Experimental thermophysical dependent mechanical analysis of earth bricks with Canarium schweinfurthii and Cocos nucifera bio-aggregates - A case study in Cameroon

Abstract

AbstractIn order to meet sustainable development goals, global policies are strongly oriented towards the recovery of local materials such as agricultural waste. In this context, a new biosourced compressed earth brick (CEB) has been produced with bio-shell aggregates of Cocos nucifera (CN) and Canarium schweinfurthii (CS), added in equal proportions for various mixtures. The main objective is to study, on one hand, the effect of CN and CS aggregates on the mechanical behaviour and the thermo-physical properties of the CEB. On the other hand, durability of CEB was investigated in terms of hydric stability, abrasion resistance and hygro-thermo-mechanical behaviour. Mixtures were prepared using earthen material and 0, 5, 10 and 15 wt% CNCS, with 8 wt% cement CEM II-B-LL 42.5, respectively. Wet and dry compressive strengths are evaluated versus compaction pressure (2.5, 5, 7 and 10 MPa). The increase of CNCS aggregate content induced a reduction of the compressive strength in both wet a\nd dry situations whatever the compaction pressure. The samples with 5 wt % CNCS and 8 wt % cement, however, offered performances close to reference samples without CNCS aggregates. The reverse trend was found for thermal conductivity: thermal conductivity of the bio-sourced CEB is below 1 W·m−2·K−1 which means good insulating performances. It is noted that the modified CEB have relatively low water absorption, which contribute to a good durability. Hygro- and thermo-mechanical tests showed however a quasi-linear decrease of compressive strength when the humidity of the environment increases but also when the temperature rises. At the end, to obtain more accurate properties by prediction, some machine learning models were used.

Keywords

• compressed earth bricks
• Cocos nucifera
• Canarium schweinfurthii
• compressive strength
• durability
• thermal conductivity and machine learning models