Experimental analysis and predictive modelling of the mechanical properties of particulate composite of Newbouldia laevis plant fibre using artificial neural networks (ANN)

The use of plant biomass as a substitute for synthetic materials in polymer composites is currently attracting significant research interest due to the negative health and environmental impacts associated with synthetic materials. However, inadequate material design hampers the performance and application of these biocomposites. In this study, fibre biomass from the Newbouldia laevis plant stem was processed and utilized to fabricate composites with varying particulate contents. The composites underwent mechanical testing, and the experimental data were analyzed and trained using artificial neural networks (ANN) to develop an accurate predictive model for enhanced material design. The calculated mean absolute percentage error (MAPE) for tensile, flexural, compression, and impact strengths was 4.4%, 3.6%, 8.7%, and 8.5%, respectively, indicating a high level of accuracy in the predictions. In comparison, the MAPE from the linear regression model was 10.9%, 5.9%, 10.4%, and 19.7% for tensile, flexural, compression, and impact strength, respectively, demonstrating that the ANN prediction model is more accurate and reliable. Therefore, for biocomposite research and product development, it is recommended that ANN be employed for optimizing material design to help reduce both the time and cost of research.