Multivariate regression approaches to predict the flexural performance of cellulose fibre reinforced extruded earth bricks for sustainable buildings

Abstract

Regression analysis is commonly used to predict the compressive strength of soil-based materials to reduce the time and cost of construction projects. This study presents the results of multivariate regression analysis on the prediction of modulus of rupture and specific energy of cement-stabilised earth bricks reinforced with bamboo cellulose fibres as a function of cellulose fibre percentage, curing temperature and curing time. Statistical modelling is carried out on the experimental data of the four-point bending test of clay brick stabilised with 10 wt% Ordinary Portland Cement reinforced with 0, 5, 7.5 and 10 wt% organosolv bamboo pulp fibres. The water content of the unreinforced specimen was adopted at 25 wt% of the solid materials based on the plastic limit value of the soil, while the value of the water content of the samples with cellulose fibres was 35 wt% due to the affinity of the cellulose with water as well as to allow good extrudability of the pulp. The bricks were cured under different conditions (23 °C, 60% RH and 60 °C, 100% RH) and at different ages (14 and 28 days). Regression analysis and two-way ANOVA were studied to assess the accuracy, correlation and effect of each variable on the prediction of the random response variable. The results showed that non-linear regression analysis provides the best-fitting statistical model with an accuracy of 96%. In addition, the curing temperature and the percentage of cellulose fibres significantly affect the bricks bending performance, while the effect of curing time is the least visible. This non-linear model can be adopted as a suitable model to predict the flexural properties of cellulose pulp fibre-reinforced earth bricks for a sustainable building solution in developing countries. For better generalization and practical application of this method to predict the flexural performance of cellulose pulp fibre-reinforced earth bricks, a large data set of a broader range should be explored.