Comparative study of flexural properties prediction of Washingtonia filifera rachis biochar bio-mortar by ANN and RSM models

Abstract

The production of environmentally friendly and sustainable materials from natural materials is growing considerably. It is therefore important to use materials that capture and store CO2. The valorization of natural wastes in the field of construction can be a good alternative to natural aggregates in concrete and mortar. Many researchers have investigated on developing biochar applications and the potential use of biochar resulting from biomass pyrolysis to reduce the carbon footprint of based cement construction materials. The local accessibility and low cost of environmental waste lead to its incorporation as a substitute for cement in a form of biochar, which is highly suggested in many applications such as pavement concrete, coating, joint mortar and non-structural concrete elements. Thus, this study investigated the possibility of utilizing waste from pyrolysis of the Washingtonia filifera rachis biochar (WFRB) at temperatures of 300 °C, 400 °C, and 500 °C in cementitious mortars as a substitute for cement. For this purpose, different mortar mixtures were developed as cement substitutes (1%, 2%, 3%, 4% and 5% WFRB). Flexural strength tests on the samples manufactured through the addition of biochar in the matrix were carried out as well as SEM analysis of the bio-mortar in order to know their internal structure. Statistical processing of the results using response surface method (RSM) and artificial neural networks (ANN) on developed cement bio-mortars was performed to find optimal addition WFRB. It was found that cement substitution of 1% pyrolysis WFRB at a temperature of 500 °C, is a promising combination offering better mechanical performance using RSM and the WFRB500-1% samples showed increases in flexural strength, displacement and flexural modulus of 5%, 20% and 265% respectively compared to the reference mortar. Model RSM as well as ANN correlate well with those obtained experimentally. However, the performance of the ANN model is much more adequate. ANN model showed a much more accurate prediction than RSM in terms of the values R2 and RMSE.