gain insight into the underlying prediction mechanism, a permutation-based sensitivity analysis is performed to quantify the contribution of each input variable to model accuracy. In addition, two-dimensional partial dependence plots (PDP 2D) are employed to further explore the interactive effects between key mixture variables, providing a more comprehensive understanding of their combined influence on compressive strength. The analysis reveals that curing duration, water content, and binder-related variables exert the strongest influence on compressive strength, whereas the effects of other mixture parameters are comparatively less pronounced. By combining reliable prediction with a transparent assessment of variable importance, the proposed MLP-based framework provides a practical and robust tool for strength prediction and data-driven optimization of sustainable RHA concrete mixtures.

Keywords

Multilayer perceptron, Rice husk ash concrete, Strength prediction, Permutation sensitivity analysis, Sustainable cementitious materials

Download

Share

COinS